The 2005 SSM Research Series 14

Social Stratification and Social Mobility in

Late-Industrializing Countries 後発産業社会の社会階層と社会移動

Hiroshi Ishida (ed.) 石田 浩 編

March 2008

The 2005 SSM Research Committee

２００５年 SSM 調査研究会

科学研究費補助金 特別推進研究（16001001）

「現代日本階層システムの構造と変動に関する総合的研究」成果報告書

The 2005 SSM Research Series 14

Social Stratification and Social Mobility in

Late-Industrializing Countries 後発産業社会の社会階層と社会移動

Hiroshi Ishida (ed.) 石田 浩 編

March 2008

The 2005 SSM Research Committee

２００５年 SSM 調査研究会

科学研究費補助金 特別推進研究（16001001）

「現代日本階層システムの構造と変動に関する総合的研究」成果報告書

刊行のことば

本書は、文部科学省科学研究費補助金（特別推進研究）「現代日本階層システムの構造と

変動に関する総合的研究」の助成を得て行われた 2005 年社会階層と社会移動調査（SSM 調

査）の研究成果報告書『2005 年 SSM 調査シリーズ』（全 15 巻）のうちの一冊である。

SSM 調査は 1955 年以来 10 年毎に行われている全国調査である。このような継続性を持っ

た社会階層と社会移動に関する調査は世界に類を見ない。もちろんそれぞれの年の SSM 調

査プロジェクトは独自の研究テーマを持っているが、親や本人の階層などの基本変数は継

続的に測定されているので、長期にわたるトレンド分析が可能になる。本シリーズの中に

も、このようなトレンド分析を行っている論文が多数収録されている。

この継続性は SSM 調査の貴重な財産である。2005 年 SSM 調査研究プロジェクトでは、こ

のことを踏まえた上で、新たな方向に踏み出した。それは本格的な国際比較と若年層調査

である。本プロジェクトの基本的なねらいは、次のような問題群に解答を与えることであ

った。グローバリゼーションと新自由主義の進行する中で、労働市場の流動性は高まって

いるのか、それともそうではないのか。また高まっているとすれば、それはどの階層を流

動的にしているのか。特定の階層は保護的制度に守られて流動化していないのではないか。

このような「流動化」と「階層の固定化」という一見すると相反する問題にアタックする

ことが、本プロジェクトの基本的なテーマであった。

このテーマを追求するために、国際比較と若年層調査は不可欠であった。グローバリゼ

ーションと新自由主義はいわば普遍的な変動要因である。ただしこれらは直接的に個々の

社会の社会階層・社会移動に影響を及ぼすのではなく、それぞれの社会のローカルな制度

との相互作用を通じて、社会階層・社会移動に影響を及ぼしたり、及ぼさなかったりする。

また新自由主義や労働市場の流動化に対する人々の評価（これは公共性問題といえよう）

も社会によって異なりうる。これらの問題に答えるためには、国際比較が必要になる。し

かしあまりに異なる社会と日本を比べることは意味をなさない。そこでわれわれは、同じ

儒教文化圏に属し、教育制度も類似しているが、日本よりも早くグローバリゼーションに

さらされている韓国と台湾を比較の対象とした。

労働市場の流動化は若年層にもっとも影響を及ぼすと考えられる。フリーターやニート

の問題をはじめとして、流動化の矛盾は若年層に集中しているといえよう。この問題に関

しては既に多くの研究がなされているが、本プロジェクトでは、SSM 調査の蓄積を活用して、

社会階層と社会移動という視点からこの問題にアプローチすることにした。たとえば、誰

でもフリーターになるわけではなく、出身階層や本人の学歴によってフリーターになる確

率は異なると考えられる。このような社会階層論・社会移動論の道具を用いることで、フ

リーター・ニート問題に新しい光を当てることができるだろう。

このような理論的関心に基づいて、国際比較と若年層調査を行った。国際比較では、韓

i

国と台湾の階層研究者 6 名に研究プロジェクトメンバーとなってもらい、彼ら・彼女らの

全面的な協力の下に韓国と台湾で SSM 調査を実施した。調査票は日本調査とかなりの部分

を共通にして、日本・韓国・台湾で厳密な比較分析が行えるようにした。また産業や職業

の国際比較ができるように、それぞれの社会のデータに国際標準産業分類コードと国際標

準職業分類コードを割り当てた。日本側メンバーにも東アジアの専門家がいて、膨大な時

間を費やしてくれたが、これらの作業は困難を極めた。調査票設計段階の調整から始まり、

調査票の翻訳やバックトランスレーション、調査設計の調整、コーディングにおける無数

ともいえる細かい確認事項などの作業を経て、調査データが完成した。

若年層調査も多くの困難に直面した。大阪大学の太郎丸博氏をヘッドとする若年層調査

タスクグループが実査を担当したが、低い回収率の問題や、郵送調査・ウェブ調査ゆえの

データ・クリーニング、コーディングの難しさがあった。しかし太郎丸氏をはじめとする

タスクグループの献身的な努力により、若年層調査データも完成した。

本シリーズに収録されている論文は、このような調査データの分析に基づいたものであ

る。本プロジェクトでは、8つの研究会からなる研究体制をとって、それぞれの研究会でメ

ンバーが論文構想を報告して相互にコメントをしあい、より良い論文を執筆することをめ

ざしてきた。その成果が本シリーズに集められている。これらの論文を通じて、日本のみ

ならず、韓国と台湾の階層状況に対する理解が深まることを期待する。

本プロジェクトを推進するに当たり、実に多くの方々のお世話になった。あえて一人一

人のお名前をあげることはしないが、ここに感謝の意を表します。また調査にご協力いた

だいた対象者の方々にも心より御礼申し上げます。

２００８年３月 ２００５年 SSM 調査研究会

ii

付記１．本研究会による刊行物のリスト 『２００５年 SSM 日本調査 コード・ブック』 2007 年 11 月 『２００５年 SSM 日本調査 基礎集計表』 2007 年 11 月 ２００５年 SSM 調査シリーズ（研究成果報告書集）（2008 年 3 月刊）

第１巻 第２巻 第３巻 渡邊 勉 第４巻 第５巻 第６巻 第７巻 土場 学 第８巻 第９巻 第 10巻 第 11巻 第 12巻 第 13巻 第 14巻 第 15巻

三 輪 哲

小林 大祐 高田 洋

阿形 健司 米澤 彰純 中村 高康

轟 亮 中井 美樹 杉野 勇 菅野 剛 太郎丸 博 前田 忠彦 有田 伸 石田 浩 佐藤 嘉倫

編 『2005 年 SSM 日本調査の基礎分析

―構造・趨勢・方法―』 編 『階層・階級構造と地位達成』 編 『世代間移動と世代内移動』 編 『働き方とキャリア形成』 編 『教育達成の構造』 編 『階層社会の中の教育現象』 編 『公共性と格差』 編 『階層意識の現在』 編 『ライフコース・ライフスタイルから見た社会階層』 編 『階層と生活格差』 編 『若年層の社会移動と階層化』 編 『社会調査における測定と分析をめぐる諸問題』 編 『東アジアの階層ダイナミクス』 編 『後発産業社会の社会階層と社会移動』 編 『流動性と格差の階層論』

iii

『２００６年 SSM 若年層郵送調査 コ－ド・ブック』 2008 年 3 月 『２００６年 SSM 若年層郵送調査 基礎集計表』 2008 年 3 月 『２００５年 SSM 韓国調査 コード・ブック』 2008 年 3 月 『２００５年 SSM 韓国調査 基礎集計表』 2008 年 3 月 “Taiwan Social Change Survey, 2005: Social Stratification and Social

Mobility in Three Countries, User Guide and Codebook” February, 2008 付記２．文部科学省科学研究費補助金研究組織等 研究課題「現代日本階層システムの構造と変動に関する総合的研究」（16001001） 研究種目 特別推進研究 研究組織

研究代表者：佐藤 嘉倫 （東北大学大学院文学研究科教授） 研究分担者：近藤 博之 （大阪大学大学院人間科学研究科教授） 研究分担者：尾嶋 史章 （同志社大学社会学部教授） 研究分担者：斎藤 友里子（法政大学社会学部教授） 研究分担者：三隅 一百 （九州大学大学院比較社会文化研究院教授） 研究分担者：石田 浩 （東京大学社会科学研究所教授） 研究分担者：中尾 啓子 （首都大学東京都市教養学部教授） （研究協力者については、全リストを第 15 巻に掲載した。）

研究経費（単位 千円） 直接経費 間接経費 総額

平成１６年度 19,700 5,910 25,610 平成１７年度 186,600 55,980 242,580 平成１８年度 29,400 8,820 38,220 平成１９年度 32,700 9,810 42,510

計 268,400 80,520 348,920

研究発表

全リストを第 15 巻に掲載。

iv

Preface

Social stratification and social mobility constitute a central concern in modern sociological inquiry. The study of intergenerational mobility, in particular, represents one of the most sustained traditions in sociological investigation. The accumulation of research in this field is centered around the intellectual community known as RC28, that is, Research Committee 28 on Social Stratification and Mobility of the International Sociological Association. The RC28 community contributed to the advancement of the study of social mobility in the world by identifying important research questions, stimulating the innovation in analytic technique, and promoting large-scale cross-national comparative research. Notable examples of comparative research on social mobility include The Constant Flux: A Study of Class Mobility in Industrial Societies, written by Robert Erikson and John Goldthorpe, and Social Mobility in Europe, edited by Richard Breen. The authors of the chapters in this volume are all associated with the RC28 community, and we owe a great deal to the researchers in this community. We build on the knowledge accumulated by previous accomplishments. We use the same class schema and educational coding system and reply on similar analytical methods for the analysis of mobility tables. With respect to the countries studied in the volume, we depart from the previous work. Our nations, late-industrializing nations, were not systematically studied in the previous comparative work. The motivation for this study was to throw light on the late-developers and compare the experience of mobility among nations with different path of industrialization. This volume grew out of the project on social mobility in late-industrializing nations that was part of the 2005 SSM research enterprise. The requirements to join the project were the availability of the large-scale national surveys that are suited for the analysis of intergenerational mobility and the familiarity with the analytic framework and methods of mobility studies carried out in the past. My connection with Columbia University where I had appointment in the 1990s was instrumental in attracting people to this project. I was fortunate to find scholars trained at Columbia, Carlos Costa-Ribeiro and Florencia Torche, who shared interests in this project. Through my Columbia connection, I was also able to recruit Kuo-hsien Su. The RC28 committee proved to a useful connection as well. The participants of the RC28 meetings, Jongchun Cha, Hyunjoon Park, Wei-hsin Yu, and Xiaogang Wu, joined the project. In the end, this project of social mobility in late-industrializing nations was able to gather researchers from Brazil, Chile, China, Japan, Korea, and Taiwan. The

v

Japanese, Korean and Taiwanese teams are associated with the 2005 SSM project and use the surveys conducted by the 2005 SSM project in their respective nations. We had three workshop meetings in Tokyo: 19-20 February, 2005, 17-18 February, 2006, and 15-16 December, 2007. In addition, some of the members were able to hold an informal meeting on 28 May, 2007, the day after the RC28 meeting in Brno, Czech Republic. The chapters included in this volume are based on the findings reported in these meetings. In addition to the project members who wrote chapters in this volume, there were other scholars who contributed to our project. Celi Scalon, then the University Research Institute of Rio de Janeiro (IUPERJ), was part of the Brazilian team at the initial stage of the project. Yu Xie of the University of Michigan took part in the workshop in February, 2006 and offered valuable technical and substantive advice. Richard Breen of Yale University, Mike Hout of the University of California, Berkeley, and Robert Mare of the UCLA inspired us with good examples when they presented their work in Tokyo in March 2006. Yoshimichi Sato of Tohoku University and many others from the 2005 SSM project participated in the workshops and offered useful comments. Their inputs are highly appreciated.

March 2008 Hiroshi Ishida

vi

The 2005 SSM Research Series/ Volume 14

Social Stratification and Social Mobility in Late-Industrializing Countries

Edited by Hiroshi Ishida

CONTENTS

Preface to the 2005 SSM Research Series Preface to Volume 14 1 Trends in Intergenerational Class Mobility and Education in Japan

Hiroshi Ishida and Satoshi Miwa 1 2 Intergenerational Mobility Patterns in Taiwan: The Case of a Rapidly Industrializing Economy

Wei-hsin Yu and Kuo-hsien Su 49 3 Trends in Intergenerational Social Mobility across Four Cohorts in South Korea

Hyunjoon Park and Jongchun Cha 79 4 Trends in Class Mobility and Education in China

Xiaogang Wu 105 5 Social Mobility Trends and Education in Brazil

Carlos A. Costa-Ribeiro 133 6 Social Mobility and Education in Contemporary Chile

Florencia Torche 177

Publication & Presentation List 206

vii

Trends in Intergenerational Class Mobility

and Education in Japan*

Hiroshi Ishida and Satoshi Miwa

(University of Tokyo)

Abstract This paper examines change and stability in the pattern of intergenerational class mobility

in Japan in the late 20th century. The first set of analysis examines the trends in social fluidity. Japanese economy experienced high-speed growth in the 1960s and early 1970s, followed by a recession and sustained economic growth until early 1990s when serious recession hit the country. By 2005 the country slowly moved out of the phase of recession. Despite these fluctuations in the economy, there seems to be stability in the pattern of association between class origin and class destination. The uniform difference model applied to mobility tables created from six surveys does not show any systematic trend. There is no clear tendency towards greater openness in post-war Japan, contrary to the prediction of the industrialism thesis. We then analyze more detailed pattern of mobility by using the core social fluidity model of the CASMIN project. The findings suggest no systematic change in the core social fluidity model across six surveys. Finally, birth cohort and age group are replaced with survey year to examine further trend in the data. Although there seem to be some fluctuations by birth cohorts and age groups, there is no noticeable trend and the dominant pattern is the stability in the pattern of association between origin and destination in the late 20th century Japan.

The second set of analysis focuses on the role of education in intergenerational mobility. It asks whether education mediates the process of intergenerational inheritance and mobility. Our findings suggest that education mediates the effect of class inheritance of the service class (I+II) and the routine non-manual class (III), but does not affect the inheritance effect of other classes. Education also mediates the hierarchy effect and the positive affinity effect. This mediating role of education in social fluidity does not seem to change across survey years. The stability in the role of education is probably explained by the fact that both the pattern of association between class origin and education and the pattern of association between education and class destination after controlling for origin have not changed in the post-war period. Key words: intergenerational class mobility, education, trends

1. Introduction A lively debate on the Japanese distinctiveness has been concerned with reference to various features of Japanese society, such as family and kinship, attitudes and consciousness, industrial management and labor relations, or educational system (e.g., * Paper presented at the workshops on Social Mobility in Late-industrializing Nations, held at the University of Tokyo, Tokyo, Japan.

－ 1 －

Nakane 1970; Dale 1986; Koike 1988; Ishida 1993; Sugimoto 1997). This study will concentrate on the further issue of social mobility, that is, the patterns of intergenerational class mobility in contemporary Japan. Theories of industrial society claim that an industrial technology and economy has profound influence on social structure and process (Kerr et al 1960; Kerr 1983). Industrialism is believed to have brought about the range of changes including the pattern of social mobility. This study will examine whether the pattern of social mobility and its relationship to education are shaped by the forces of industrialization in post-war Japan. The first part of this study will analyze empirical data on the trends in social mobility in post-war Japan in order to verify the hypotheses about long-term trends in mobility among industrial nations. The second part of this study focuses on the role of education in social mobility. It will examine the relationship between class origin and education and between education and class destination by presenting the results of cross-temporal comparisons of five surveys conducted in post-war Japan.

2. Japanese Industrialization Process Japan has experienced dramatic changes in social structure in the 20th century. Figure 1 shows the changes in the industrial distribution of the labor force in Japan in the 20th century. In 1920 over 50 percent (55%) of the labor force was engaged in primary production, and since then its proportion declined gradually until the Second World War. However, the agricultural population grew immediately after the war because many men who returned from the battle fields went back to the original farming villages, and the destruction of industries in urban areas and food shortage forced people to move back to farming areas. The trend was quickly reversed by the end of 1940s when the flux of population began moving from the farming areas to cities. The movement out of the farming took place rapidly, and by 1960 only a third of the labor force was in the primary production and by 1970 less than 20 percent (19%) engaged in the primary production. The farming population continued to decline until the 1990s when the proportion reached six percent in 1995. If we use the declining proportion of the primary production as an indicator of the industrialization, Japan experienced the rapid course of industrialization in the late-1950s and 1960s. The trends in the proportion of the secondary sector where the manufacturing industries are concentrated show the impact of the War. Industrial production steadily increased until the beginning of the War in 1941, and more than a quarter (26%) of the labor force was engaged in the secondary production. However, industrial production was quickly reduced due to war-time destruction. It was not until mid-1950s when the industrial production recovered to the level of pre-war. By 1955 the proportion of the labor force in the secondary sector was 23 percent. In response to the outflow of population from the farming areas in the late 1950s and 1960s, the proportion of people engaged in the secondary production increased to 34 percent in 1970 after which the rate stabilizes. The tertiary sector in Japan increased steadily in the post-war period (1945-2005), but the increase is most dramatic in the early part until 1975. By 1975, the majority (52%) of the labor force was engaged in the tertiary sector. Japan’s industrial structure in the late-1970s already resembles that of other industrial countries. However, the increase in the tertiary sector was not accompanied by the substantial

－ 2 －

decline in the secondary sector, unlike the experience of many western nations (Cole and Tominaga 1976). The process of Japanese industrialization is characterized by almost simultaneous expansion of the secondary and tertiary sector. In other words, the declining farming population is accompanied by the expansion of both blue-collar and white-collar employment. These features are related to the Japanese experience of the late and rapid course of industrialization. The process of economic growth in Japan, however, did not follow a linear pattern. Figure 2 shows the trend in the rate of economic growth during the post-war period. The figures represent the changes in the Gross Domestic Products (GDP). The post-war economic development can be divided into three distinct stages: (1) high economic growth period (1955-1973), (2) low-growth period (1974-1990), and (3) economic recession period (1991-2004). Beginning in the late 1950s, Japan achieved a rapid and substantial economic expansion until November 1973 when the oil crisis hit Japan. The GDP growth rate during this period averaged 9 percent, exceptionally high growth. It is in this period when most dramatic changes in the industrial structure happened in Japan. Despite the oil shock, Japan quickly recovered and entered the stable and sustained low-growth period until 1990. The average GDP growth rate was 4 percent. During the 1990s and early 2000s, Japan experienced the worst recession in the post-war period, and the average GDP increase was only one percent. These economic indicators suggest that the pace of the Japanese economic development during the post-war period differs among the three stages: the accelerated fast growth during the first stage, the stable economic expansion during the second stage, and finally the stagnated period during the last stage. Another aspect of societal transformation which is relevant to our study is the expansion of the Japanese educational system. Figure 3 shows the trend of expansion in three levels of education: the proportion of those who attended minimum compulsory education of primary school, the proportion of those who advanced to higher level of education after the compulsory education, and the proportion of those who were enrolled in the institutions of higher education. The spread of the primary education is already over 90 percent at the turn of the century in 1905. The advancement rate beyond the compulsory education jumped after the introduction in 1947 of the new educational system modeled after the American comprehensive educational system. The compulsory education was extended from six years of primary school to nine years of primary and middle schools, and there was an upgrade of the minimum level of education. The advancement beyond the compulsory level escalated during the 1960s, and the expansion of the higher education system took place in the 1960s and early 1970s and again beginning in the late 1990s. Different stages of the educational system expanded at varying time points, and the process of the educational expansion followed a non-linear pattern.

3. Hypotheses about Change and Stability in Intergenerational Mobility A number of theories and hypotheses have been advanced about the trends in intergenerational class mobility among industrial nations. We present five hypotheses or predictions about the long-term trends in class mobility. These hypotheses are not always written explicitly in the works of scholars cited below, but can be derived from their various writings (see, Breen 2004; Erikson and Goldthorpe 1992b; Goldthorpe

－ 3 －

1985b; Vallet 2001 for further discussions on these hypotheses). The first hypothesis is called a “threshold hypothesis.” This hypothesis argues

that intergenerational mobility rates increase dramatically when a society transforms from the “pre-industrial” to “industrial” stage. An historical increase in mobility rates is observed when a society reaches a level of industrial maturity (Lipset and Zetterberg 1959; Davis 1962). Urbanization and massive migration from the farming to the urban areas lead to a sudden increase in the intergenerational mobility. Japan experienced rapidly increasing rate of mobility following the transition from a feudal society to an industrial society in Meiji Japan (Mitani 1977) and also at the time of rapid economic development of the 1950s and 1960s (Yasuda 1971; Tominaga 1992). According to this hypothesis, we expect a dramatic increase in absolute mobility rates (especially total mobility and outflow rates) in the 1950s and 1960s when Japan turned into a truly mature industrial society. The second hypothesis, which derives from the work of industrialism thesis, predicts the “continuous” increase in mobility rates and the weakening of the association between class origin and class destination, as societies experience industrialization (Blau and Duncan 1967; Bell 1973; Treiman 1970, 1990; Treiman and Yip 1989). The industrialism thesis predicts that all industrial societies converge towards a more mobile and fluid society, exhibiting an increased rate of both absolute mobility (such as total mobility and outflow rates) and relative mobility (or social fluidity). According to this hypothesis, there is a positive correlation between the level of industrialization and the openness of societies. This prediction is based on the notion that the principle of allocation of labor moves from particularistic criteria of social background to universalistic criteria of educational attainment (Parsons 1951; Levy 1966). As the level of industrialization progresses, jobs are more likely to be allocated by people’s ability and achievement than by family background. Industrialization promotes a meritocratic form of social selection, and produces greater fluidity and openness in industrial societies. American and Japanese scholars who subscribed to the modernization theory (Jansen 1965, Ward 1967, and Dore 1967) claimed that the Japanese society has transformed and caught up to the Western nations in the 1950s and 1960s because of the increased use of achievement in allocating people to positions. As a strong believer of the industrialism thesis, Tominaga (1979, p.63) argued that Japan experienced a “rapid and consistent increase” in mobility rates during the post-war period, and that Japanese society became increasingly open in the 1950s and 1960s. As shown in Figure 2, the economic growth was particularly marked in the postwar period until the oil crisis in 1973. Therefore, the second hypothesis predicts a continuous increase in relative mobility rates especially in the period of rapid economic expansion until the mid-1970s in Japan. The trend of openness should be weakened during the time of economic recession during the 1990s and 2000s. The third hypothesis, in contrast, emphasizes “stability” in trends of intergenerational mobility. In his classical study of social mobility, Sorokin (1959) claims that there are short-term fluctuations of mobility rates which reflect historical events and contingencies. However, over the long term, there is stability and “no perpetual trend in the fluctuations” (Sorokin 1959, p. 63). Featherman, Jones and Hauser (1975), and more recently, Erikson and Goldthorpe (1992b) take up a similar position. Although there are changes in occupational and class structures following industrialization, the strength and pattern of association between class origin and class destination remain remarkably stable in industrial societies. Class origins continue to

－ 4 －

affect the allocation of class positions in a similar way over time. Therefore, the third hypothesis predicts that there is a cross-temporal stability in relative mobility rates in post-war Japan (see also Kojima and Hamana 1984; Kanomata 1987, 1997; Imada 1989, 1997; Seiyama et al 1990; Hara and Seiyama 1999; Ishida 2001). The fourth hypothesis is called the “post-industrial rigidity” thesis, and claims that there is an increased rigidity in class structure in post-industrial Japan. This hypothesis became increasingly popular among the media and the scholarly after the collapse of the bubble economy in the late-1980s. Toshiaki Tachibanaki (1998), an award-winning economist, argued that Japan’s income inequality has increased greatly in the late 1980s and that Japan has become one of the most unequal nations in the world. Toshiki Sato (2000), a sociologist, claimed that there was an increased rigidity in the upper non-manual class, or what he called the intellectual elite, in the 1990s and that the intergenerational barriers to entry into the upper non-manual class became increasingly high. As shown in Figure 2, the Japanese society entered the recessionary period beginning in the 1990s, and both the media and the scholarly work claim the increased economic gaps and inequality in Japanese society (see, for example, Ohtake 2005; Shirahase 2005, 2006). Therefore, the post-industrial rigidity hypothesis predicts a trend of increasing rigidity and decreasing openness in Japan beginning in the 1990s. Finally, the fifth hypothesis pertains to the “historical institutional” hypothesis which focuses on the impact of historical institutional changes on mobility rates. The most dramatic changes in the Japanese social structure took place immediately after the defeat in the Second World War. The American Occupation Forces introduced a number of social policies which are likely to have had impact on social structure and processes. The policies were introduced for the purpose of democratizing Japanese society. The dissolution of financial cliques (big business groups) and the land reform which distributed pieces of land to peasants should have increased fluidity in the society and reduced the reproduction of the top owners. Therefore, according to the historical institutional hypothesis, we should expect to see increased fluidity among those who experienced the post-war social policies. These five hypotheses will be examined using empirical data of intergenerational class mobility in post-war Japan. Some of the hypotheses are not necessarily incompatible with each other. It is possible that Japan experienced a historic increase in observed mobility rates in the 1950s (the threshold hypothesis) and, at the same time, an increased rigidity in the 1990s (the post-industrial rigidity hypothesis). However, the second “continuous” hypothesis and the third “stability” hypothesis are not compatible to each other.

4. Hypotheses about Class Origin, Class Destination, and Education

The second part of this study examines the relationship among class origin, class destination, and education. In particular, it will empirically investigate the effect of class origin on the attainment of educational credentials, the effect of education on the allocation of individuals to different class destinations, and the role of education in intergenerational mobility. The analyses include cross-temporal comparison of six national surveys conducted in post-war Japan. First, we will focus on the studies regarding the effect of class origin on educational attainment. The industrialism thesis predicts that particularistic criteria,

－ 5 －

such as class origin, play a diminishing role in educational and socio-economic attainment among industrial societies (Parsons 1951; Levy 1966; Blau and Duncan 1967; Bell 1973). Because of an increased urbanization and development of free mass education of industrialized nations, the dependence of educational attainment upon class origin is reduced (Treiman 1970). Critics of the industrialism thesis, however, point out that socio-economic as well as cultural obstacles to equality of educational opportunity persist in industrial nations (e.g., Bowles and Gintis 1976). The educational success of children is often dependent upon familial well-being and the cultural environment of their home (Featherman and Hauser 1978; Halsey et al. 1980). The advocates of cultural capital theory argue that children from advantaged cultural backgrounds benefit more from their schooling because they are already equipped with the linguistic and cultural competence which is necessary to succeed in school (Bourdieu 1973; 1974; Bourdieu and Passeron 1977; Bernstein 1977). These theories predict that the linkage between class origin and educational attainment is far from being diminished in industrial societies. A number of studies of cross-cohort trends in educational attainment (e.g., Mare 1981; Halsey et al. 1980; Raftery and Hout 1993; Shavit and Blossfeld 1993) also report a finding that is not consistent with the industrialism thesis: a cross-cohort stability or increase in the effects of social origins on school-grade progressions. In Japan studies analyzing trends in the relationship between social background and education report the results that are consistent with those in other nations. The effect of social background on educational attainment appears to be stable in post-war Japan (Ushiogi 1975, 1978; Fujita 1981; Ehara 1984; Kariya 1995; Ishida 2007). Furthermore, Ishida (1993) shows that the overall impact of all social background characteristics on educational progression is not particularly smaller in Japan than in Britain and the United States, despite the prevalence of educational credentialism which seems to imply that the attainment of education is open to any talented individual. Second, we will focus on the association between education and class destination among industrial societies. The industrialism thesis (Kerr et al. 1960; Treiman 1970; Treiman and Yip 1989; Hout 1989) claims that individuals who attain higher level of education are placed at the top of the occupational hierarchy. It also predicts an increasing effect of education on the allocation of individuals to class positions as society continues to industrialize. According to the functionalist account (Davis and Moore 1945), these highly educated people perform important functions in the society which cannot be fulfilled by other people. It predicts a positive correlation between educational level and desirability and prestige of positions in the society. This positive correlation, however, does not need to come from a functionalist explanation. The highly educated possess an enhanced ability and resources to obtain the most desirable positions in the society. They could restrict entry into the highly desirable positions, such as professional jobs, to those who possess appropriate certifications (Collins 1979; Parkin 1979). Studies dealing with the relationship between education and labor market outcomes (occupational status, income and other rewards) are too numerous to report. The human capital theory (Mincer 1974; Becker 1975), for example, views education as the investment in human capital and predicts a positive correlation between years of schooling and wage. Within the framework of class analysis, the most comprehensive and up-to-date study is that of Shavit and Muller (1998). They try to explain cross-national variation in the effect of educational qualification on entry class and entry

－ 6 －

occupational status by institutional features of industrial societies. Ishida (1993) also shows that the effects of educational credentials on occupational status and income are not particularly greater in Japan than in Britain and the United States. The results, according to Ishida, are evidence against the educational credentialism thesis which imply greater socio-economic returns to education in Japan. Third, we will examine the role of education in intergenerational class reproduction and mobility. We consider the three pairs of association between class origin and destination, between class origin and education, and between class destination and education at the same time, and determine how much of the association between class origin and class destination is mediated by education. Previous research reports that the extent of mediation by education depends on the types of class reproduction and mobility (Robinson and Kelly 1979; Kerckhoff, Campbell, and Trott 1982; Robinson 1984; Yamaguchi 1983; Ishida, Muller, and Ridge 1995). The reproduction of ownership of production (such as farming and petty business) is likely to go through the direct transmission of physical capital without the means of education, but the reproduction of control over labor is likely to be facilitated by educational achievement. By representing the association between class origin and class destination by different effect matrices, we will evaluate whether education uniformly affects different effect matrices. The role of education may turn out to be very different depending on the kinds of association between origin and destination, represented by effect matrices. Three themes discussed above will be examined within the context of change. We will examine whether the linkage between class origin and education and between education and class destination has changed in postwar Japan. The examination of the trend in the effect of class origin on educational attainment in post-war Japan provides an opportunity to empirically verify the prediction of the industrialism thesis that the effect of class origin weakens as society continues to industrialize. The persistence and stability in the association between class origin and education will be consistent with the cultural capital argument. Similarly, the trend analysis of the effect of education on class destination will empirically evaluate the prediction of strengthening the linkage between achieved level of education and class positions in industrial nations.

We will also examine whether the mechanism through which class origin affects class destination has changed during the post-war period in Japan. In particular, we will evaluate whether the mediating role of education has changed in Japan. The effective use of education in the reproduction of the professional-managerial class, for example, might have strengthened in the post-war period due to the increased importance of education in allocating people to class positions. As the tendency for the association between education and class position to be strengthened, the mediating role of education might have strengthened.

5. Data and Variables The Japanese data sets used in this study are derived from the Social Stratification and Social Mobility (SSM) surveys conducted in Japan. These surveys were conducted in 1955, 1965, 1975, 1985, 1995, and 2005 with virtually same questions on the core items, such as the respondents’ employment conditions and social background. These surveys, therefore, provide us with the unique opportunity to conduct cross-temporal comparisons using virtually identical variables. The age range is set to 30 to 64, so that the analyses

－ 7 －

include only those respondents who have completed their educational attainment. We also restrict analysis to men because female respondents were excluded in the SSM surveys prior to 1985. This study uses three variables: class origin, class destination and education. Class origin refers to the class of the respondent’s father when the respondent was growing up. 1 Class destination refers to the respondent’s current class. In order to determine class categories, the following four questions are used: occupation, employment status, managerial status, and firm size. The class schema is shown in Table 1. 2 Our analyses are based on the six-category version of Erikson-Goldthorpe- Portocarero class schema (Erikson, Goldthorpe, and Portocarero 1979): the professional managerial class or the “service class” (I+II), the routine non-manual class (III), the urban petty bourgeoisie (IVab), the farming class (IVc+VIIb), the skilled manual class (V+VI), and the unskilled manual class (VIIa). Education refers to educational credentials which include both academic and vocational qualifications. Konig, Luttinger, and Muller (1988) describe the procedure for constructing a comparable educational classification across countries. We use three levels of education: the lowest is defined as the social minimum of elementary education (Low); the middle level includes higher levels of vocational qualifications and intermediate academic qualifications (Middle); and the highest level includes all tertiary qualifications (High). 3 We will use the terms education and qualifications interchangeably throughout the paper.

6. Trends in Absolute Rates of Intergenerational Class Mobility

We first present the trends in the distributions of class origin and class destination across six survey years from 1955 to 2005 (Table 2). The distribution of class destination represents the (male) class structure of the Japanese society at the time of the survey. There are several noticeable trends apparent in changes in the class destination. First, the most striking trend deals with the rapid contraction of the farming class in post-war 1 In the Japanese data sets, slightly different operationalization of the father’s class is used depending on the survey years. The 1955 Survey used the father’s longest employment as the measure of father’s class. The rest of the surveys used the father’s main employment as the measure of father’s class. An additional caution is required in the use of the 1955 Survey. It did not ask the question of managerial status to the respondent nor to the father. In other words, the proportion of the professional managerial class is probably slightly underestimated at the expense of the routine non-manual class because some of the clerical job holders might have lower managerial status. Similarly, the proportion of the skilled manual workers is probably slightly underestimated at the expense of the non-skilled manual workers because some manual workers in class VIIa might hold a foreman status which entitles them to be assigned in class V. 2 For details of class schema, see Erikson and Goldthorpe (1992b, chapter 2). For justification of collapsing the full 10-category version of the class schema, see Ganzeboom, Luijkx, and Treiman (1989). On the use of more disaggregated tables, see Erkison and Goldthorpe (1992a) and Hout and Hauser (1992). On service class, see Goldthorpe (1982). 3 For details of the educational classification scheme, see Konig, Luttinger, and Muller (1988) and Ishida, Muller, and Ridge (1995, appendix). In the Japanese data sets, the first category includes those with the compulsory level of education, the second category high school graduates and old system junior high school graduates, and the third category old system high school graduates, junior college and university graduates. Note that the Japanese classification used in this study differs from that used in Ishida, Muller, and Ridge (1995).

－ 8 －

Japan. The largest drop in the share of the farming class took place from 41 percent in 1955 to 24 percent in 1965, and the share continued to drop until 1985. Second, the skilled manual working class expanded rapidly from 7 percent in 1955 to 15 percent in 1965, and remained stable around 18 percent from 1975 to 2005. Third, the professional-managerial class expanded steadily throughout the post-war period. The share of the professional-managerial class was only 10 percent in 1955, but it doubled to 20 percent by 1965 and has continued to grow, reaching 37 percent by 1995. The professional-managerial class was already the largest class in the class structure by 1975. Judging from these changing shares of various classes, the most dramatic changes in the Japanese labor market probably took place from 1955 to 1965. The rapid migration of population out of the rural area was absorbed into the urban areas. Moreover, what is apparent in the trend of class destination distributions is that there was a simultaneous expansion of the white-collar sector (the professional-managerial class) and the blue-collar sector (the skilled working class) during the rapid industrialization process from 1955 to 1965. In contrast to many other industrial nations that experienced the expansion of the blue-collar sector first and much later that of the white-collar sector, the white-collar sector in Japan began its expansion almost at the same time of the blue-collar sector, largely due to the late and rapid industrialization process. Fourth, the share of the routine non-manual class, the petty bourgeoisie, and the non-skilled working class remained fairly stable across survey years. It is important to notice that there was no sign of declining significance of the urban petty bourgeoisie in the postwar Japanese class structure, at least until 1995. The share of the urban proprietors remained around 20 percent from the 1950s to 1990s. We already know that Japan had the relatively large petty bourgeoisie sector in the 1970s, compared to European nations (Ishida, Goldthorpe and Erikson 1991), and the persistence of this sector across survey years suggests the continued importance of this sector in Japan. However, there is evidence that the urban self-employment has declined from the late-1990s, primarily due to the declining value of the assets during the recession (Ishida 2004). The share of the non-skilled working class remained stable at about 10 percent. This trend suggests that this class never expanded to constitute a demographically significant group in Japan, unlike many industrial nations where the non-skilled working class was at one point in time the major social force within the class structure. The distributions of class origin do not represent the class structure of any given time period since the age of the fathers varies substantially and men without any son never appear in the distributions (Blau and Duncan 1967). Instead, they show how the origins of the respondents in a particular survey year have changed over time. 4 There are changes parallel to those which were observed in the class destination distribution: the contraction of the farming class and the corresponding increase in the shares of the skilled working class and the professional managerial class. However, compared with the class destination distributions, there seems to be a time-lag in the changes in the distribution. The gradual decline of the farming class was observed from 1955 to 2005. There is also a gradual increase in the share of the professional managerial from 1955 to 1995. The share of the skilled working class increased most rapidly from 1975 to 1985. Table 2 reports the total mobility rates for the six survey years. The rate increased dramatically from 1955 to 1965 and continued to increase modestly until 1985. 4 It should be remembered that the operationalization of the father’s class is slightly different in 1955.

－ 9 －

From 1985 to 2000, there is a plateau in total mobility rate at about 68 percent. The trend is closely related to the changes in the class origin and class destination distributions. In 1955, over 60 percent of the fathers were engaged in primary production and about 40 percent of the respondents were in the farming class (IVc+VIIb). A large share of the farming class in both generations implied high intergenerational inheritance. However, rapid contraction from 1955 to 1965 of the farm sector both in the class origin and class destination distributions meant the mobility out of the farming class, and the total mobility rate jumped from 48 percent in 1955 to 60 percent in 1965. This finding is consistent with the prediction by Lipset and Zetterburg about the historic increase in mobility rate when a society enters a mature industrial stage. Trends in the dissimilarity index between class origin and class destination parallel those of total mobility rates. When the dissimilarity index is low, the total mobility rate is low. As the dissimilarity index increases, there is a corresponding increase in total mobility rate. However, the reduction in the dissimilarity index after 1985 does not accompany the reduction in the total mobility rate probably because the farming class has sufficiently shrunk by 1985 in the class destination. Table 3 presents outflow mobility rates which are computed from the 6 by 6 Japanese mobility tables for the six survey years. Features of cross-temporal changes across survey years more or less correspond to the characteristics of the changes in the class destination distributions. Outflows to the farming class (IVc+VIIb) decreased dramatically, especially from 1955 to 1965. On the other hand, outflows to the professional managerial class increased steadily. Outflow rates to the petty bourgeoisie remained at a high level throughout the period. Table 4 presents inflow mobility rates for the six survey years. Trends in inflow rates generally parallel those in outflow rates, but they are much less apparent. Inflows from the farming class have declined, and inflows from the professional managerial class increased. Our particular interests pertain to the outflows from the farming class (see figures in italic in Table 3). The intergenerational stability (that is, the sons of the farming class who themselves became the farming class) is dramatically reduced from 61 percent in 1955 to 40 percent in 1965, following the rapid contraction of the farming class during this period (see distribution of class destination). It is worth noting that these sons of the farming class were found in both the blue-collar sector (the skilled and non-skilled working classes) and the white-collar sector (the professional-managerial class) since the outflows to these classes increased from 1955 to 1965. Therefore, migration out of the farming implied mobility into both the manual working class and the upper white-collar class. Figure 4 shows the changes in the outflow rates from the farming class and the expansion of education. The figure was drawn by dividing our respondents into different birth cohorts and computing rates by cohorts. From Figure 4 we can observe that the outflow rates from the farming to the manual working class (V/VI and VIIa combined) increased as the proportion of those who had more than senior high school education expanded. Similarly, the changes in the outflow rates from the farming to the professional-managerial class parallel the changes in the proportion of those who attended the institutions of higher education. These figures suggest that the outflows from the farming class are probably facilitated by education.

Table 5 presents how education was used to achieve such mobility out of farming. The table disaggregates the outflow rates from the farming class by education level. From 1955 to 1975, there is a dramatic reduction in the intergenerational stability of the

－ 10 －

farming class, especially among those with low education (from 56 percent to 24 percent). In return, the increased proportion of the sons of the farming class with low level of education moved into the manual working class from 1955 to 1975 (from 11 percent to 26 percent). We should bear in mind that the new educational system which extended the compulsory education to middle school was introduced in 1947 in Japan. Therefore, the respondents who were in the lowest level of education (compulsory education) in the 1965 and 1975 survey contained those who completed middle school. Middle schools under the new educational system assisted students (especially from the farm background) find non-farm employment, and the students from the farm background migrated into urban areas on the massive scale (Kariya, Sugayama, and Ishida 2000). The result is the massive outflow of the sons of the farmers with middle school education. After 1975, the sons of the farming class who had senior high school education moved into the manual working class and the professional-managerial class. Education, therefore, appears to have helped the sons of the farming class move out of their class of origin. We have already learned the distinctive feature of the Japanese manual working class in comparison with European nations: a very weak demographic stability or “demographic identity” (Goldthorpe 1985a, 1987; Ishida, Goldthorpe and Erikson 1991). In particular, the Japanese working class is characterized by its low level of intergenerational stability (outflow rate) and its low level of intergenerational self-recruitment (inflow rate), compared with the working class in European nations. The demographic character of the Japanese working class is clearly separated from that of the European working classes (Ishida 2001). It is therefore important to examine whether this feature has changed over the course of economic development in post-war Japan. Table 6 presents outflow rates from and inflow rates to the two manual working classes, that is, class V+VI and VIIa combined. By looking at outflow rates, we find the intergenerational stability of the working class has not changed much: that is, the sons of the working class who were themselves becoming the working class constitute about 45 percent. Although there is a temporal decline in the intergenerational stability in 1995, the level of stability in 2005 remains at 49 percent, the same level observed in 1975 and before. From the inflow recruitment perspective, the percentage of self-recruitment shows some fluctuations. Self-recruitment declined from 22 percent in 1955 to 15 percent in 1975 and then increased from 1975 to 35 percent in 2005. There is a steady declining trend in the share of the farming class of 54 percent in 1975 to 29 percent in 2005. However, the distinctive inflow pattern in Japan, vis-a-vis Europe, is still apparent in 2005. The petty bourgeoisie (IVab) and the farming class account for about the majority (50 percent) of the class origin among the working class. In other words, the recruitment into the working class is still predominantly from the self-employment sector even in 2005. Furthermore, the recruitment into the working class from the white collar class (I+I and III) increased steadily from 6 percent in 1955 to 15 percent in 2005. However, it is worth noting that the self-recruitment increased sharply from 25 percent in 1995 to 35 percent in 2005. The increased proportion implies the increased closure, at least in absolute rate, of the manual working class. In order to examine the role of education in the intergenerational stability and self-recruitment of the manual working class, we present in Table 7 the outflow rates from and the inflow rates into the working class by education and survey year. The examination of the outflow rates shows that the main destinations, aside from staying in

－ 11 －

the working class, were the two self-employed classes of the farming and urban petty bourgeoisie, especially for those sons with minimum education, in 1955 and 1965. However, after 1965, the sons of the working class who had high school education or attended higher education moved into the professional-managerial class. In other words, higher level of education provided avenues of upward mobility for the sons of the working class. The examination of the inflow rates into the working class shows that the sons of the farming class with low level of education constituted the dominant recruitment component until 1975: over 40 percent of the working class came from the farming class with low education. However, as the recruitment from the farming class declined following the shrinkage of the farm origin, the self-recruitment from the sons of the working class gradually increased. Furthermore, as the overall self-recruitment rate increased from 1995 to 2005, the self-recruitment of the sons of the working class with high school education increased sharply from 13 to 25 percent. In other words, high school education did not necessarily imply the means of moving out of the working class background in 2005. In summary, the two distinctive features of the Japanese manual working class, a low level of both intergenerational stability and self-recruitment, were apparent throughout the post-war period. The intergenerational stability of the working class remains at a relatively low level in Japan, and there is evidence that education facilitated the mobility out of the working class. The Japanese working class continues to be recruited extensively from other classes, but the extent of self-recruitment seemed to have increased from 1995 to 2005. Throughout the post-war period, the Japanese manual working class has not shown a fully developed “demographic identity” that is found in many western nations.

7. Models of Association between Class Origin and Class Destination

We employ a series of log-linear and log-multiplicative models to examine the trends in relative mobility. The most fundamental model is called the constant social fluidity model (CSF model). It sets the odds ratio pattern in the mobility table exactly the same across the survey years. In other words, the marginal distributions of the mobility tables, that is, the distribution of class origin and of class destination, may differ across survey years, but the relative mobility rates are set exactly identical across years. Formally, the CSF model may be written as the following multiplicative form: Fijt = η τi

O τjD τt

Y τitOY τjt

DY τijOD, (1)

where Fijt refers to the expected frequency in cell (i,j,t) of the origin by destination by survey year 3-way mobility table, η is a scale term, τi

O is the main effect of class origin, τj

D is the main effect of class destination, τtY is the main effect of survey year, and the

remaining two-way terms (τitOY, τjt

DY, τijOD) indicate the association between origin and

year, destination and year, and origin and destination, respectively. The CSF model does not include the three-way term (τijt

ODY) implying that the association between origin and destination does not differ by survey year. The association between class origin and class destination may be represented by the model of core fluidity. The model is composed of different effects which are intended to capture different aspects of mobility. These effects are informed by

－ 12 －

sociological ideas about the process of intergenerational mobility in industrial nations, and the core model implies that these effects operate in mobility tables constructed from any industrial nation. The original core model is constructed for the 7 by 7 mobility table, but the present study uses the modified version for the 6 by 6 table (Ishida, Muller, and Ridge 1995). The details of the model and the sociological rationale for deriving the model can be found elsewhere (Erikson and Goldthorpe 1987a, 1992b). Different effects of the model are represented by matrices in Figure 5. First of all, there is an inheritance effect. This effect simply refers to the propensity of individuals to stay in their class of origin rather than to move out of it. Each class is allowed to have different propensity of inheritance because social processes which generate inheritance are likely to be different depending on the class in question (Robinson and Kelly 1979; Yamaguchi 1983; Robinson 1984; Grusky and Hauser 1984). For example, the inheritance of the farming class and the petty bourgeoisie often involves handing over the physical capital in the form of land or a factory to the offspring. The inheritance of the professional managerial class (I+II) is often facilitated by economic resources as well as “cultural capital” (Bourdieu 1973, 1974; Bourdieu and Passeron 1977; Bernstein 1977). The second effect is called hierarchy effect and is captured by two matrices (HI1 and HI2) shown in Figure 5. It intends to divide six classes into three hierarchical levels by separating the professional managerial class (I+II) at the top and the non-skilled working class (VIIa) at the bottom of the hierarchy. There is an additional asymmetry in hierarchy 2 (HI2) matrix involving the farming class. The asymmetric assignment tries to capture the change in the status of farming between the two generations. In the father’s generation, the farming class is mostly composed of peasants based on largely subsistence agriculture while in the son’s generation, farming is more commercialized and market-oriented. In order to take into account this transformation of the farming sector, the hierarchy effect assigns the farming class as the class of origin to the least advantaged position in the hierarchy along with the non-skilled working class but assigns the farming class as the class of destination to the middle level of the hierarchy. The third effect refers to positive affinity effects which are meant to capture relatively easy flow of individuals between particular classes. The positive affinity A recognizes exchange movement between the professional managerial class and the routine non-manual class, as forming a “white-collar bloc.” The positive affinity B captures movement involving the two propertied classes (IVab and IVc+VIIb) and the two working classes (V+VI and VIIa). The exchange between the petty bourgeoisie and the farming class arises out of the possibility of transferring capital, and the exchange between the skilled and non-skilled working class is facilitated by the similarity in manual labor forming a “blue-collar bloc.” The positive affinity B also includes two other kinds of movement. The exchange between the professional managerial class and the petty bourgeoisie reflects the fact that some individuals who belong to I+II are owners of professional practices or large business. An additional asymmetry indicating a flow from the farming class to the non-skilled working class recognizes the propensity for the sons of the farmers to engage in non-skilled work when they move out of farming. The core model may be written as a log-linear model expressed in the following multiplicative form: Fij = η τi

O τjD τ(ij)

DIGk τ(ij)HI1 τ(ij)

HI2 τ(ij)AF2A τ(ij)

AF2B, (2)

－ 13 －

where Fij refers to the expected frequency in cell (i,j) of the mobility table, η is a scale term, τi

O is the main effect of class origin, τjD is the main effect of class destination, and

the rest of the parameters represent effect matrices described above. The CSF model may be constructed by using the effect matrices of the core fluidity model. The association between origin and destination may be represented by a series of effect matrices, instead of the full association as described in Figure 5. These effects may be fixed across survey years. The log-linear model representing the CSF effect matrix model may be written as: Fijt = η τi

O τjD τt

Y τitOY τjt

DY τ(ij)DIGk τ(ij)

HI1 τ(ij)HI2 τ(ij)

AF2A τ(ij)AF2B , (3)

where Fijt refers to the expected frequency in cell (i,j,t), η is a scale term, τi

O is the main effect of class origin, τj

D is the main effect of class destination, the two-way terms represent the main effects allowed to vary by year, and the remaining terms represent the effect matrices of the core model. The effect matrices, however, do not vary by year (t). There are national variants of the effect matrices of the core fluidity model. The Japanese variant introduces an additional negative affinity (AF1J) which represents low propensity for the sons of the professional-managerial class to be downwardly mobile into the ranks of the manual working class (see Figure 5 for matrix representation). The model can be written as: Fijt = η τi

O τjD τt

Y τitOY τjt

DY τ(ij)DIGk τ(ij)

HI1 τ(ij)HI2 τ(ij)

AF2A τ(ij)AF2B τ(ij)

AF1J , (4) We could construct a log-linear model in which the association between origin and destination is represented by the same effect matrices but the extent of the effect is allowed to vary across survey years. The log multiplicative form of the equation is the following: Fijt = η τi

O τjD τt

Y τitOY τjt

DY τ(ijt)DIGk τ(ijt)

HI1 τ(ijt)HI2 τ(ijt)

AF2A τ(ijt)AF2B τ(ijt)

AF1J , (5) where Fijt refers to the expected frequency in cell (i,j,t), η is a scale term, τi

O is the main effect of class origin, τj

D is the main effect of class destination, the two-way terms represent the main effects allowed to vary by year, and the remaining terms represent the effect matrices of the core model which are allowed to vary by year (t). Finally, we employ a log-multiplicative model of uniform difference, called “Unidiff model” (Erikson and Goldthorpe 1992b; Xie 1992). This model represents the difference between two survey years in the pattern of association between class origin and class destination by a single uniform difference parameter. For each pair of comparison between years, a Unidiff parameter can be estimated as follows: Fijt = η τi

O τjD τt

Y τitOY τjt

DYexp(ψ ijODφ t

Y), (6)

8. Trends in Social Fluidity across Survey Years Table 8 shows the fit of the various log-linear and log-multiplicative models described above to the 6 by 6 by 6 (origin by destination by survey year) table in Japan. The constant social fluidity model (model (2) in Table 8 - equation 1) fits the data, with the G

－ 14 －

2 value of 163.5 and the associated p-value of .012. The percentage of cases misclassified by the CSF model is only 4.3 percent and the reduction in G 2 value from the conditional independence model is over 93 percent. When we add either linear change (2a) or curvilinear change (2b) to the CSF model, the fit does not improve significantly. These models impose that the overall association between origin and destination (which is represented by the Unidiff parameter) is either linear or curvilinear from 1955 to 2005. When we allow the overall association between origin and destination represented by the Unidiff parameter to vary among six survey years (using five degrees of freedom over the CSF model), the model (2c) does not show any significant improvement over the CSF model. In other words, none of the survey years depart significantly from the CSF model, when the overall association is represented by the Unidiff parameter. Table 8 also shows the fit of the models using effect matrices. We fit two versions of the CSF model with effect matrix representation. The first version is the core fluidity model with unmodified form of matrices (model (3) – equation 3). This model does not fit the table (G 2 =233.70, df=140, p<.001). The second version is the national variant of the core fluidity model with modified effect matrices (model (4) – equation 4). This model does not fit the table, either (G 2 =211.4, df=139, p<.001). We then allow the parameter estimates of the national variant of the core fluidity model to vary by survey years (model 4a). In other words, the strength of each effect matrix is different by survey year. This variable effect model does not fit the data well (G 2 =134.6, df=84, p<.001). The last model of the table (model 4b) is the variant of the variable effect matrix model of equation (5) above. The strength of the effect matrices is allowed to vary only for selected years. The fit of this model is adequate at the one percent significant level (G 2 value of 173.63 and the associated p-value of .010). Table 9 reports the parameter estimates for the effect matrices from the last model, that is, the effect matrix model with variable parameters for selected years. It is important to notice first that of the 66 possible parameters (11 separate effects times 6 years) there are only 11 parameters which are significantly different from the CSF parameters. In other words, there are some departures from the CSF model, but the extent of deviation is not at all pervasive. Furthermore, there is no systematic trend in the deviations. For every survey year, there is at least one parameter which is different from the CSF model, but these deviations do not necessarily indicate greater openness across the survey years. For example, in 1955 the extent of class inheritance among the petty bourgeoisie class (IVab) is weaker than that in other years implying greater fluidity at least out of the petty bourgeoisie class. However, all other parameters in 1955 are the same as the constant effects, so there is hardly any evidence of exceptional fluidity. The results of the changes in the effect matrix parameters suggest that there is no noticeable trend in relative mobility. If anything, the results are consistent with Sorokin’s prediction of “trendless fluctuation.” In order to detect any change including minor ones in the odds ratio pattern, we examine the trend of all the individual odds ratios. We report the results of comparing all the 225 odds ratios that can be computed from the 6 by 6 table across the pair of survey years. The odds ratios will fall into three distinct patterns, as shown in Figure 6. The first pattern is that the odds ratio becomes closer to 1.0 or the log of odds ratio becomes close to zero. This trend suggests an increasing fluidity from one year to the next. The second pattern is the exact opposite where the log of odds ratio becomes further away from zero between two survey years thereby indicating a trend of

－ 15 －

decreasing fluidity. The third pattern occurs when the log of odds ratio goes through zero. The log of odds ratio becomes closer to zero and then away from zero, as shown in the last panel of Figure 6. Table 10 presents the results of classifying every pair of odds ratio into one of the patterns shown in Figure 6 and computing the proportion of three patterns. From 1955 to 1965, of the 225 log of odds ratios, 49 percent were moving close to zero, 32 percent moving away from zero, and 19 percent crossing zero. Almost the majority of odds ratios shows a trend of increasing fluidity and openness. From 1965 to 1975, the modal pattern is that of an increasing fluidity but these odds ratios do not constitute the majority (44%) and there is almost the same proportion of the odds ratios which are in the opposite direction (40%). From 1975 to 1985, the trend is reversed; the modal pattern is that of a decreasing fluidity. From 1985 to 1995, the trend is reversed again with the modal pattern of an increasing fluidity (49%). Finally, from 1995 to 2005, the trend is reserved yet again because the model pattern is that of a decreasing fluidity. The reversed trend from 1995 to 2005 appears to support the “post-industrial rigidity” thesis. However, we should interpret these fluctuations with caution because first the fit of the CSF model for the years 1995 and 2005 is good and second the statistical test of the difference in the overall odds ratio pattern between the these two survey years is not significant (see Table 10). Therefore, the apparent change in the direction of trend from 1995 to 2005 is not significant and may not be real. Table 10 also reports the results of running Unidiff model in order to assess whether all odds ratios are moving uniformly in the same direction (rather than to assess whether individual odds ratio is moving in the same direction). The most important finding is that the direction of the Unidiff parameter is consistent with the breakdown of the patterns, except for the period from 1965 to 1975. From 1955 to 1965, the parameter is negative implying a trend of increasing fluidity. From 1965 to 1975, the parameter is positive indicating decreasing fluidity, although it is very small and almost zero. From 1975 to 1985 the sign of the parameter is positive again, implying a trend of decreasing fluidity. From 1985 to 1995, the sign is reversed again (a minus sign) indicating a trend of increasing fluidity. Finally, from 1995 to 2005, the sign is positive, implying a trend of decreasing fluidity. However, the significance testing of the Unidiff parameters shows that the Unidiff models do not significantly improve the CSF model. Therefore, these changes implied by the Unidiff parameters should not be taken so seriously, and the pattern of relative mobility is basically stable in the postwar period in Japan. In summary, the overall picture which emerges from all these analyses of relative mobility is the stability and constancy in the pattern of social fluidity. Therefore, the prediction of Sorokin and the FJH hypothesis are consistent with our findings. The “continuous” hypothesis of the industrialism thesis predicted that social fluidity would increase during the period of rapid economic development, but there is no clear evidence that the Japanese society became more open in late-1950s and 1960s when the economy expanded rapidly. The hypothesis of the “post-industrial rigidity” was not supported by the analyses, either. There is no noticeable trend of increased rigidity or decreased openness after the late-1980s of the recessionary period.

9. Trends in Social Fluidity across Cohorts

－ 16 －

So far our analysis has been concentrated on the trends using survey year as the unit of analysis. In this section, we will use birth cohort, instead of survey year, in order to detect any trend in intergenerational mobility. The use of cohort will allow us to identify the impact of historical events and institutional changes on mobility patterns. Specifically, we attempt to test the post-industrial rigidity hypothesis which claims the increased rigidity in class structure in the 1990s, and the historical institutional hypothesis which predicts the increased fluidity among those who experienced the post-war social policies. We combine six surveys and divide all the respondents into the following six cohorts: (1) the respondents born in 1920 or before, (2) born in 1921-30, (3) born in 1931-40, (4) born in 1941-50, (5) born in 1951-60, and (6) born in 1961-75.

Table 11 shows the results of fitting various log-linear and log multiplicative models to the 6 by 6 by 6 (class origin by destination by cohort) table in Japan. The constant social fluidity model with full association does not fit the data at the one percent significance level. When we introduce the effect matrices to represent the association, neither the constant social fluidity model using the core model nor the constant social fluidity model using the national variant core fit the data at the one percent level. However, using the bic statistics, the most preferred model is the CSF model with national variant core. This model misclassifies only 4.7 percent of cases. The parameter estimates of effect matrices are also shown in the table. These parameters are very similar to those reported in the analysis of trend using survey year as the unit of analysis (see Table 9).

We introduced a Uni-diff parameter to represent the difference in the strength of the origin-destination association across cohorts. Since there are six cohorts, five Uni-diff parameters are added to the constant social fluidity model using full association and to the constant social fluidity model using the national variant core. Both models improve the CSF model, but the improvements are not significant at one percent significance level. The Uni-diff parameters from the core national variant model are shown in the table. The origin-destination association became stronger from the earliest cohort to the 1921-30 cohort, but the same association became weaker from the 1921-30 cohort to the 1931-40 cohort. The trend is reversed (that is, weakening of the association) from the 1941-50 cohort to the 1951-60 cohort. However, from the 1951-60 to the 1961-75 cohort, the association is strengthened again. The pair-wise comparisons of the strength of the origin-destination association between adjacent cohorts using Uni-diff parameters show that these comparisons are not significant at the one percent level of significance (table not shown). Just like the results from the analysis using survey year, there is fluctuation in the strength of association between origin and destination across cohorts. There is no clear tendency towards either openness or rigidity.

The youngest cohort members, that is those born between 1961 and 1975, are most likely to be affected by the recent down turn in Japanese economy, and the post-industrial rigidity hypothesis predicts the increased rigidity among these members. The Uni-diff parameter representing the pair-wise comparison of the strength of the origin-destination association between the 1951-60 cohort and the 1961-75 cohort suggests that there is a slight increase in rigidity (parameter value=1.077). 5 However, 5 The pair-wise comparison is based on the analysis of applying the Uni-diff model to the 6 by 6 by 2 (origin by destination by cohort) table. The Japanese version of the core model is used to represent the association between origin and destination. The two cohorts are the 1951-60 cohort and the

－ 17 －

the change is so small and is not statistically significant (p=.512). Therefore, there is no clear empirical evidence to support the post-industrial rigidity hypothesis.

The respondents who were born between 1921 and 1930 are most likely to be affected by the social policies introduced immediately following the defeat in the war because these people were 15 to 24 years old when the war ended in 1945 and entered the labor market when these policies took effect. The pair-wise comparison of the strength of the origin-destination association, however, does not suggest the weakening of the association among this particular cohort. If anything, the trend is the opposite, suggesting an increased association among this cohort. The historical institutional hypothesis, therefore, does not receive empirical support from our analyses.

10. Trends in Social Fluidity across Cohorts and Age Groups So far our analysis has been concentrated on the trends using survey year as the unit of analysis. Every survey includes male respondents of aged 30 to 64 who were at different career stages, and we evaluated the pattern of social fluidity for all the active members of the society at large. By using the broad cross-sectional coverage, it allowed us to compare the state of social mobility in the society at large at six different historical periods. We assumed that, by restricting our sample to those of aged 30 and above, the impacts of career stages and age are minimal. However, this assumption may not hold true, and controlling for career stage and age may affect our assessment about the trends in social fluidity. Furthermore, it is difficult to identify the effect to historical institutional changes or events on mobility pattern because they may affect particular cohort or age groups. In order to address this issue, we merged the data from six surveys and grouped the respondents into age-cohort groups, as shown in Table 12. This design distinguishes seven cohorts and four age groups. Each cohort is observed at the minimum of two different age groups, and at the maximum of four different age groups. The youngest age group includes those aged 30 to 34, thereby the age range is five years rather than ten years. The design is quasi-panel because we do not follow up the same individuals. We simply compare men who were born in the same cohort at different points in time, even though they are not exactly the same people. Table 13 presents the results of examining the trends in age-cohort design. The analysis includes 22 origin-by-destination mobility matrices, each matrix representing the combination of cohort and age group. The conditional independence model (model 1) which allows the distribution of origin and destination to vary by cohort and age (OAC and DAC three-way terms) but does not allow any association between origin and destination (omit OD term). The constant social fluidity (CSF) model adds the two-way OD term to the independence model and imposes the association between origin and destination to be constant across birth cohorts and age groups. Using the one percent significance level, it fits the data with the G 2 value of 597.2 and the associated p-value of .016. The percentage of cases misclassified by the CSF model is 7.7 percent and the reduction in G 2 value from the conditional independence model is 79 percent. 6 1961-75 cohort. 6 These figures are lower than those of the analysis of survey years because we have larger tables and more degrees of freedom to work with.

－ 18 －

Given that we use a large number of tables (6 by 6 by 22), the fit is fairly good. We relax the condition of constancy by introducing one parameter for each combination of cohort and age group and represent the difference (change) in the pattern of association by a Unidiff parameter. With 22 mobility matrices, uniform change model looses 21 degrees of freedom from the CSF model with the G2 value of 564.1 (df=504) and the associated p-value of .033. The fit is significantly improved over the CSF model at .05 level of significance (difference in G2=33.1, df=21, p=.045). When we introduce separately cohort change or age change over the CSF model, the improvement in fit is not significant (for cohort change: the difference in G2=11.3, df=6, p=.080; for age group change: the difference in G2=4.8, df=3, p=.185). Therefore, by judging from the improvement of fit, we are left with the CSF model and the Unidiff model of change with respect to both cohort and age. Using the bic statistics, the CSF model (bic=-4199.6) is preferred over the Unidiff model (bic=-4040.8) since it is more parsimonious. Figure 7 displays the estimates of uniform change (Unidiff) parameters of model (3), that is, the model with the uniform change with respect to both cohort and age. As shown in these parameters, there is no clear trend across the birth cohorts or age groups. The dominant pattern is the trendless fluctuation among the parameter estimates.

In summary, we are inclined to conclude that the stability in social fluidity observed in the analysis using survey years is re-confirmed with the analysis using birth cohorts and age groups. Although there seem to be some fluctuations by birth cohorts and age groups, there is no noticeable trend and the dominant pattern is the stability in the pattern of association between origin and destination in the late 20th century Japan. Our results are consistent with the hypothesis proposed by Sorokin and FJH. The “historical institutional” hypothesis predicted that social policies introduced immediately after the Second World War would create greater openness in the society. Two birth cohorts of 1911-1920 and 1921-1930 are most likely to be affected by these policies, but there is no empirical evidence to suggest that these cohorts exhibit more fluidity than other cohorts.

11. Models of Associations between Class Origin and Education and

between Education and Class Destination

The second part of this study will examine the linkage between class origin and education and between education and class destination in postwar Japan. In order to empirically analyze these linkages, we need models to represent the pattern of association between class origin and education and the pattern between education and class destination. The models of association are concerned with capturing the relative pattern, not absolute. In other words, we do not intend to examine the inflow and outflow patterns regarding the relationship between class origin and education and between education and class destination. We have already learned about the differences in the distributions of class origin and class destination across our surveys. The main purpose of our exercise aims to examine the pattern of association, net of differences in the marginal distributions. The models which are used in this study are adopted from the log-linear models proposed in Ishida, Muller, and Ridge (1995). Following the same logic of the core

－ 19 －

social fluidity model (Erikson and Goldthorpe 1987a, 1987b, 1992b), the associations between class origin and education and between education and class destination are represented by a set of different effects which are shown in Figure 8. The multiplicative equation for the model of association between class origin and education is shown below: Fik = η τi

O τkQ τ(ik)

SO1 τ(ik)SO2 τ(ik)

MO1 τ(ik)MO2P τ(ik)

MO2F τ(ik)MO2S τ(ik)

MO2U, (7) where Fik refers to the expected frequency in cell (i,k) of the class origin by education table, η is a scale term, τi

O is the main effect of class origin, τkQ is the main effect of

education, and the rest of the parameters represent effect matrices described in Figure 8. Similarly, the multiplicative equation for the model of association between education and class destination is: Fjk = η τj

D τkQ τ(jk)

HQ1 τ(jk)HQ2 τ(jk)

LQ1 τ(jk)LQ2P τ(jk)

LQ2F τ(jk)LQ2S τ(jk)

LQ2U, (8) where Fjk refers to the expected frequency in cell (j,k) of the class destination by education table, η is a scale term, τj

D is the main effect of class destination, τkQ is the

main effect of education, and the rest of the parameters represent effect matrices described in Figure 8. The main features of the model of association between class origin and education include the following. The advantages of the sons of the professional managerial class are represented by two effects: they are more likely to be found in the highest level of education (SO1) and less likely to end up in the lowest education level (SO2) than the sons of other classes. The disadvantages faced by the farming class and the manual working classes are represented by two kinds of effect: the sons of these classes are less likely to attain the highest level of education (MO1) and more likely to be found in the lowest level of education (MO2) than the sons of other classes. Along with the petty bourgeoisie, the extent of the disadvantage in ending up in the lowest education level varies by the farming class and the working classes. Similarly, there are symmetric effects in the model of the association between education and class destination. The advantages of the highly educated are captured by HQ1 and HQ2 effects: they tend to have better access to professional managerial positions and to avoid manual destinations. The disadvantages faced by those who had the lowest level of education are captured by LQ1 and LQ2 effects: less chance of entry into the professional managerial class and more chance of entry into the farming, petty bourgeoisie, and manual destinations.

12. Trends in Associations Involving Education One of the major social transformations in post-war Japan is the expansion of educational system -- first the expansion of the post-compulsory education of senior high schools and later the expansion of the tertiary education sector. As shown in Table 14, educational distributions of the respondents to the six national surveys reflect these changes. The percentage of respondents in the middle level of education (senior high school level) increased steadily from 1955 to 1985, with an attenuated increase after 1985. The percentage for the highest level of education which includes junior colleges and universities began its increase from 1965, with a particularly large jump from 1975 to

－ 20 －

1985. The major question which will be examined in this section is whether the associations between education and class origin or between education and class destination have undergone major changes, in response to the changes in the educational distribution. With respect to the effect of class origin on education, we ask whether educational opportunities became more open and less affected by social origin. With respect to the effect of education on class destination, we ask whether the allocation of class positions became increasingly more affected by education. Table 15 reports the results of applying various log-linear models to the three-way table of class origin by education by survey year (first panel) and to the three-way table of education by class destination by survey year (second panel). The first model in each panel is the conditional independence model which is used as the basis of evaluating the fit of other models. The second model is the constant social fluidity (CSF) model. It is the same model used in the analysis of class origin by class destination tables (as shown in equation 1), but either class origin or class destination is replaced by education. The CSF model fits well the three-way table of class origin by education by year. This suggests that the association is characterized by the same pattern across six survey years. In contrast, the CSF model does not fit the three-way table of education by class destination by year. However, when we allow the odds ratio pattern in 1955 to be different and fix the pattern for the rest of the years, the CSF model (with 1955 excluded) fits the education by class destination tables (see model 2a: G 2 =42.32, df=40, p=.371). The third model in each panel is the constant social fluidity model using effect matrices described in Figure 8. This model is similar to the model in equation (3) and can be written as the following multiplicative form. For the table of class origin by education by year: Fikt = η τi

O τkQ τt

Y τitOY τkt

QY τ(ik)SO1 τ(ik)

SO2 τ(ik)MO1 τ(ik)

MO2P τ(ik)MO2F τ(ik)

MO2S τ(ik)MO2U, (9)

and for the table of education by class destination by year: Fjkt = η τj

D τkQ τt

Y τjtDY τkt

QY τ(jk)HQ1 τ(jk)

HQ2 τ(jk)LQ1 τ(jk)

LQ2P τ(jk)LQ2F τ(jk)

LQ2S τ(jk)LQ2U, (10)

where Fikt or Fjkt refer to the expected frequency in cell (i,k,t) or (j,k,t) of the three-way table, η is a scale term, τi

O is the main effect of class origin, τkQ is the main effect of

education, τjD is the main effect of class destination, τt

Y is the main effect of year, two-way terms represent the marginal distributions (O, D, and Q) varied by survey year, and the rest of the parameters represent effect matrices described in Figure 8. The most important feature of this model is that effect matrices do not vary by year (t). Again the fit of this model to the class origin by education table is very good. The model misclassifies only 2.7 percent of cases and the reduction in G2 from the independence model is 96 percent. The fit of this model to the education by class destination table is not adequate with p-value of less than .001. However, again if we exclude 1955, this model produces an adequate fit as shown in row 3a of Table 15 (G2=59.6, df=46, p=.086). These results imply that the association between education and class destination was very stable from 1965 to 2005 while there were some significant changes in the association between education and class destination from 1955 to 1965. In order to specify these changes, the fourth model in Table 15 allows some

－ 21 －

effect parameters to vary between 1955 and the rest of the years. Based on the statistical tests of the difference in the effect matrix parameters, HQ1, HQ2, LQ2U, LQ2S, and LQ2F effects are different between 1955 and the rest of the years. The parameter estimates from the CSF model using effect matrices for the origin by education table are shown in Table 16, and by construction of the model the effect parameter estimates do not vary by survey year. The sons of the professional-managerial class are advantaged in educational opportunities in two ways. The most conspicuous advantage deals with their ability to have better access to tertiary education (SO1), while their ability to avoid the lowest level of education is relatively weak (SO2). The sons of the farming class (MO2F) are most disadvantaged in educational attainment followed by the sons of the unskilled working class (MO2U). The major source of disadvantage derives from their tendency to end up in the lowest educational level. The barriers to educational attainment faced by the sons of the farming class are particularly striking throughout the post-war period. It should be emphasized that the stability in the association between class origin and education is conspicuous because the educational system expanded substantially during the post-war period in Japan (see Table 14). Despite the continuous expansion of educational institutions in Japan, class origin remains to affect educational attainment. There is no evidence of weakening effect of class origin on educational advancement. Table 17 presents parameter estimates from the model which allows some cross-year variations in effect matrices between 1955 and the other years for the education by destination table. Five out of seven effect parameters are different in 1955. Furthermore, except for HQ2 effect, HQ1, LQ2U, LQ2S, and LQ2F effects are all stronger in 1955 than in other years. In other words, the linkage between education and class positions is much tighter in 1955 than in later years. Despite the increasing level of industrialization and accompanied changes in the class structure and the educational system in post-war Japan, there is no empirical evidence to support the increasing effect of education on the allocation of class positions. In addition to the general decline in the strength of some effects, the major change between 1955 and 1965 relates to the mechanism through which the highly educated assure their advantage. In 1955 the major source of advantage of the highly educated came from the privileged access to professional-managerial positions. However, after 1965 the major source of advantage came from their ability to avoid manual class positions. This change may be related to the trend of the rapid expansion of the professional-managerial class in the class destination distribution from 1955 to 1965. Since the disadvantage faced by those who have low level of education in access to professional-managerial positions has been consistently weak in Japan, the expansion of the professional-managerial class might have opened up chances of entry to this class by those who had less than higher education. The other possibility for explaining the change pertains to the composition of those who had tertiary education. In 1955, they are composed predominantly by those who attended the pre-war educational system. The graduates of pre-war universities and high schools probably occupied the distinctively advantaged positions which resulted in the very strong linkage between high educational attainment and professional managerial positions. Finally, the association between education and class destination is examined after controlling for class origin. This exercise allows us to estimate the net effect of education on class destination, after controlling for the spurious component due to class origin. In order to control for class origin, we need to consider a four-way table of

－ 22 －

education by class destination by class origin by year. For each pair of association, we use effect matrix representation discussed above. The multiplicative equation is as follow: Fijkt = η τi

O τjD τk

Q τtY τit

OY τjtDY τkt

QY τ(ik)SO1 τ(ik)

SO2 τ(ik)MO1 τ(ik)

MO2P τ(ik)MO2F τ(ik)

MO2S τ(ik)

MO2U τ(jkt)HQ1 τ(jkt)

HQ2 τ(jk)LQ1 τ(jk)

LQ2P τ(jkt)LQ2F τ(jkt)

LQ2S τ(jkt)LQ2U τ(ijt)

DIGk τ(ij)HI1 τ(ij)

HI2 τ(ij)

AF2A τ(ij)AF2B τ(ijt)

AF1J , (11) where Fijkt refers to the expected frequency in cell (i,j,k,t) of the four-way table, η is a scale term, τi

O is the main effect of class origin, τkQ is the main effect of education, τj

D is the main effect of class destination, τt

Y is the main effect of year, two-way terms represent the marginal distributions (O, D, and Q) varied by survey year, and the rest of the parameters represent effect matrices described in Figure 5 and 8. Some of the education-destination effect matrices vary by year (represented by subscript t), and some of the origin-destination effect matrices also vary by year (represented by subscript t). Year-specific effect matrices are determined by the result of three-way table shown in Table 9 (for origin-destination association) and Table 17 (for education-destination association). The fit of model (9) to the four-way table is good: G2=464.32, df=440, p=.204. The percentage of cases misclassified by the model is 7.5 percent, and the reduction in G2 from the independence model is over 90 percent. Table 18 reports parameter estimates for education-destination effect matrices from model (9). As shown in Table 18, this model allows HQ1, HQ2, LQ2U, LQ2S and LQ2F to be different in 1955, so that the same pattern is used for the 3-way and 4-way tables. First, the most important finding in Table 18 is that the effects are very similar to those in Table 17 (3-way table). In other words, controlling for class origin does not change the overall pattern of association between education and class destination. The change (reduction) in the effect is minimal, and therefore the spurious component is very small. Second, the impact of controlling for class origin is uniform across survey years. Although some of the parameters in 1955 are allowed to be different from those in other years, the percentage reductions in 1955 are about the same as those in other years. Third, there are two effect matrices that are affected by the control by class origin although the extent of change is small. The association between low education and farm destination (LQ2F) is reduced by 15 (in 1955) and by 25 percent (in rest of years). This is probably due to the fact that men of farm origin are more likely to have low education (MO2F) and, at the same time, are more likely to inherit farm class (inheritance parameter of farm - DIG(IVc+VIIb)) than those of other class origins. Similarly, the small part of the advantage enjoyed by the highly educated (HQ2) can be explained by class origin. The part of the reason why the highly educated avoid manual destinations derives from the fact that the highly educated came from the professional-managerial origin. We repeated the trend analysis involving education by replacing survey year with cohort. We combined the respondents from six surveys and group them into six birth cohorts: (1) the respondents born in 1920 or before, (2) born in 1921-30, (3) born in 1931-40, (4) born in 1941-50, (5) born in 1951-60, and (6) born in 1961-75. The results are shown in Appendix 1. The substantive findings replicate those already reported in the analysis using survey year as the unit of analysis, and we do not need to qualify any of our conclusions even though we used cohort as the unit of analysis. In summary, what we found in the gross association between education and class

－ 23 －

destination (Table 17) was not altered in the net association between education and destination controlling for class origin. Although there were major transformation in the class structure and expansion in the educational system in post-war Japan, the linkage between education and class positions (both in gross and net terms) is stable from 1965 to 1995. The linkage seems to have weakened from 1955 to 1965. These results are not consistent with the industrialism thesis which predicted the strengthening association between education and class positions in industrial societies.

13. Role of Education in Intergenerational Reproduction and Mobility This final section addresses whether the pattern of association between class origin and class destination is mediated by education. Our strategy is to look at the change in the parameters of the effect matrices representing origin to destination transition (Figure 5), before and after controlling for education. Some of the effect matrices are allowed to vary by survey year (as reported in Table 9). Similarly, both the association between class origin and education and the association between education and class destination are represented by effect matrices (Figure 8), and some effect matrices for education to destination transition are allowed to vary for 1955 (as reported in Table 17). 7

Table 19 reports the parameter estimates for origin to destination transition and the percentage reduction in the estimates after controlling for education. Four findings stand out in Table 19. First, the role of education in mediating class inheritance varies by class being reproduced. The inheritance of the professional-managerial class is facilitated by education. The large part (half to two-thirds of the effect of inheritance) can be explained by education. The professional-managerial class uses education most effectively to reproduce its privileged position in the labor market. Similarly, the tendency for the reproduction of the routine non-manual class is in part explained by educational attainment. Close to a half (40 percent) of the inheritance effect of the routine non-manual class is mediated by education. In contrast, the inheritance of other classes is not affected by education. For example, the inheritance of the petty bourgeoisie class is nothing to do with education. It is probably the result of direct transmission of physical capital, in the form of small business or financial assets. The inheritance of the farming class is likely to take place by passing the farming land from one generation to the next. The reproduction of manual working classes is probably affected by the lack of economic resources and social capital. Second, the effect of hierarchy 2 (HI2) is explained entirely by educational attainment. This effect implies a low propensity for the sons of the professional-managerial class to be downwardly mobile to the unskilled manual class and a correspondingly low propensity for the sons of the unskilled manual class and farming class to be upwardly mobile to the professional-managerial class. The sons of the professional-managerial class are able to avoid the long-range downward mobility because of their superior education, and the sons of the farming and unskilled manual class face barriers to long-range upward mobility due to their lack of superior education. The effect of hierarchy 1 (HI1) is not much affected by education, suggesting that educational attainment is not the major reason for the hierarchical barriers expressed in hierarchy 1. 7 The fit statistics for this model are already reported above.

－ 24 －

Third, the positive affinity A (AF2A), the propensity for movement between the professional-managerial class and the routine non-manual class, is facilitated by education. The sons of the professional-managerial class show a high propensity to move into the routine non-manual class, and the sons of the routine non-manual class show a high propensity to move into the professional-managerial class. These exchanges are likely to be the result of educational attainment. The other positive affinity (AF2B) and the negative affinity (AF1) are in part affected by education. However, these effects are also likely to be produced by mechanisms not related to education. The effect of AF2B relates to three types of movement: mobility between the professional-managerial class and the petty bourgeoisie (probably due large part to an increase or a decrease in the scale of business); mobility between the petty bourgeoisie and the farming class (probably due to the transfer of capital between the farm and non-farm sectors); and mobility among the skilled manual, unskilled manual, and farming class (probably due to the similarity in manual work). The effect of AF1 relates to a low propensity for the sons of the professional-managerial class to be downwardly mobile to the ranks of manual working classes. This barrier to downward mobility is due in part to superior education of the sons of the professional-managerial class, but it is also probably due to superior economic and social resources of the professional-managerial class. Fourth, the role of education in mediating the association between class origin and class destination is very similar across the survey years. In other words, these results suggest that the mechanism of class inheritance and mobility is stable throughout the post-war Japan. The inheritance of the professional-managerial class has always been facilitated by educational attainment, and the inheritance of the petty bourgeoisie, the farming and the manual working classes has always been produced by the process largely independent of education. The long-range upward and downward mobility represented in hierarchy 2 effect is also explained by educational attainment throughout the post-war period. These results imply that not only the pattern of intergenerational reproduction and mobility (that is, the model of core social fluidity expressed in Figure 5) but also the mechanism through which the pattern is produced are stable in post-war Japan. We again repeated the analysis on the role of education by replacing survey year with cohort. We combined the respondents from six surveys and group them into six birth cohorts: (1) the respondents born in 1920 or before, (2) born in 1921-30, (3) born in 1931-40, (4) born in 1941-50, (5) born in 1951-60, and (6) born in 1961-75. The results are shown in Appendix 2. The reduction in the strength of the O-D effect matrices after controlling for education follows a very similar picture. In other words, using cohort, instead of survey year, does not alter our major findings.

14. Summary and Conclusion This paper first analyzed the trends in intergenerational class mobility in late 20th century Japan using six national surveys conducted in post-war Japan. Japanese economy experienced high-speed growth in the 1960s and early 1970s, followed by a recession and sustained economic growth until early 1990s when serious recession hit the country. By 2005 the country slowly moved out of the phase of recession. Reflecting these changes, the Japanese society has experienced dramatic and rapid changes in its

－ 25 －

class structure both among the sons’ and the fathers’ generation during the postwar period. In particular, by following the path of late but rapid industrialization, rapid contraction of the farming sector was accompanied by the expansion of both the blue-collar industrial sector and the white-collar sector almost at the same time. This particular path of development sets Japan apart from European nations.

Absolute mobility rates are influenced by this path and showed some systematic trends across the six survey years. The changes in absolute mobility rates are most pronounced during the high-speed economic growth period of the late-1950s and 1960s. Total mobility rates increased sharply from 1955 to 1965 and continued to increase modestly until 1985. Outflow rates to the farming class decreased dramatically during the 40-year span, especially from 1955 to 1965, while outflow rates to the professional managerial class increased steadily. These patterns reflect the changes in the class destination distribution across survey years. There is evidence that massive outflows from the farming class were facilitated by education. Men from the farming origin who completed middle schools under the new educational system introduced in 1947 benefited from the assistance of schools in finding urban employment. The two distinctive features of the Japanese working class that were reported in the previous cross-national studies (Ishida, Goldthorpe, and Erikson 1991; Ishida 2001) were reconfirmed in our trend analysis: a low level of both intergenerational stability and self-recruitment. The Japanese manual working class never fully developed its “demographic identity” throughout the entire post-war period. When we shift our attention to relative mobility rates or social fluidity, a very different picture emerges. Despite the fluctuations in the economy and different pace of industrialization during the post-war period, there seems to be stability in the pattern of association between class origin and class destination. The uniform difference model applied to mobility tables created from six surveys does not show any systematic trend. There is no clear tendency towards greater openness in post-war Japan, contrary to the prediction of the industrialism thesis. Even when we analyzed more detailed pattern of mobility by using the core social fluidity model of the CASMIN project, the findings suggest no systematic change in the core social fluidity model across six surveys. Although there are some fluctuations in the parameter estimates across some survey years, there is no systematic trend of either increasing or decreasing fluidity. Finally, birth cohort and age group are replaced with survey year to examine further trend in the data. Although there seem to be some fluctuations by birth cohorts and age groups, there is no noticeable trend and the dominant pattern is the stability in the pattern of association between origin and destination in the late 20th century Japan. In summary, it is the combination of rapidly changing absolute mobility rates and stability in relative mobility rates that characterizes the post-war Japanese mobility experience. The stable core pattern of the association between class origin and class destination coexisted with the changing context of class structure. The second part of this study examined the relationship of education to social background and class destination. The results of cross-temporal comparisons of associations involving education are generally not supportive of the industrialism thesis. Despite the rapid expansion of the educational system and changes in class structure in post-war Japan, the effect of class origin on the chances of educational attainment is remarkably stable from 1955 to 2005. There is no clear empirical evidence to suggest the declining significance of social background in the process of educational attainment in post-war Japan. The effect of education on the allocation of class positions shows

－ 26 －

some changes from 1955 to 1965. However, from 1965 to 2005 the effect is very stable. Furthermore, the changes from 1955 to 1965 are primarily in the direction of loosening the linkage between education and class positions, rather than strengthening the association. There is no general trend of increasing importance of education in the allocation of class positions in post-war Japan. In summary, the thesis of industrialism receives very little support from the findings from cross-temporal comparisons. Finally, this study addressed the issue of the role of education in intergenerational reproduction and mobility. Our results indicate that the role of education in reproducing class relations and in generating mobility processes varies by the type of class relations reproduced and types of mobility in question. The inheritance of the professional-managerial class has always been facilitated by educational attainment, and the inheritance of the petty bourgeoisie, the farming and the manual working classes has always been produced by the process largely independent of education. The long-range downward mobility from the professional-managerial class to unskilled manual class and the long-range upward mobility from the unskilled manual class and farming class to the professional-managerial class are also explained by educational attainment throughout the post-war period. These results suggest that both the pattern of intergenerational reproduction and mobility as expressed by the model of core social fluidity and the mechanism through which the core fluidity pattern is produced are stable in post-war Japan.

References Becker, Gary. 1975. Human Capital. New York: National Bureau of Economic Research. Bell, Daniel. 1973. The Coming of Post-industrial Society. New York: Basic Books. Bernstein, Basil. 1977. Class, Codes and Control. Volume 3, 2nd ed. London: Routledge and Kegan Paul. Blau, Peter M. and Otis Dudley Duncan. 1967. The American Occupational Structure. New York: Wiley. Bourdieu, Pierre. 1973. “Cultural Reproduction and Social Reproduction.” Pp. 7-112 in Knowledge, Education and Cultural Change, edited by Richard Brown. London: Tavistock. Bourdieu, Pierre and Jean-Claude Passeron. 1977. Reproduction in Education, Society and Culture. Beverly Hills: Sage. Bowles, Samuel and Herbert Gintis. 1976. Schooling in Capitalist America. New York: Basic Books. Breen, Richard. 2004. Social Mobility in Europe. Oxford: Oxford University Press. Cole, Robert and Ken’ichi Tominaga. 1976. “Japan’s Changing Occupational Structure

－ 27 －

and Its Significance.” Pp. 53-95 in Japanese Industrialization and Its Social Consequences, edtied by Hugh Patrick. Berkeley: University of California Press. Collins, Randall. 1979. The Credential Society. New York: Academic Press. Dale, Peter N. 1986. The Myth of Japanese Uniqueness. New York: St Martin’s Press. Davis, Kingsley. 1962. “The Role of Class Mobility in Economic Development.” Population Review 6. Davis, Kingsley and Wilbert Moore. 1945. “Some Principles of Stratification.” American Sociological Review 10: 242-9. Dore, Ronald P. (ed.). 1967. Aspects of Social Change in Modern Japan. Princeton: Princeton University Press. Ehara, Takekazu. 1984. Gendai Koto Kyoiku no Kozo (The Structure of Contemporary Higher Education). Tokyo: University of Tokyo Press. Erikson, Robert and John Goldthorpe. 1987a. “Commonality and Variation in Social Fluidity in Industrial Nations: A Model for Evaluating the 'FJH Hypothesis'.” European Sociological Review 3: 54-77. _____ 1987b. “Commonality and Variation in Social Fluidity in Industrial Nations: The Model of Core Fluidity Applied.” European Sociological Review 3: 145-66. _____ .1992a. “The CASMIN Project and the American Dream.” European Sociological Review 8: 283-305. _____ .1992b. The Constant Flux: A Study of Class Mobility in Industrial Societies. Oxford: Clarendon Press. Erikson, Robert, John Goldthorpe, and Lucienne Portocarero. 1979. “Intergenerational Class Mobility in Three Western European Societies.” British Journal of Sociology 30: 415-41. Featherman, David L. and Robert M. Hauser. 1978. Opportunity and Change. New York: Academic Press. Featherman, David, Frank L. Jones and Robert M. Hauser. 1975. “Assumptions of Social Mobility Research in the United States: The Case of Occupational Status.” Social Science Research 4: 339-60. Fujita, Hidenori. 1981. “Kyoiku Kikai wa Kaihoteki ka” (Are the Educational Opportunities Open?). Pp. 27-44 in Nihon no Gakureki Shakai wa Kawaru (Change in the Japanese Educational Credential Society), edited by Yo Takeuchi and Makoto Aso. Tokyo: Yuhikaku.

－ 28 －

Goldthorpe, John H. 1982. “On the Service Class, Its Formation and Future.” Pp. 162-85 in Social Class and the Division of Labour, edited by Anthony Giddens and Gavin Mackenzie. Cambridge: Cambridge University Press. _____. 1985a. “Soziale Mobilitat und Klassenbildung: Zur Erneuerung einer Tradition soziologischer Forschung.” Pp. 174-204 in Die Analyse Sozialer Ungleichheit, edited by Hermann Strasser and John H. Goldthorpe. Opladen: Westdeutscher Verlag. _____. 1985b. “On Economic Development and Social Mobility.” British Journal of Sociology 36: 549-73. _____ (with C. Llewellyn and C. Payne) 1987. Social Mobility and Class Structure in Modern Britain. 2nd ed. Oxford: Clarendon Press. Goodman, Leo. 1978. The Log-Linear Analysis of Cross-Classified Data. Boston: Abt Press. Grusky, David B. and Robert M. Hauser. 1984. “Comparative Social Mobility Revised: Models of Convergence and Divergence in 16 Countries.” American Sociological Review 49: 19-38. Halsey, A. H., A. F. Heath, and J. M. Ridge. 1980. Origins and Destinations. Oxford: Clarendon Press. Hout, Michael. 1989. Following in Father's Footsteps. Cambridge, Mass.: Harvard University Press. Imada, Takatoshi. 1989. Shakai Kaiso to Seiji (Social Stratification and Politics). Tokyo: University of Tokyo Press. _____. 1997. “Industrialization and the Regime of Social Mobility in Postwar Japan: 1955-95.” Paper presented at the meeting of the ISA Research Committee 28, August 14-15 in Quebec City, Canada. Ishida, Hiroshi. 1993. Social Mobility in Contemporary Japan: Educational Credentials, Class and the Labour Market in a Cross-National Perspective. London: Macmillan _____ . 2001. “Industrialization, Class Structure, and Social Mobility in Postwar Japan.” British Journal of Sociology 52: 579-604. _____ .2004. “Entry into and Exit from Self-Employment in Japan,” in The Reemergence of Self-Employment: A Comparative Study of Self-Employment Dynamics and Social Inequality, edited by Richard Arum and Walter Muller (Princeton: Princeton University Press), pp. 348-87. _____ .2007. “Japan: Educational Expansion and Inequality in Access to Higher Education,” in Stratification in Higher Education: A Comparative Study, edited by Yossi Shavit, Richard Arum, and Adam Gamoran (Stanford: Stanford University Press).

－ 29 －

Ishida, Hiroshi, John Goldthorpe and Robert Erikson. 1991. “Intergenerational Class Mobility in Post-war Japan.” American Journal of Sociology 96: 954-92. Ishida, Hiroshi, Walter Muller, and John Ridge. 1995. “Class Origin, Class Destination, and Education: A Cross-national Study of Industrial Nations.” American Journal of Sociology 101: 145-93. Jansen, Marius (ed.). 1965. Changing Japanese Attitudes Towards Modernization. Princeton: Princeton University Press. Kanomata, Nobuo. 1987. “Shakai Kaiso to Ido Kenkyu ni okeru Sangyoka Meidai no Saikento” (Re-evaluation of Industrialism Thesis in the Study of Social Stratification and Mobility). Gendai Shakaigaku 13: 156-81. _____ . 1997. “Sengo Nihon niokeru Sedaikan Ido no Hendo” (Changes of Intergenerational Mobility). Kodo Keiryogaku 24: 20-7. Kariya, Takehiko. 1995. Taishu Kyoiku Shakai no Yukue (Mass Education Society). Tokyo: Chuo Koronsha. Kariya, Takehiko, Shinji Sugayama, and Hiroshi Ishida. 2000. Gakko, Shokuan to Rodoshijyo (Schools, Public Employment Offices and the Labor Market in Post-war Japan) (University of Tokyo Press). Kelly, William W. 1991. “Directions in the Anthropology of Contemporary Japan.” Annual Review of Anthropology 20: 395-431. Kerckhoff, Alan C., Richard T. Campbell, and Jerry M. Trott. 1982. “Dimensions of Educational and Occupational Attainment in Great Britain.” American Sociological Review 47: 347-64. Kerr, C., Dunlop, J. T., Harbinson, F. H., and Myers, C. A. 1960. Industrialism and Industrial Man. Cambridge: Harvard University Press. Kerr, C. 1983. The Future of Industrial Societies. Cambridge, Mass.: Harvard University Press. Koike, Kazuo. 1988. Understanding Industrial Relations in Modern Japan. London: Macmillan. Kojima, Hideo and Atsushi Hamana. 1984. “Shokugyo Ido no Keiko Bunseki” (Trend Analysis of Occupational Mobility). Ibaraki Daigaku Kyoiku Gakubu Kiyo 33: 17-32. Konig, Wolfgang, Paul Luttinger, and Walter Muller. 1988. “Comparative Analysis of the Development and Structure of Educational Systems.” CASMIN Working Paper No. 12. Mannheim: CASMIN Project.

－ 30 －

Levy, Marion J. 1966. Modernization and the Structure of Societies. Princeton: Princeton University Press. Lipset, Seymour M. and Hans L. Zetterberg. 1959. “Social Mobility in Industrial Societies.” Pp. 11-75 in Social Mobility in Industrial Society, edited by Seymour M. Lipset and Reinhard Bendix. Berkeley: University of California Press. Mare, Robert D. 1981. “Change and Stability in Educational Stratification.” American Sociological Review 46: 72-87. Mincer, Jacob. 1974. Schooling, Experience, and Earnings. New York: National Bureau of Economic Research. Mitani, Hiroshi. 1977. “Shakai Ido to Kyoiku” (Social Mobility and Education). Pp. 156-89 in Kindai Nihon Kenkyu Nyumon (The Introduction to the Study of Modern Japan), edited by Takafusa Nakamura and Takashi Ito. Tokyo: University of Tokyo Press. Murakami, Yasusuke. 1984. Shin Chukan Taishu no Jidai (The Age of New Middle Mass). Tokyo: Chuo Koronsha. Nakane, Chie. 1970. The Japanese Society. Berkeley: University of California Press. Naoi, Atsushi. 1989. “Kuzure Hajimeta Heijunka Shinwa” (The Breakdown of the Myth of Equalization). The Asahi Journal (April): 14-9. _____ . 1990. “Joron” (Introduction). Pp. 1-14 in Gendai Nihon no Kaiso Kozo (The Structure of Stratification in Contemporary Japan), vol. 1, edited by Atsushi Naoi and Kazuo Seiyama. Tokyo: University of Tokyo Press. Ohtake, Fumio. 2005. Nihon no Fubyodo (Japan’s Inequality). Tokyo: Nihon Keizai Shinbunsha. Ozawa, Masako. 1985. Shin Kaiso Shohi no Jidai (The Age of the New Stratified Pattern of Consumption). Tokyo: Nihon Keizai Shinbunsha. Parkin, Frank. 1979. Marxism and Class Theory. New York: Columbia University Press. Parsons, Talcott. 1951. The Social System. Glencoe, Il.: Free Press. Raftery, Adrian and Michael Hout. 1993. “Maximally Maintained Inequality: Expansion, Reform, and Opportunity in Irish Education 1921-75.” Sociology of Education 66: 41-62. Robinson, Robert V. 1984. “Reproduction of Class Relations in Industrial Capitalism.” American Sociological Review 49: 182-96. Robinson, Robert V. and Jonathan Kelly. 1979. “Class as Conceived by Marx and Dahrendorf.” American Sociological Review 44: 38-58.

－ 31 －

Sato, Toshiki. 2000. Japan as an Unequal Society. Tokyo: Chuokoron Shinsha. Seiyama, Kazuo, Atsushi Naoi, Yoshimichi Sato, Kazuharu Tsuzuki and Hideo Kojima. 1990. “Gendai Nihon no Kaiso Kozo to sono Susei” (The Structure of Stratification in Contemporary Japan and its Trends). Pp. 15-50 in Gendai Nihon no Kaiso Kozo (The Structure of Stratification in Contemporary Japan), vol. 1, edited by Atsushi Naoi and Kazuo Seiyama. Tokyo: University of Tokyo Press. Shavit, Yossi and Hans-Peter Blossfeld (eds.). 1993. Persistent Inequality. Boulder, CO: Westview Press. Shavit, Yossi and Walter Muller (eds.). 1998. From School to Work: A Comparative Study of Educational Qualifications and Occupational Destinations. Oxford: Clarendon Press. Shirahase, Sawako. 2005. Shoshi Korei Shakai no Mienai Kakusa (The Unseen Gaps in an Aging Society). Tokyo: University of Tokyo Press. Shirahase, Sawako (ed.). 2006. Henka Suru Shakai no Fubyodo (Inequality in a Changing Society). Tokyo: University of Tokyo Press. Sorokin, Pitirim A. 1959 (1927). Social and Cultural Mobility. Glencoe, Il.: The Free Press. Sugimoto, Yoshio. 1997. An Introduction to Japanese Society. Cambridge: Cambridge University Press. Tachibanaki, Toshiaki. 1988. Japan’s Economic Inequality. Tokyo: Iwanami Shoten. Treiman, Donald. 1970. “Industrialization and Social Stratification.” Pp. 207-34 in Social Stratification: Research and Theory for the 1970s, edited by Edward O. Laumann. Indianapolis: Bobbs-Merrill. _____. 1977. Occupational Prestige in Comparative Perspective. New York: Academic Press. _____. 1990. “Industrialization and Social Stratification.” Sociological Inquiry 40: 207-34. Treiman, Donald and Kam-Bor Yip. 1989. “Educational and Occupational Attainment in 21 Countries.” Pp. 373-94 in Cross-National Research in Sociology, edited by Melvin L. Kohn. Newbury Park: Sage. Tominaga, Ken'ichi. 1979. “Shakai Kaiso to Shakai Ido no Susei Bunseki” (Trend Analysis of Social Stratification and Social Mobility). Pp. 33-87 in Nihon no Kaiso Kozo (The Structure of Stratification in Japan), edited by Ken'ichi Tominaga. Tokyo: University of Tokyo Press.

－ 32 －

_____. 1992. “Sengo Nihon no Shakai Kaiso to sono Hendo: 1955-1985” (Social Stratification and its Changes in Post-war Japan: 1955-1985). Pp. 429-95 in Gendai Nihon Shakai, vol. 6, (Contemporary Japanese Society) edited by Institute of Social Science, University of Tokyo. Tokyo: University of Tokyo Press. Ushiogi, Morikazu. 1975. “Shinro Kettei Katei no Pasu Kaiseki” (A Path Analysis of the Process of High School Attendance). Kyoiku Shakaigaku Kenkyu 30: 75-86. _____. 1978. Gakureki Shakai no Tenkan (A Change of the Educational Credential Society). Tokyo: University of Tokyo Press. _____. 1980. Yureru Gakureki Shakai (Shifting Educational Credential Society). Tokyo: Shiseido. Vallet, Louis-Andre. 2001. “Forty Years of Social Mobility in France: Change in Social Fluidity in the Light of Recent Models.” Revue Francaise de Sociologie 42 (Supplement): 5-64. Ward. Robert E (ed.). 1967. Political Development in Modern Japan. Princeton: Princeton University Press. Xie, Yu. 1992. “The Log-Multiplicative Layer Effect Model for Contingency Mobility Table.” American Sociological Review 57: 380-95. Yamaguchi, Kazuo. 1983. “The Structure of Intergenerational Occupational Mobility.” American Journal of Sociology 88: 718-45. Yasuda, Saburo. 1971. Shakai Ido no Kenkyu (The Study of Social Mobility). Tokyo: University of Tokyo Press.

－ 33 －

0%

10%

20%

30%

40%

50%

60%

70%

80%

90%

100%

1920 1940 1955 1965 1975 1985 1995

Figure 1 Industrial Distribution in Japan

Tertiary

Secndary

Primary

Figure 2 Percent Change in GDP

-2

0

2

4

6

8

10

12

14

1956 1961 1966 1971 1976 1981 1986 1991 1996 2001

Year

% C

han

ge

%GDP

average 1956-73(9 .1%)

average 1974-90 (3 .8%)

average 1991-04 (1 .2%)

－ 34 －

Figure 3 Educational Expansion in Japan

0

10

20

30

40

50

60

70

80

90

100

1890 1900 1910 1920 1930 1940 1950 1960 1970 1980 1990 2000

Enro

llment

rate

Compulsory

Secondary

Higher

Figure 4. Educational Expansion and Outflow Rates from theFarming Class by Birth Cohort

0.000

0.100

0.200

0.300

0.400

0.500

0.600

0.700

0.800

1901-05

1906-10

1911-15

1916-20

1921-25

1926-30

1931-35

1936-40

1941-45

1946-50

1951-55

1956-60

1961-65

1966-70

1971-75

0

10

20

30

40

50

60

70

80

90

100

Ⅳc→Ⅰ+Ⅱ

Ⅳc→Ⅳc

Ⅳc→Ⅴ/Ⅵ/Ⅶa

Education: High

Education: Middle

－ 35 －

F

igure 5 Core Social Fluidity Model (Model of Association between Origin and Destination)

and the National Variant of the Core Social Fluidity Model

__________________________________________________________________________________________________

Core Model:

Inheritance Effect Matrix (DIG)

I+II III IVab IVc+VIIb V+VI VIIa

I+II 2 1 1 1 1 1

III 1 3 1 1 1 1

IVab 1 1 4 1 1 1

IVc+VIIb 1 1 1 5 1 1

V+VI 1 1 1 1 6 1

VIIa 1 1 1 1 1 7

Hierarchy 1 Effect Matrix (HI1) Hierarchy 2 Effect Matrix (HI2)

I+II III IVab IVc+VIIb V+VI VIIa I+II III IVab IVc+VIIb V+VI VIIa

I+II 1 2 2 2 2 2 I+II 1 1 1 1 1 2

III 2 1 1 1 1 2 III 1 1 1 1 1 1

IVab 2 1 1 1 1 2 IVab 1 1 1 1 1 1

IVc+VIIb 2 2 2 2 2 1 IVc+VIIb 2 1 1 1 1 1

V+VI 2 1 1 1 1 2 V+VI 1 1 1 1 1 1

VIIa 2 2 2 2 2 1 VIIa 2 1 1 1 1 1

Positive Affinity A Effect Matrix (AF2A) Positive Affinity B Effect Matrix (AF2B)

I+II III IVab IVc+VIIb V+VI VIIa I+II III IVab IVc+VIIb V+VI VIIa

I+II 1 2 1 1 1 1 I+II 1 1 2 1 1 1

III 2 1 1 1 1 1 III 1 1 1 1 1 1

IVab 1 1 1 1 1 1 IVab 2 1 1 2 1 1

IVc+VIIb 1 1 1 1 1 1 IVc+VIIb 1 1 2 1 1 2

V+VI 1 1 1 1 1 1 V+VI 1 1 1 1 1 2

VIIa 1 1 1 1 1 1 VIIa 1 1 1 1 2 1

National Variant

Negative Affinity for Japan (AF1-JAP)

I+II III IVab IVc+VIIb V+VI VIIa

I+II 1 1 1 1 2 2

III 1 1 1 1 1 1

IVab 1 1 1 1 1 1

IVc+VIIb 1 1 1 1 1 1

V+VI 1 1 1 1 1 1

VIIa 1 1 1 1 1 1

__________________________________________________________________________________________________

－ 36 －

－ 37 －

────────────────────────────

First Pattern]

Log Odds Ratio

0

Survey year

Figure 6 Three Patterns of the Trend in Log Odds Ratios

────

[

[Second Pattern]

Log Odds Ratio

0

Survey year

[

Third Pattern]

Log Odds Ratio

0

Survey year

────────────────────────────────

Figure 7 Unidiif Parameter in Age-Cohort Design

0

0.2

0.4

0.6

0.8

1

1.2

-1910 1911-1920 1921-1930 1931-1940 1941-1950 1951-1960 1961-

Birth cohorts

30-34

35-44

45-54

55-64

Figure 8 Models of Association between Class Origin and Education and between Education and Class Destination

________________________________________________________________________________

A. Origin-Education Effect Matrices

Education

______________________________________

Class Origin High Medium Low

I+II Professional-managerial class SO1 - SO2

III Routine non-manual class - - -

IVab Petty bourgeoisie - - MO2P

IVc+VIIb Farming class MO1 - MO2F

V+VI Skilled working class MO1 - MO2S

VIIa Unskilled working class MO1 - MO2U

B. Education-Destination Effect Matrices

Class destination

_______________________________________________________

Education I+II III IVab IVc+VIIb V+VI VIIa

High HQ1 - - HQ2 HQ2 HQ2

Medium - - - - - -

Low LQ1 - LQ2P LQ2F LQ2S LQ2U

________________________________________________________________________________

－ 38 －

Table 1 The Class Schema

O I I I I I I

V

VI

VI

VI

riginal Ten-category version Seven Sixcategory category

Higher grade professionals, administrators and officials; managers in large industrial establishments; large proprietors I+II I+II 'professional-managerial'I Lower-grade professionals, administrators and officials; higher-grade technicians; managers in small industrial establishments; supervisors of nonmanual employeesII Routine nonmaual employees in administration and commerce; III III 'routine nonmanual' sales personnel; other rank-and-file service workersVa Small proprietors, artisans etc. with employees IVa+IVb IVa+IVb 'petty bourgeoisie'Vb Small proprietors, artisans etc. without employeesVc Farmers and small holders; other self-employed workers in primary production IVc IVc+VIIb 'farming'

Lower-grade technicians; supervisors of manual workers V+VI V+VI 'skilled workers' Skilled manual workers

Ia Semi- and unskilled manual workers (not in agriculture etc.) VIIa VIIa 'unskilled workers'

Ib Agricultural and other workers (including family workers) VIIb in primary production

Table 2 Percentage Distributions of Class Origin and Class Destination

and Total Mobility Rates by Survey Year

1955 1965 1975 1985 1995 2005Class Origin: I+II 6.7 10.1 13.4 15.6 19.7 20.3 III 3.4 3.6 4.6 5.7 5.0 8.0 IVab 22.3 24.9 25.2 26.6 27.1 26.3 IVc+VIIb 60.3 51.7 48.1 37.1 30.7 23.1 V+VI 2.8 5.9 4.8 8.7 10.5 12.6 VIIa 4.5 3.8 3.9 6.3 7.0 9.7

lass Destination: I+II 9.9 19.8 24.5 30.0 37.3 36.7 III 10.4 10.7 9.9 10.2 8.5 11.2 IVab 22.5 20.9 21.1 21.2 21.2 16.2 IVc+VIIb 41.1 23.6 16.4 7.3 5.3 4.9 V+VI 7.1 15.0 16.6 18.6 17.4 17.8 VIIa 9.0 10.0 11.5 12.6 10.3 13.3 otal Mobility Rates: 47.7 60.3 65.4 68.5 68.3 68.6

Index of Dissimilarity

C

T

: 19.2 32.2 35.8 35.2 31.4 28.3 N (sample size) 1339 1379 1691 1628 1579 1594

－ 39 －

Table 3 Outflow Rates by Survey Year

F I+II III IV IV V+VI

V

Outflow Rates to Class Destination:I+II III IVab IVc+VIIb V+VI VIIa

rom Class Origin:

1955 39.8 23.2 15.5 15.5 3.5 2.41965 53.9 15.1 13.7 5.1 7.2 5.01975 54.0 12.8 18.6 3.1 7.5 4.01985 61.3 13.8 11.0 0.1 8.3 5.51995 66.8 5.8 13.5 0.4 9.3 4.22005 60.2 15.1 11.1 0.9 4.9 7.7

1955 13.6 30.7 26.3 11.2 6.8 11.41965 30.4 16.2 22.3 8.5 8.3 14.41975 33.5 11.6 23.4 4.1 16.9 10.51985 37.5 18.2 12.9 1.3 21.4 8.71995 44.6 15.4 14.1 2.8 15.3 7.82005 53.8 13.3 5.5 1.6 14.0 11.7

ab1955 12.7 14.4 43.1 9.7 11.4 8.71965 17.4 15.1 40.7 5.8 14.5 6.41975 25.0 12.4 36.8 3.1 13.6 9.21985 27.7 7.9 37.9 1.4 15.9 9.21995 30.5 8.2 34.3 0.7 18.4 7.92005 33.2 10.0 31.0 1.2 13.8 10.7

c+VIIb1955 5.7 6.1 15.7 60.8 4.6 7.21965 13.9 7.4 14.0 39.5 14.4 10.81975 15.5 8.0 14.4 30.5 16.8 14.91985 19.8 6.8 18.3 17.4 21.8 15.91995 24.1 7.4 21.6 15.2 17.3 14.42005 23.1 7.3 13.3 17.1 21.7 17.4

1955 8.4 16.5 22.2 3.2 27.6 22.21965 19.7 7.4 14.8 6.3 37.0 14.81975 20.9 7.5 18.4 3.8 44.3 5.01985 26.2 17.7 12.7 2.2 29.0 12.11995 37.5 12.1 10.9 0.7 25.5 13.32005 31.0 12.5 10.0 0.6 31.0 15.0

IIa1955 6.8 10.1 20.0 16.7 11.7 34.81965 17.1 13.5 11.6 13.7 19.2 25.01975 25.6 7.7 12.2 4.8 28.6 21.21985 20.5 14.6 12.8 3.1 19.6 29.41995 32.6 11.8 11.0 1.9 26.4 16.42005 22.6 11.6 10.3 2.0 32.2 21.3

－ 40 －

Table 4 Inflow Rates by Survey Year

To Class Destination: I+II III IV IV V+VI

V

Inflow Rates from Class Origin::I+II III IVab IVc+VIIb V+VI VIIa

1955 27.0 4.6 28.5 34.5 2.3 3.11965 27.4 5.5 21.9 36.1 5.8 3.31975 29.4 6.3 25.8 30.4 4.1 4.11985 31.9 7.2 24.5 24.5 7.6 4.31995 35.3 5.9 22.2 19.9 10.5 6.12005 33.3 11.8 23.8 14.5 10.6 6.0

1955 15.1 10.1 30.9 35.2 4.4 4.41965 14.3 5.4 35.3 36.1 4.1 4.81975 17.3 5.4 31.7 38.9 3.6 3.11985 21.0 10.2 20.4 24.6 15.0 9.01995 13.5 9.0 26.1 26.8 14.9 9.72005 27.5 9.6 23.6 15.2 14.0 10.1

ab1955 4.6 4.0 42.7 42.0 2.7 4.01965 6.6 3.8 48.6 34.7 4.2 2.11975 11.8 5.1 44.0 32.8 4.2 2.31985 8.1 3.5 47.4 32.1 5.2 3.81995 12.5 3.3 43.8 31.3 5.4 3.62005 13.9 2.8 50.4 19.0 7.7 6.2

c+VIIb1955 2.5 0.9 5.3 89.2 0.2 1.81965 2.2 1.3 6.2 86.6 1.6 2.21975 2.6 1.2 4.7 89.3 1.1 1.21985 0.3 1.0 5.1 88.3 2.6 2.71995 1.4 2.6 3.7 88.3 1.4 2.52005 3.9 2.7 6.6 81.3 1.5 4.0

1955 3.4 3.3 35.9 39.1 10.8 7.51965 4.8 2.0 24.1 49.7 14.5 4.91975 6.1 4.7 20.7 48.9 12.9 6.81985 6.9 6.6 22.8 43.5 13.5 6.61995 10.5 4.4 28.7 30.5 15.3 10.62005 5.6 6.3 20.4 28.2 21.8 17.6

IIa1955 1.8 4.3 21.5 48.1 6.8 17.51965 5.1 5.1 16.0 55.7 8.7 9.51975 4.6 4.1 20.0 61.9 2.1 7.21985 6.8 3.9 19.5 46.8 8.3 14.71995 8.0 3.7 20.8 42.9 13.5 11.12005 11.8 7.1 21.2 30.2 14.2 15.6

－ 41 －

Table 5 Outflow rates from Ⅳc+Ⅶb

Survey year N High Middle Low High Middle Low High Middle Low High Middle Low High Middle Low

1955 809 3.0 1.5 1.2 0.5 2.7 2.8 0.7 2.1 12.9 1.2 3.3 56.2 0.2 0.5 11.01965 714 4.3 4.3 5.2 0.0 2.4 5.0 0.1 1.8 12.0 0.3 3.2 36.0 0.1 2.9 22.11975 814 3.6 7.7 4.2 0.2 3.6 4.2 0.6 3.4 10.3 1.1 5.2 24.2 0.2 5.7 25.81985 605 6.6 11.1 2.1 1.2 3.8 1.8 0.8 5.8 11.7 0.5 5.8 11.1 0.7 14.9 22.11995 486 7.2 14.2 2.7 1.9 3.3 2.3 0.8 8.4 12.3 0.8 7.2 7.2 0.6 15.4 15.62005 368 9.5 11.4 2.2 1.1 4.9 1.4 1.4 5.7 6.3 1.6 10.1 5.4 1.4 25.5 12.2

to Class Destination and Education

Ⅴ+Ⅵ/ⅦaⅠ+Ⅱ Ⅲ Ⅳab Ⅳc+Ⅶb

Table 6 Outflow Rates from and Inflow Rates to Class V+VI and VIIa

Combined for Japan by Survey Year

ow Rates from Class V+VI and VIIa Combined Class Destination:

I+II III IVab IVc+VIIb V+VI/VIIa 1955 7.4 12.5 20.8 11.6 47.71965 18.7 9.8 13.5 9.2 48.81975 23.0 7.6 15.6 4.3 49.61985 23.8 16.4 12.8 2.6 44.41995 35.5 12.0 10.9 1.2 40.42005 27.3 12.1 10.1 1.2 49.3

low Rates to Class V+VI and VIIa Combinedm Class Origin:

I+II III IVab IVc+VIIb V+VI/VIIa 1955 2.5 3.9 27.9 44.1 21.61965 4.9 3.2 20.9 52.1 18.91975 5.5 4.5 20.4 54.3 15.41985 6.9 5.5 21.4 44.8 21.31995 9.6 4.1 25.8 35.1 25.42005 8.3 6.7 20.8 29.0 35.3

Outfl to

Inf

fro

Table 7 Outflow Rates from and Inflow Rates to Class V+VI/VIIa and Education　

Survey yea

r N High Middle Low High Middle Low High Middle Low High Middle Low High Middle Low

1955 96 3.1 2.1 2.1 3.1 3.1 6.3 0.0 2.1 18.8 1.0 0.0 10.4 0.0 3.1 44.81965 133 3.0 9.8 6.0 0.8 6.0 3.0 0.8 2.3 10.5 0.8 0.0 8.3 1.5 7.5 39.81975 147 4.8 10.9 7.5 0.7 5.4 1.4 0.7 7.5 7.5 0.0 1.4 2.7 0.7 11.6 37.41985 243 9.1 12.8 2.1 3.7 9.9 2.9 0.8 5.3 6.6 0.8 0.0 1.6 0.4 18.9 25.11995 275 13.5 20.7 1.5 2.9 8.0 1.1 0.0 6.2 4.7 0.0 0.4 0.7 0.7 20.4 19.32005 355 11.5 14.4 1.4 2.5 8.5 1.1 2.0 5.9 2.3 0.3 0.8 0.0 2.3 34.6 12.4

Surveyr N High Middle Low High Middle Low High Middle Low High Middle Low High Middle Low

1955 214 0.0 0.0 2.3 0.0 0.9 2.8 0.9 3.3 23.8 0.9 1.9 41.6 0.0 1.4 20.11965 345 0.6 1.7 2.6 0.0 0.9 2.3 0.6 3.2 17.1 0.3 6.1 45.8 0.6 2.9 15.41975 475 0.6 2.7 2.1 0.0 1.7 2.7 0.6 5.9 13.9 0.4 9.7 44.2 0.2 3.6 11.61985 508 1.0 4.7 1.2 0.2 2.6 2.8 1.8 9.4 10.2 0.8 17.7 26.4 0.2 9.1 12.01995 438 1.1 5.7 2.7 0.2 2.7 1.1 2.1 13.9 9.8 0.7 17.1 17.4 0.5 12.8 12.12005 496 2.2 5.2 0.8 0.6 4.0 2.0 2.0 12.7 6.0 1.0 19.0 9.1 1.6 24.8 8.9

from Class Origin and EducationⅠ+Ⅱ Ⅲ Ⅳab Ⅳc+Ⅶb Ⅴ+Ⅵ/Ⅶa

to Class Destination and Education

Ⅴ+Ⅵ/ⅦaⅠ+Ⅱ Ⅲ Ⅳab Ⅳc+Ⅶb

yea

－ 42 －

Table 8 Fit Statistics of the Constant Social Fluidity (CSF) Model and

(1)

(2)

(2a

(2b) Curvilinear chang (2c

(3)

(4)

( ( (

Other Models to the Origin by Destination by Year Japanese Table% %

G2 df p misc. reduction bic G

2 p Conditional Independence Model 2487.17 150 0.000 20.70 ---- 1117.80 ----

Constant Social Fluidity Model 163.50 125 0.012 4.31 93.43 -977.64 ----

) Linear change 163.01 124 0.011 4.29 93.45 -969.00 0.48 0.486

e 163.00 123 0.009 4.29 93.45 -959.88 0.50 0.780

) Unidiff 160.64 120 0.008 4.24 93.54 -934.86 2.86 0.722(1955) (1965) (1975) (1985) (1995) (2005)

1 0.965 0.995 1.082 0.975 1.054

Constant Social Fluidity Model 233.70 140 0.000 5.21 90.60 -1044.38 70.20 0.000 with Effect Matrices

(vs 3) Constant Social Fluidity Model 211.38 139 0.000 5.04 91.50 -1057.57 22.32 0.000

with Effect Matrices (Japanese variant)(vs 4)

4a) Variable Effect Matrices Model 134.64 84 0.000 3.22 94.59 -632.21 76.74 0.028

4b) Effect Matrices Unidiff Model 208.80 134 0.000 4.96 91.60 -1014.50 2.58 0.765(1955) (1965) (1975) (1985) (1995) (2005)

1 0.962 1.000 1.083 0.980 1.041

4c) Effect Matrices Model with Variable 173.63 133 0.010 4.27 93.02 -1040.54 37.75 0.000 Paramters for Selected Years

diff. vs 2

Table 9 Estimates of Origin-Destination Effect Matrix Parameters in the Origin

by Destination by Year Japanese Table (model 4c)model 4constant 1955 1965 1975 1985 1995 2005

Effect Matrix Parameter: ** ** DIG(I+II) 1.024 1.163 1.163 0.757 1.163 1.163 0.757 (0.133) (0.145) (0.145) (0.156) (0.145) (0.145) (0.156) DIG(III) 0.458 0.460 0.460 0.460 0.460 0.460 0.460 (0.141) (0.141) (0.141) (0.141) (0.141) (0.141) (0.141) ** DIG(IVab) 1.058 0.706 1.112 1.112 1.112 1.112 1.112 (0.072) (0.155) (0.075) (0.075) (0.075) (0.075) (0.075) ** DIG(IVc+VIIb) 2.420 2.542 2.080 2.542 2.542 2.542 2.542 (0.096) (0.110) (0.180) (0.110) (0.110) (0.110) (0.110) ** ** DIG(V+VI) 0.479 0.701 0.701 0.701 0.186 0.186 0.701 (0.092) (0.118) (0.118) (0.118) (0.141) (0.141) (0.118) ** ** ** DIG(VIIa) 0.563 0.931 0.931 0.260 0.931 0.260 0.260 (0.125) (0.169) (0.169) (0.163) (0.169) (0.163) (0.163) HI1 -0.099 -0.104 -0.104 -0.104 -0.104 -0.104 -0.104 (0.042) (0.042) (0.042) (0.042) (0.042) (0.042) (0.042) HI2 -0.286 -0.201 -0.201 -0.201 -0.201 -0.201 -0.201 (0.103) (0.081) (0.081) (0.081) (0.081) (0.081) (0.081) AF2A 0.538 0.523 0.523 0.523 0.523 0.523 0.523 (0.083) (0.083) (0.083) (0.083) (0.083) (0.083) (0.083) AF2B 0.255 0.249 0.249 0.249 0.249 0.249 0.249 (0.047) (0.047) (0.047) (0.047) (0.047) (0.047) (0.047) ** ** AF1J -0.497 -0.695 -0.695 -0.695 -0.263 -0.263 -0.695 (0.107) (0.136) (0.136) (0.136) (0.152) (0.152) (0.136)

** indicates that the parameter is significantly different from the CSF parameter.

model 4c

－ 43 －

Table 10 Trends in Log Odds Ratios between Two Adjacent Survey Years

5 1955-65 1965-75 1975-85 1985-95 1995-200

[First Pattern] 49% 44% 34% 49% 35%

ond Pattern] 32% 40% 42% 32% 49% Pattern] 19% 16% 24% 19% 16%

arameter -0.079 0.026 0.080 -0.095 0.081(0.075) (0.076) (0.078) (0.084) (0.083)

nge in G-square 1.033 0.111 1.012 1.268 0.956

[Sec[Third UniDiff p

St. errorCha

Table 11 Fit Statistics and Parameter Estimates of Log-linear and Log-multiplicative models

to the Origin by Destination by Birth Cohort Japanese Table

G2 df misc. bic

) conditional independence [OC, DC] 2411.9 150 20.29 1042.3) CSF with full association 177.5 125 4.16 -963.8) CSF using core (no national variant) 245.7 140 5.00 -1032.6) CSF using core national variant 219.5 139 4.73 -1049.6

DIG1 DIG2 DIG3 DIG4 DIG5 DIG6coef. 0.942 0.445 1.054 2.335 0.454 0.539

s.e. 0.134 0.141 0.073 0.097 0.092 0.126HI1 HI2 AF2A AF2B AF1J

coef. -0.105 -0.233 0.478 0.209 -0.538s.e. 0.042 0.081 0.083 0.047 0.107

diff with full association 172.1 120 3.99 -923.6diff with core national variant 212.8 134 4.57 -1010.7

(-1920) (1921-30) (1931-40) (1941-50) (1951-60) (1961-)unidiff 1 1.162 1.037 1.152 1.088 1.159

ODC

(1(2

(3 (4

(5) uni (6) uni

Table 12 Observed Cell Frequencies for Age-Period-Cohort Design

irth cohort

30-34 35-44 45-54 55-64 Total0 0 391 490 8810 420 362 265 1047

260 476 482 323 1541319 604 510 401 1834340 558 548 507 1953237 464 470 0 1171361 422 0 0 783

al 1517 2944 2763 1986 9210

Age B

1891-19101911-1920 1921-1930

1931-1940

1941-19501951-1960

1961-1975

Tot

－ 44 －

Table 13 Fit Statistics of the Constant Social Fluidity (CSF) Model and Other Models to the Origin by Destination with Cohort and Age Japanese Table

% %G-square df p misc. reduction bic G2 p

(1) Conditional Independence Model 2789.1 550 0.000 21.55 ---- -2236.0 ----

(2) Constant Social Fluidity Model 597.2 525 0.016 7.72 78.59 -4199.6 ----

(3) Unidiff (by age*cohort) 564.1 504 0.033 7.20 79.78 -4040.8 33.1 0.045

(4) Unidiff (by age) 592.3 522 0.018 7.64 78.76 -4177.0 4.8 0.185

(5) Unidiff (by cohort) 585.9 519 0.022 7.51 78.99 -4156.0 11.3 0.080

diff vs(2)

Table 14 Percentage Distributions of Education by Survey Year

1955 1965 1975 1985 1995 2005 High

11.0 11.4 15.1 22.9 27.2 32.7 ium 15.1 22.0 31.9 42.9 47.6 52.2

w 73.9 66.6 53.0 34.2 25.3 15.1 mple size) 1339 1379 1691 1628 1579 1594

le 15 Fit Statistics of the Constant Social Fluidity (CSF) Model and Other Models for the Origin by Education by Year Table and for the Origin by Destination by Year Table

% % G2 df p misc. reduction bic G2 p

y Education by Year Table: onditional Independence Model 1588.80 60 0.000 14.80 ---- 1041.06 ----

onstant Social Fluidity Model 61.02 50 0.137 2.64 96.16 -395.43 ----

) Unidiff 59.91 45 0.068 2.55 96.23 -350.90

onstant Social Fluidity Model 64.66 53 0.131 2.70 95.93 -419.18 3.64 0.303ith Effect Matrices

) Effect Matrix Unidiff Model 63.81 48 0.063 2.61 95.98 -374.39(1955) (1965) (1975) (1985) (1995) (2005)

1 0.975 0.999 0.937 0.991 1.013

% % G2 df p misc. reduction bic G2 p

by Destination by Year Tabl

Med

Lo

N (sa

Tab

Origin b(1) C

(2) C

(2a

(3) C w

(3a

Education e

onditional Independence Model 2913.79 60 0.000 20.70 ---- 2366.04 ----

onstant Social Fluidity Model 77.88 50 0.007 2.97 96.86 -378.58 ----

) CSF Model (1955 is different) 42.32 40 0.371 2.08 98.29 -322.84 35.56 0.000 onstant Social Fluidity Model 97.51 53 0.000 3.26 96.07 -386.33 19.64 0.000ith Effect Matrices (vs 3)

) CSF Model with Effect Matrices 59.60 46 0.086 2.47 97.60 -360.34 37.91 0.000is different)

(vs 3a)fect Matrices Model with Variable 63.31 48 0.068 2.63 97.45 -374.89 3.71 0.156

ameters for Selected Matrices

diff. vs 2

diff. vs 2

(1) C

(2) C (2a (3) C w (3a (1955 (4) Ef Par in 1955

－ 45 －

Table 16 Estimates of Origin-Education Effect Matrix Parameters in the

Effect Matrix

Origin by Education by Year Japanese Table

1955 1965 1975 1985 1995 2005 Parameter:

SO1 1.054 1.054 1.054 1.054 1.054 1.054 (0.077) (0.077) (0.077) (0.077) (0.077) (0.077) SO2 -0.552 -0.552 -0.552 -0.552 -0.552 -0.552 (0.152) (0.152) (0.152) (0.152) (0.152) (0.152) MO1 -0.706 -0.706 -0.706 -0.706 -0.706 -0.706 (0.071) (0.071) (0.071) (0.071) (0.071) (0.071) MO2U 0.774 0.774 0.774 0.774 0.774 0.774 (0.151) (0.151) (0.151) (0.151) (0.151) (0.151) MO2S 0.551 0.551 0.551 0.551 0.551 0.551 (0.145) (0.145) (0.145) (0.145) (0.145) (0.145) MO2F 1.200 1.200 1.200 1.200 1.200 1.200 (0.122) (0.122) (0.122) (0.122) (0.122) (0.122) MO2P 0.426 0.426 0.426 0.426 0.426 0.426

(0.124) (0.124) (0.124) (0.124) (0.124) (0.124)

Table 17 Estimates of Education-Destination Effect Matrix Parameters in the

Education by Destination by Year Japanese Table

1955 1965 1975 1985 1995 2005 Parameter:

** HQ1 1.731 1.027 1.027 1.027 1.027 1.027 (0.237) (0.072) (0.072) (0.072) (0.072) (0.072) ** HQ2 -0.060 -1.379 -1.379 -1.379 -1.379 -1.379 (0.317) (0.109) (0.109) (0.109) (0.109) (0.109) LQ1 -0.484 -0.484 -0.484 -0.484 -0.484 -0.484 (0.111) (0.111) (0.111) (0.111) (0.111) (0.111) ** LQ2U 2.655 1.541 1.541 1.541 1.541 1.541 (0.351) (0.113) (0.113) (0.113) (0.113) (0.113) ** LQ2S 2.395 1.388 1.388 1.388 1.388 1.388 (0.356) (0.105) (0.105) (0.105) (0.105) (0.105) ** LQ2F 2.710 1.824 1.824 1.824 1.824 1.824 (0.210) (0.118) (0.118) (0.118) (0.118) (0.118) LQ2P 1.205 1.205 1.205 1.205 1.205 1.205

(0.095) (0.095) (0.095) (0.095) (0.095) (0.095)

indicates that the parameter is significantly different from the CSF parameters.

Effect Matrix

**

－ 46 －

Table 18 Estimates of Education-Destination Effect Matrix Parameters in the

Ef

Perc

**

EllipN

Education by Destination by Origin by Year Japanese Table

1955 1965 1975 1985 1995 2005fect Matrix Parameter

** HQ1 1.636 0.950 0.950 0.950 0.950 0.950 (0.238) (0.074) (0.074) (0.074) (0.074) (0.074) HQ2 0.250 -1.195 -1.195 -1.195 -1.195 -1.195 (0.318) (0.111) (0.111) (0.111) (0.111) (0.111) LQ1 -0.461 -0.461 -0.461 -0.461 -0.461 -0.461 (0.112) (0.112) (0.112) (0.112) (0.112) (0.112) ** LQ2U 2.592 1.435 1.435 1.435 1.435 1.435 (0.354) (0.115) (0.115) (0.115) (0.115) (0.115) ** LQ2S 2.404 1.337 1.337 1.337 1.337 1.337 (0.360) (0.106) (0.106) (0.106) (0.106) (0.106) ** LQ2F 2.360 1.465 1.465 1.465 1.465 1.465 (0.211) (0.121) (0.121) (0.121) (0.121) (0.121) LQ2P 1.223 1.223 1.223 1.223 1.223 1.223

(0.097) (0.097) (0.097) (0.097) (0.097) (0.097)ent Reduction

HQ1 6 8 8 8 8 8 HQ2 … 15 15 15 15 15

LQ1 5 5 5 5 5 5 LQ2U 2 7 7 7 7 7 LQ2S 0 4 4 4 4 4 LQ2F 15 25 25 25 25 25 LQ2P 0 0 0 0 0 0

indicates that the parameter is significantly different from the CSF parameters.ses designate that the parameters were non-significant in the two-way table.

OTE: When the effect is increased after controlling for classs origin, the percent reduction is indicated as 0.

－ 47 －

Table 19 Estimates of Origin-Destination Effect Matrix Parameters After Controlling for

Eff P ** i

ElliN

Origin-Education and Education-Destination Associations and Percent Reduction in O-D Effect Parameters After the Control

1955 1965 1975 1985 1995 2005ect Matrix Parameter:

** ** DIG(I+II) 0.561 0.561 0.252 0.561 0.561 0.252 (0.152) (0.152) (0.162) (0.152) (0.152) (0.162) DIG(III) 0.276 0.276 0.276 0.276 0.276 0.276 (0.144) (0.144) (0.144) (0.144) (0.144) (0.144) ** DIG(IVab) 0.697 1.113 1.113 1.113 1.113 1.113 (0.156) (0.076) (0.076) (0.076) (0.076) (0.076) ** DIG(IVc+VIIb) 2.316 1.880 2.316 2.316 2.316 2.316 (0.111) (0.180) (0.111) (0.111) (0.111) (0.111) ** ** DIG(V+VI) 0.676 0.676 0.676 0.156 0.156 0.676 (0.119) (0.119) (0.119) (0.143) (0.143) (0.119) ** ** ** DIG(VIIa) 0.919 0.919 0.258 0.919 0.258 0.258 (0.170) (0.170) (0.164) (0.170) (0.164) (0.164) HI1 -0.078 -0.078 -0.078 -0.078 -0.078 -0.078 (0.043) (0.043) (0.043) (0.043) (0.043) (0.043) HI2 0.018 0.018 0.018 0.018 0.018 0.018 (0.084) (0.084) (0.084) (0.084) (0.084) (0.084) AF2A 0.241 0.241 0.241 0.241 0.241 0.241 (0.088) (0.088) (0.088) (0.088) (0.088) (0.088) AF2B 0.171 0.171 0.171 0.171 0.171 0.171 (0.048) (0.048) (0.048) (0.048) (0.048) (0.048) ** ** AF1J -0.567 -0.567 -0.567 -0.188 -0.188 -0.567 (0.141) (0.141) (0.141) (0.154) (0.154) (0.141)ercentage Reduction: DIG(I+II) 52 52 67 52 52 67 DIG(III) 40 40 40 40 40 40 DIG(IVab) 1 0 0 0 0 0 DIG(IVc+VIIb) 9 10 9 9 9 9 DIG(V+VI) 3 3 3 … … 3 DIG(VIIa) 1 1 1 1 1 1 HI1 26 26 26 26 26 26 HI2 100 100 100 100 100 100 AF2A 54 54 54 54 54 54 AF2B 31 31 31 31 31 31 AF1J 18 18 18 29 29 18

ndicates that the parameter is significantly different from the CSF parameter.pses designate that the parameters were non-significant in the two-way table.

OTE: When the effect is increased after controlling for education the percent reduction is indicated as 0.

－ 48 －

Intergenerational Mobility Patterns in Taiwan:

The Case of a Rapidly Industrializing Economy

Wei-hsin Yu (The University of Texas at Austin)

and

Kuo-hsien Su (National Taiwan University)

Abstract Previous research on class mobility has been disproportionately based on Western advanced countries. Much less known is how well the existing frameworks may be extended to late-industrializing countries, particularly those having experienced rapid economic growth and relatively condensed trajectories of industrialization. In this paper we examine intergenerational class mobility in Taiwan, a country that has undergone dramatic social and economic changes during last half century. Using combined data from multiple waves of the Taiwan Social Change Survey, we show that the core model of social fluidity, which is argued to capture the general similarities in mobility patterns across the industrial world, fits rather poorly in Taiwan. A further analysis indicates that this poor fit has to do with the implicit assumption of a stable class hierarchy by the core model. In Taiwan, the rapid economic changes have reshaped the relative difficulties of entering various social classes from fathers’ to children’s generations. Because of Taiwan’s condensed trajectory of industrialization, the relative prestige and desirability of a social class could alter substantially within merely one generation of time. We therefore need to allow the relative statuses of a few social classes to differ between fathers and children, in order to capture the patterns of social fluidity in Taiwan. Moreover, our analysis reveals that the trends and patterns of relative class mobility differ considerably by gender. In particular, there is a clear trend of increasing social fluidity for women, largely due to the loosening association between their class origin and educational attainment, but not for men. We attribute this gender difference to the fact that industrialization brought much greater improvement in educational and job opportunities for women than men in Taiwan. Overall, our findings provide little support for the argument of long-term stability and cross-national similarity in the underlying process of intergenerational class mobility. We argue that the rapid social economic changes largely account for the differences in Taiwan’s experience from those observed in Western industrialized countries. Given so, we believe that it is important for mobility research to pay more attention to the experiences of newly industrializing countries. Key words: social class, intergenerational mobility, social fluidity, education, Taiwan

1. Introduction Sociologists have long been interested in patterns of intergenerational mobility in industrial societies. While mobility research has often adopted a comparative approach, the theories guiding this research have been disproportionately derived from the

－ 49 －

experiences of Western advanced countries (Breen and Luijkx 2004; Erikson and Goldthorpe 1992; Featherman, Jones and Hauser 1975). Much less known is whether the existing frameworks can be extended to late-industrializing countries, particularly those having experienced rapid economic growth and thus relatively condensed trajectories of industrialization. In this paper we examine intergenerational class mobility in Taiwan, whose economy shifted from a primarily agricultural one in the 1950s to a drastically industrializing one in the 1970s and 1980s. By the 1990s, Taiwan, along with South Korea, Singapore, and Hong Kong, has been argued to have outgrown the “newly industrializing economy” status and become a newly industrialized society instead.

Taiwan’s miraculous speed of economic development makes it an interesting setting to test the possibility of generalizing the class mobility patterns found in more developed countries, whose industrialization has stretched over a longer period of time. Other than fast growth of new occupational opportunities, Taiwan’s condensed trajectory of development also implies dramatic changes in the economy and society. In particular, with the economy shifting its focus and mass education prevailing, the thresholds for entering different social classes may have altered with time, resulting in changes in the social prestige associated with each class. For example, when the majority of the society was illiterate and nonfarm job options were few, the differences between white-collar and skilled blue-collar workers in qualifications, income, and prestige were likely to be small. Such class differences are likely to increase substantially with economic development and educational expansion. As a result, within a relatively short period of time—from parents’ to children’s generations—the boundaries between classes may become much less penetrable. The shifts in class thresholds and prestige may also reshape the socially defined distance between a given pair of classes, hence altering the hierarchical relations among social classes. Thus, unlike countries that had undergone lengthier industrialization, Taiwan’s experience renders a possibility that redefinitions of class boundaries and hierarchy might occur from one generation to the next. Such redefinitions should play important roles in shaping the patterns of intergenerational class mobility.

Research on intergenerational mobility generally distinguishes individuals’ chances of absolute mobility from those of relative mobility. While absolute social mobility measures individuals’ rates of moving to classes other than their class origins, relative social mobility, also referred as social fluidity, captures individuals’ likelihood of altering classes after taking into account changes in the society’s occupational composition. The rapid changes that Taiwan has undergone can certainly be expected to increase the level of absolute mobility in the society, as the adjustments of occupational structures would have led many individuals to move to class destinations different from their origins. By contrast, the relation between a condensed trajectory of industrialization and shifts in individuals’ relative mobility rates is less clear. Previous research based on advanced industrial countries suggests that the underlying process of intergenerational mobility, observed by the advantages and disadvantages each social class passes onto the next generation net of changes in occupational structures, is similar across industrial societies and generally stable over time (Featherman, Jones and Hauser 1975). Tsay’s (1997; 1999) findings from Taiwan provide some support for the claim of cross-temporal stability. No research on Taiwan, however, has tested the applicability of the “core model of social fluidity,” which is argued to capture the general similarities in mobility patterns across the industrial world (Erikson and Goldthorpe 1992). In this sense, it remains questionable whether Taiwan’s pattern of relatively social mobility is similar to that derived from more industrialized countries. The lack of strictly comparable analyses also

－ 50 －

makes it difficult to conclude whether Taiwan’s long-term trend in relative social mobility would be similar to those found in Western industrialized countries (Breen and Luijkx 2004).

In this study we take three steps to provide a more comprehensive understanding of social mobility in Taiwan. First, we draw on descriptive statistics to assess changes in opportunity structures for men and women. Such an assessment sheds light on changes in individuals’ absolute chances of mobility. The second part of the analysis turns to relative social mobility. We test the hypothesis of a cross-national similarity in the pattern of social fluidity by fitting a modified version of the core model of social fluidity proposed by Ishida and colleagues (1995). We also address the question of how Taiwan’s pattern of fluidity may differ from this core model, if any. To test the hypothesis of cross-temporal stability, we further investigate whether and how the overall level of fluidity has changed for men and women during Taiwan’s rapid economic development. Specifically, we compare the chances of relative mobility among four birth cohorts of individuals who came of age during substantially different phases of industrialization. Finally, because Taiwan’s industrialization was accompanied by a dramatic expansion of educational opportunities (Huang 2001; Yu and Su 2006), we examine the roles education plays in shaping the level of social fluidity. This part of the analysis helps reveal whether education mediates the process of class mobility as much in newly industrialized economies as in more advanced industrial societies (see Ishida, Muller, and Ridge [1995] for example).

Throughout the statistical analysis we compare the patterns between men and women. Previous research on intergenerational mobility in Taiwan rarely pays attention to women’s mobility processes (e.g., Tsay 1997; 1999). While Taiwan’s overall level of female participation in the labor force has lagged behind those of many industrial societies in the West, there have been considerable increases in gender equality in education and employment opportunities in recent decades (Yu forthcoming). In fact, it is arguable that economic development has brought much greater level of improvement in Taiwanese women’s job opportunities than men’s. For this reason, we may expect the trends regarding class mobility to differ between men and women in Taiwan. We thus investigate the extent to which the mobility pattern of employed women resembles that of their male counterparts. Nevertheless, any conclusion drawn from such comparisons requires caution, as women are more likely to leave the labor force than men, and their decisions to leave the labor force are rarely made at random.

2. Background Taiwan’s industrialization is thought to begin before World War II, under Japan’s colonial rule (Cumings 1987). Nevertheless, the development of its industrial sector had been rather gradual until the early 1960s. As Figure 1 shows, Taiwan’s manufacturing industries expanded quickly and dominated the economy from the early 1960s to late 1980s. In the mean time, the percentage of workers in the agricultural sector shrank. By the late 1990s, no more than ten percent of the labor force held agricultural jobs. Since the early 1990s the share of the service sector has begun to surpass the manufacturing sector, signifying Taiwan’s transition to a postindustrial economy. Thus, during the last half century Taiwan essentially went from an agricultural to an industrial society, and then to a postindustrial society. These transitions brought substantial changes in

－ 51 －

occupational opportunities for both men and women. Joined with Taiwan’s rapid economic development was also an expansion of

educational opportunities (Woo 1991). From the cohort born in the early 1930s to that in the late 1960s, the average years of education nearly doubled for men and tripled for women (Huang 2001). The increases in the educational attainment, especially among women, are partly attributable to Taiwan’s rapid decrease in fertility rate since the 1960s. A dramatic demographic transition that considerably shrank the family size, in conjunction with the rapidly changing economic structure, has encouraged parents to focus on the “quality,” rather than the “quantity,” of children. The reduced family size also allowed parents to allocate financial resources to not only sons but also daughters, enabling girls to have equal educational opportunities to boys’ (Parish and Willis 1993; Yu and Su 2006).

While the female population on the whole still lags behind the male one regarding educational attainment, the gender gap in education is nearly closed for the recent generation. With the decrease in educational inequality, Taiwanese women’s employment opportunities also improved in both quantity and quality (Yu forthcoming). Not only have the gender gaps in occupational status and wages narrowed during last several decades in Taiwan, but these gaps have also closed at a relatively fast rate. By 1995, Taiwan’s female-to-male wage ratio reached 70%, which was higher than those in both Japan and South Korea, two countries with similar cultural values and paths of economic development (Brinton 2001).

The dramatic changes in educational and economic opportunities for men and women have important implications for the class structure in Taiwan. Unlike in countries with more stable industrial structures and educational opportunities, the compositions of upper, middle, and lower social strata are likely to have shifted over time in Taiwan. In particular, we expect that the status of skilled manual laborers and self-employed workers to have declined with Taiwan’s economic development. For the generations matured before World War II, skilled manual jobs required more than the average level of skills, as the majority of the population was involved in agricultural work and poorly educated. Such skill requirements made it likely that those made up the skilled manual class were relatively well off and that their social status was comparatively high. By the time their children came of age, however, the illiteracy rate had been largely reduced and the mandatory educational level had shifted from six to nine years. With the entire work-age population equipped with sufficient qualifications for skilled factory jobs, such jobs can no longer be considered as relatively privileged. Moreover, the influence of Confucianism on Taiwanese society has long attached a cultural value to education, as well as jobs requiring high educational credentials. At the same time, Taiwan’s history of development is featured by a focus on low-skilled, labor-intensive industries and exploitation of manual workers. The “intellectual” supremacy, along with the social image of low-skilled, poorly paid factory work, has caused further devaluation of the working class (regardless of its skill levels). As a result, the skilled labor class is likely to have slipped down the social hierarchy in postwar Taiwan, moving farther away from the upper class (i.e., service class). The devaluation of skilled manual jobs may be particularly pronounced among women, whose opportunities of routine nonmanual jobs, such as service and sales jobs, have increased rapidly with the expansion of the service sector in Taiwan.

Similarly, the relative difficulties of becoming self-employed also shifted over time. In early twentieth century, most of the self-employed workers were likely to be

－ 52 －

merchants, whose possession of skills and assets were both above the majority of the population, whose livelihood lied in farming. These merchants’ potential for wealth accumulation might even surpass that of many nonmanual employees in a time when professions were underdeveloped and modestly rewarded. The educational expansion in the postwar era nevertheless made the qualification requirements for the petite bourgeoisie class less difficult to achieve. The unique trajectory of industrialization in Taiwan has further contributed to the lowering of the threshold for entering the self-employed class. Not only was the economy’s focus on labor-intensive industries conducive to small-scale firms with low capital investment, but the state has also deliberately promoted small-to-medium enterprises in Taiwan (Hamilton and Biggart 1988). The dominance of small firms has ensured low barriers to starting one’s own business, leading to considerable flows of low-skilled workers to self-employment in seeking higher socioeconomic status (Yu and Su 2004). Hence, the self-employed as a social class had lost its semi-elite social image as Taiwan industrialized. While today’s self-employed workers in Taiwan are not necessarily worse off than lower-white-collar workers in terms of earnings, they generally work longer hours and experience greater hardship with economic fluctuations (Yu 2001). These characteristics have made self-employment increasingly undesirable among individuals from relatively high class backgrounds.

The history of rapid industrialization also leads to farmers’ relatively low social status in Taiwan. As the miraculous economic growth increased industrial wages, agricultural work quickly became a relatively undesirable option. The difficulty of making profit out of farming is also a result of the land reform in the 1950s, which allowed all farm workers to own their lands and virtually eliminated owners of large farmlands. Without large-farm owners, few working in the agricultural sector might have greater income potential than those working in the industrial or service sector in Taiwan. Farming thus becomes the last resort for those who have nearly no alternative occupational options.

Finally, social and economic changes in Taiwan also raise the likelihood of dissimilarity between class origins and destinations particularly for women. As stated earlier, women’s educational and economic opportunities have grown with Taiwan’s development. The increase in women’s employment has nevertheless concentrated on a few occupations that are consistent with the traditionally feminine image (e.g., teachers, secretaries, sales and service workers). As a result, the class of routine nonmanual workers, including clerical and sales workers, has become highly feminized over time. This feminization is also attributable to increase in service workers within the routine nonmanual class as the Taiwanese economy transits to a postindustrial one. In addition, the fact that women’s decisions to be in the labor force are not randomly distributed across occupations further raises the probability that their social classes will differ from their fathers’ (i.e., their class origins). Specifically, Taiwanese women with white-collar jobs are more likely to remain in the labor force after marriage and childbirth (Yu 2005). Because our class analysis for women excludes those away from the labor force, we are likely to find women’s class distribution tilt more toward the nonmanual classes, including professional, managerial, and routine nonmanual workers, than men’s. Just as women’s class distribution differs from men’s, their class destinations are also likely to differ from their origins, which are measured by their fathers’ occupations.

While a high level of dissimilarity between women’s class origins and destinations can be expected, it is more difficult to predict how the dissimilar

－ 53 －

occupational opportunities between fathers and daughters, as well as the selection mechanisms for women’s employment, may affect the pattern of social fluidity among women. Because women who have moved from lower classes to the nonmanual classes are more likely to stay in the labor force than those who have not, it is possible that we may overestimate social fluidity for women. That is to say, when women from manual classes follow their fathers’ steps to take manual jobs, they may choose to leave the labor force after marriage, thus lowering the estimate of relative immobility. We, however, have no reason to expect the impact of this selection bias to change over time. Previous research on Taiwanese women’s employment decisions shows relatively little cohort differences in their likelihood of participating in the labor force, after controlling for period differences in women’s occupational opportunities (Yu forthcoming). If the selection bias is consistent across women of different birth cohorts, then we may still be able to accurately estimate the long-term change in women’s chances of relative mobility, despite our overestimation of their overall fluidity at any given time.

3. Data and Methods

The data for this study are drawn from multiple waves of the Taiwan Social Change (TSC) surveys, conducted annually since 1989 by Academia Sinica in Taiwan. The TSC project collects two nationally representative surveys on separate topics every year. Despite some variation over time, each of the samples consists of approximately 2,000 men and women aged 20 and above. To maximize the cases available for the analysis, we pool together the samples from the TSC surveys that include sufficient details on respondents’ current jobs and their fathers’ occupations when they grew up. Such information is available in eight of the TSC surveys conducted in eight different years between 1992 and 2005.1 We further select respondents who were between 30 and 64 years of age to limit the analysis to working-age individuals who are likely to have completed schooling and mandatory military service.2 We restrict the analysis to those aged thirty or older also because they are more likely to have reached their final class destinations. To be able to identify respondents’ class destinations, we also exclude those who did not have a current occupation or failed to report it. After these selections, the final sample consists of 9,467 men and 6,130 women.

The variables included in the analysis are class origin, class destination, birth cohort, and education. Class origin refers to the father’s class when a respondent was growing up. We use the detailed information on respondents’ current jobs to measure their class destinations. We code both variables using a modification of the class schema developed by the CASMIN project. This revised classification contains six categories, instead of the original seven categories, of social class: I+II (service class), III (routine nonmanual class), IVab (self-employed workers), V+VI (manual supervisors and skilled workers), IVc+VIIb (farmers and farm workers). The only difference our classification has is that farmers and farm workers are combined into one class. The reason for this 1 Specifically, the pooled sample is from both surveys in 1992 and 1997, Survey I in 1994, Survey II in 1996, Survey I in 1999, Survey I in 2002, Survey II in 2004, and Survey III in 2005 (also known as the East Asia Social Stratification and Social Mobility Survey—Taiwan). 2 In Taiwan, all men over age 20 are obligated to serve for at least two years in military. Because those in school can postpone the military service until finishing schooling, the actual age when men complete this service may vary.

－ 54 －

change is that the number of farm workers has been very small since the land reform in the 1950s. Hence, separating farmers from farm workers for Taiwan is not really meaningful. For convenience we refer the combined group of workers as farmers hereafter.

The sample is divided into four birth cohorts corresponding to Taiwan’s industrialization phases and changes in educational opportunities, both of which may affect mobility patterns. These cohorts of individuals were born in: 1945 and earlier, 1946-55, 1956-65, and 1966-75. Born before the end of World War II, a portion of the cohort 1945 and earlier was born and educated in Mainland China and then followed the Kuomintang regime to Taiwan after the regime lost the civil war in 1949. Respondents born in 1946-55 came of age as Taiwan began its rapid industrialization. Those born after 1956-65 were the first to benefit from the extension of mandatory education from six to nine years. The last cohort in our sample, whose birth years were between 1966 and 1975, experienced the expansion of the tertiary sector and the economy’s shifting focus to capital- and technology-intensive industries as they entered adulthood.

Education is measured by three categories based on the highest level of school that respondents reported to have attained. Specifically, those with up to middle school of education are considered as having a low educational level. Those who have completed high schools are considered as having a middle level of education. A high educational level refers to receiving education beyond high school (i.e., some tertiary education). Our statistical analysis proceeds as follows. First, we apply the common model of social fluidity proposed by Ishida and colleagues (1995) to the case of Taiwan. For simplicity, we refer this model as the core model hereafter. Second, we assess the differences in the mobility pattern between Taiwan and advanced industrial societies. Specifically, we develop a model to take into account forces pertaining to the Taiwanese context and test whether this model improves the core model significantly. Third, we use a uniform-difference (UNIDIFF) model, or known a multiplicative layer effect model, to examine cross-temporal differences in the intergenerational mobility in Taiwan (Erikson and Goldthorpe 1992; Xie 1992). Finally, we refine our examination of changes over time by controlling for the mediating roles of education. In particular, we analyze changes in the association between class origins and educational attainment, as well as between educational attainment and class destinations. We also investigate how controlling for the mediating roles of education helps reshape the associations between fathers’ and children’s social classes.

4. Descriptive Statistics

Table 1 presents the distributions of class origins and destinations, as well as the rate of total mobility and the index of dissimilarity, by gender and birth cohort. One clear trend shown in the table is the sharp decline in the proportion of farmers in the society, which is consistent with Figure 1. Regardless of birth cohort and gender, the share of the farming class in the father’s generation is substantially greater than that in the child’s generation. By contrast, the percentage of the service class, including managerial and professional workers, is considerably greater in the child’s generation than that in the father’s generation. In addition, as expected, the percentage of women in the routine, nonmanual class is larger than that of their fathers. This father-daughter difference also becomes greater among more recent cohorts. The dissimilarity index for women indicates

－ 55 －

that the overall differences between fathers’ and daughters’ classes increased with the onset of Taiwan’s economic development. The index for those born after 1945 is clearly larger than that for those born before the end of the War. The level of dissimilarity between fathers’ and daughters’ classes rose again as Taiwan began its postindustrial phase, which was approximately when women born 1966-75 reached their class destinations.

The level of dissimilarity between class origins and destinations is generally lower for men than for women in Taiwan. Men born in 1945 and earlier, however, have more similar class destinations to their class origins, compared to those of later cohorts. We attribute this cohort difference to Taiwan’s rapid industrialization since the 1960s, as those born up to 1945 are likely to have entered their class destinations by the mid-1960s. The mobility rates revealed in Table 1 presents a similar story. A Taiwanese man’s chance of moving away from his father’s class appears to increase with Taiwan’s transition to an industrial economy. Nevertheless, the mobility rates barely change among the three birth cohorts of men who were born after the War. This similarity suggests relatively small change in men’s absolute chances of mobility once Taiwan’s rapid industrialization began.

Tables 2 and 3 provide outflow and inflow mobility rates for men and women of various birth cohorts. The differences in the rates between cohorts illustrate the temporal changes in social mobility in Taiwan. According to Table 2, outflows to the farming class have decreased drastically for both men and women over time, as a result of the contraction of the agricultural sector. Another class to which the outflow rate has declined substantially is the self-employed class. This finding is consistent with the argument that the self-employed class has gradually lost its charm in Taiwan’s process of industrialization. This tendency is particularly evident for men. There are also increases in the outflows to the skilled manual class among men, perhaps resulting from the expansion of the manufacturing sector with Taiwan’s development. Moreover, the outflow rate from the working class (V+VI and VIIa) to the service class has increased with time, suggesting a rise in upward mobility with the economic development.

Trends regarding Taiwanese men’s inflow rates are somewhat parallel to those indicated by outflow statistics. According to Table 3, inflows from the farming class have declined among men, while inflows from the working class have increased over time. By contrast, Taiwanese men’s inflow rates from the white-collar class, including the service and routine nonmanual classes, have been relatively stable.

Table 2 also indicates rises in Taiwanese women’s outflows to the service class as well as the routine, nonmanual class. Such trends corroborate the argument regarding the long-term feminization of routine, nonmanual occupations in the Taiwanese labor market. The particularly large growth in female participation in clerical occupations has made these occupations increasingly “suitable” for women, which further accelerates female concentration in these occupations. Also noteworthy is that the increase in women’s outflow rates to the service class is somewhat greater that in men’s. This gender gap is in part explained by the fact that women with white-collar occupations, particularly professional or managerial jobs, are more likely to choose to stay in the labor force in Taiwan. This mechanism leads to an over-selection of women with high occupational attainment into our sample and thus higher outflow rates to the service class than those among men.

In addition to the selection process, education also plays a part in explaining women’s more frequent outflows to the service class. Table 4 presents the distributions of

－ 56 －

men and women in the sample by their educational levels and birth cohorts. Not only has the increase in the educational level of employed women been greater than that of their male counterparts, but the educational credentials of the female workers are also slightly higher than those of male workers for the most recent cohort (born 1966-75). This considerable growth of women’s educational attainment should have contributed to their rates of outflows to the service class. Similarly, Table 3 also shows that the inflow rate from the working class to the service class has risen with the overall improvement of women’s education across time.

5. Modeling Social Fluidity in Taiwan To demonstrate the pattern of relative mobility, or social fluidity, in Taiwan, we extend beyond descriptive statistics and fit a series of log-linear models to the mobility tables for men and women. Table 5 shows the results from fitting these models. The first row in the table presents the statistics for the independent model. As a baseline model, the independence model assumes that a respondent’s class destination is independent of the father’s class and completely determined by the occupational distribution in the society. The table also includes the results from fitting the core model of social fluidity to the data (“CSF + original core”). This model adds several effect matrices to the independence model in order to capture the influences of the father’s class on the child’s class. These influences are thought to result from the existence of social hierarchy, the inheritance of family assets and class-specific advantages and disadvantages, and the affinities between certain occupations (Erikson and Goldthorpe 1992; Ishida, Muller and Ridge 1995).

The comparison between the independence and core models indicates that taking into account the hierarchy, inheritance, and affinity effects helps improve the model fit considerably. The improvement is significant for both men and women. Nevertheless, the p-value for the G2 remains smaller than .05 for both groups, suggesting that the model fit is not entirely ideal. A new model is therefore needed to fully take into account forces that are particularly relevant to the Taiwanese context.

We attempt to improve the model of social fluidity by making two major changes (see Appendix 1 for details). Our first step is to revise the hierarchy effect matrices in the core model to better capture the class hierarchy in Taiwan. We argue that Taiwan’s rapid economic development has not only altered the occupational opportunities but also reshaped where each social class locates in the overall social strata. In particular, we contend that the skilled manual class has become less desirable and prestigious from the father’s to the child’s generations. We therefore propose asymmetrical class hierarchies for fathers and children, as shown in Table 6. In the father’s generation, we assume that the skilled manual class belongs to the middle stratum, whereas the same class falls to the lower stratum by the time the child’s generation reaches thirty years of age and older. In addition, although in the core model the farming class is thought to be positioned in the middle of the social strata, we expect this class to belong to the lower stratum in Taiwan. Not only are the educational requirements for farmers low, but their economic status is generally inferior, given the lack of large-land owners among this group. During Taiwan’s economic development the state also aggressively promoted the industrial sector in replacement of the agricultural sector, thus reducing the social desirability of becoming farmers.

－ 57 －

While we propose that the self-employed class remain in the middle of the social strata from the father’s generation to the child’s, we nevertheless argue that over time this class has moved down the social hierarchy slightly. That is to say, for the father’s generation jobs of the self-employed might be closer to professional, managerial, or other nonmanual jobs in terms of their qualification requirements, income potential, and social desirability than they are for the child’s generation. Following this assumption we make the second major change to the core model. Specifically, we remove the positive affinity from the service class to self-employed class, but retaining the one in the opposite direction. In the core model, the positive affinity is assumed to exist in both directions because the service and self-employed classes share the feature of possessing assets. In the case of Taiwan, however, we argue that those from the service class may not be particularly likely to enter self-employment because the self-employed class has become less prestigious over time.

We also remove the positive affinity from the self-employed class to the farming class because in the father’s generation the self-employed class is relatively well off and thus unlikely to have its offspring to fall to the least prestigious class. For the opposite direction, from farming to self-employment, we nevertheless assume a positive affinity as in the core model. By the time the child’s generation matured, self-employment has become a popular option for many relatively unskilled workers in Taiwan. Hence those from farming backgrounds may very well become self-employed. In addition, we entered a positive affinity from the self-employed class to the routine, nonmanual class. Because self-employment tended to be associated with relatively high socioeconomic status among the previous generations, having a self-employed father has been found to be associated with greater educational attainment for the current generation (e.g., Yu and Su 2006). This educational advantage is likely to enhance a child’s possibility of entering both the service and routine, nonmanual classes.

The other adjustments made to the positive affinity matrix have to do with the associations between the farming and working classes. Because women have far lower likelihood than men to remain in manual jobs throughout their lives, our adjustments vary slightly by gender. Similar to the core model, we enter a positive affinity from the farming class to the skilled manual class, as well as from skilled to unskilled manual workers, for men. We nevertheless remove the affinity from the farming class to the unskilled manual class based on our empirical finding. Perhaps for men from farming families, it is always possible to work at their family farm. Those unable to obtain better than unskilled manual jobs may therefore opt for farm work instead. By contrast, women are less likely to work at their father’s farm, especially after marriage. Therefore, women from farming families are particularly likely to become both skilled and unskilled manual workers. We remove the positive affinity from the skilled manual class to unskilled manual class for women, however. As stated earlier, the skilled manual class in Taiwan should be considered to belong to the middle stratum for the father’s generation. Following this logic, those originated from this class would consider moving the unskilled manual class as a great slide in social status. While men who have moved downward as such may still stay in the labor force, women who experience the same mobility may find such jobs beneath them and thus choose to leave the labor force. For this reason, we are unlikely to find a positive affinity from the skilled to unskilled manual classes among women.

The results from the Taiwan model of social fluidity, as we proposed, are shown in Table 5, right below those from the core model. For both men and women, the new

－ 58 －

model significantly improves the core model and yields a satisfactory fit (△G2>100 for both gender groups; p=.165 for men and p=.120 for women). The improvement made by the Taiwan model can also be observed in Table 7, which provides the coefficients for the effect parameters included in the core and Taiwan models, respectively. The results from the core model indicate that the hierarchy effect is small for Taiwanese men. The barriers to mobility only exist between social strata that are far from each other, such as between the service class and the unskilled manual class (HI2). However, the hierarchy effect for moving between classes belonging to adjacent social strata becomes statistically significant in the Taiwan model, after the modification of the class hierarchy. This finding suggests that our revision of the hierarchy matrices better captures the hierarchical relations among social classes in Taiwan. The hierarchy effect from the Taiwan model is nevertheless consistent with that from the core model in the sense that the long-range hierarchy effect is generally greater than the hierarchy effect between two adjacent social strata. Thus, the barriers to a long-range upward move, such as from the lower to upper strata, are more than double those to a move to the immediately higher social stratum.

Other effects are more similar in the core and Taiwan models. Generally speaking, one’s class origin affects the current class through one’s inheritance of the tangible and intangible family assets. Positive affinities are also found between certain class categories as hypothesized. Nevertheless, some gender differences exist in the mobility patterns in Taiwan. According to the coefficients for the Taiwan model, women are unlikely to replicate their fathers’ classes if their fathers were self-employed or held skilled or unskilled manual jobs. The gender differences in the inheritance effects are statistically significant for the self-employed and skilled manual classes (p<.05 by t-test). The lack of inheritance effect for women from the self-employed class has to do with the fact that the patriarchal culture in Taiwan prescribes that sons, rather than daughters inherit family enterprises. Even when direct inheritance of family enterprises is not the case, previous research has found that Taiwanese women’s entry into self-employment is more likely to be related to their husbands’ employment options, whereas men’s more likely to their father’s employment status (Lu 2001; Yu 2004; Yu and Su 2004). As to the gender difference in the inheritance effect for the skilled manual class, perhaps the increasing occupational segregation by gender has made skilled manual jobs men’s jobs, while routine, nonmanual jobs women’s. This trend makes it unlikely for women from the skilled manual class to follow their fathers’ steps in occupational choices.

So far we have shown how Taiwan’s pattern of social fluidity differs from the core model suggested for advanced economies. The next task is to investigate whether Taiwan’s mobility pattern has changed over time, as well as how this trend may differ between men and women. We explore the possibility of cross-cohort changes by fitting a UNIDIFF model to the three-way mobility table (i.e., class origin by destination by birth cohort). The UNIDIFF model treats birth cohort as an additional variable and test whether there is any cohort-specific divergence from the general origin-destination association patterns. We also introduce the effect matrices from the Taiwan model to further improve the fit. Table 5 presents the fit statistics for the UNIDIFF models for men and women. We also present parameter Φ for each cohort, which indicates the extent to which each cohort deviates from the overall level of association.

The UNIDIFF models with the Taiwan version of effect parameters further reduce G2 for both men and women, but the improvement is only statistically significant for women. That is to say, there would not be a great loss if we assume no difference in the

－ 59 －

strength of the origin-destination association between birth cohorts for men. However, the Φ’s for earlier cohorts are greater in magnitude even for men, suggesting that Taiwanese society has become increasingly fluid regarding intergenerational mobility. The trend is more apparent among women. The chance of relative social mobility has improved significantly for each cohort of Taiwanese women.

6. The Roles of Education The third and final part of our analysis explores the roles of education in the mobility process in Taiwan. In particular, we examine the associations between class origin and education and between education and class destination. As described earlier, we divide education into three levels, high, middle, and low. We use the same effect matrices proposed by Ishida and colleagues (1995) to first model the associations between different class origins and levels of schooling. Table 8 reports the fit statistics for such models. For both gender groups, the specified model improves the fit of the independence model substantially. The overall fit of the model, however, is less satisfactory for women as for men (p=.093 for men and p<.001 for women).

Table 9 reports the coefficients from the origin-education model just described. The general patterns are similar for men and women. Those from the service class are very likely to achieve high educational levels, while unlikely to have low educational levels. At the same time, those from farming or working-class backgrounds are disadvantaged in obtaining high levels of education, but rather likely to have low educational attainment. The educational disadvantages are greatest for the farming class, followed by the skilled manual class, unskilled manual class, and self-employed class, in that order. This ranking is rather consistent with our argument about the relative socioeconomic status among these social classes for the father’s generation. Furthermore, the class advantages and disadvantages in education appear to be more pronounced for women than men. The unequal resource allocation between sons and daughters accounts for this gender difference. Women from families with insufficient financial resources to support both sons and daughters are particularly likely to suffer from parents’ overall preference for educating sons. By the same token, Taiwanese women from the upper class benefit more from their class origins than others, as their parents can afford to invest much in their education as well as their brothers’.

Turning back to Table 8, we also use UNIDIFF models to test whether the associations between class origin and education for men and women have changed across birth cohorts. In the table we report the fit statistics and UNIDIFF parameter for each gender group. The trends revealed here are similar to those concerning the strength of association between class origin and destination. The class advantages and disadvantages in educational attainment have not varied significantly for men, but they have become less likely to affect women’s educational levels over time. As the financial conditions of Taiwanese families improve with reduced fertility and economic growth, perhaps parents no longer need to choose between sons and daughters about educational investment. As a result, class origin plays an increasingly less important role in shaping women’s educational opportunities.

Next, Table 10 shows the statistics for models estimating pattern in which education is associated with class destination. Here we also include the effect matrices as proposed by Ishida and colleagues (1995). Additionally, we introduce a UNIDIFF

－ 60 －

parameter to the education-destination (ED) models. This parameter enables us to reveal cross-cohort differences in the strength of the association between education and class destination. The fit statistics for the ED models indicate that the model design is not entirely satisfactory. Nevertheless, for both men and women, the UNIDIFF models do improve the original models significantly. For women, the association between education and class destination has become weaker over time. The trend is similar for men, except that for the most recent cohort the association between men’s educational levels and class destinations appear to have tightened again.

Our last concern is whether taking into account the mediating roles of education may affect the trends revealed earlier regarding social fluidity in Taiwan. To shed light on this issue, we first fit the Taiwan model of social fluidity to men’s and women’s mobility tables, while controlling both the associations between origin and education and between education and destination. Table 11 presents the fit statistics of this model by gender, and Table 12 shows the effect parameters from this model. The model fit both men’s and women’s mobility patterns well. The coefficients of the effect parameters are similar to those from the Taiwan models without controlling for education (as in Table 7). One exception, however, is that the inheritance effects for the service class become small and nonsignificant for both gender groups after taking education into consideration. This result suggests that children from the service class are able to “inherit” their fathers’ social class entirely because of their class advantages in educational attainment.

To examine the roles of education in shaping trends in social fluidity, we add a UNIDIFF parameter to the Taiwan models of social fluidity, while controlling for education. The fit statistics included in Table 11 indicate no significant improvement with the UNIDIFF models for either gender. This result is consistent with what we have found for men before controlling for education, but not for women. We have observed a rise in women’s fluidity when the roles of education were not considered. How do we explain this gender difference? Here, we should note that our findings have revealed that for both men and women, the association between education and class destination has weakened over time. Only for women, however, the association between class origin and educational attainment has become looser. Taken together, these findings suggest that the loosening association between class origin and education is likely to be responsible for women’s changing rates of social fluidity. That is to say, women’s chances of relative mobility increase as their class origins become less relevant to their educational attainment. By contrast, as class advantages and disadvantages on men’s educational attainment remain rather stable, we generally observe no change in their relative mobility rates.

7. Conclusions We began this paper by asking whether the existing frameworks about class mobility, generally derived from the experiences of Western advanced countries, can be readily extended to a rapidly and newly industrialized country like Taiwan. Our answer to this question is negative. The main difference of Taiwan’s experience has to do with its patterns of social fluidity. To elaborate, our analysis of Taiwan’s absolute mobility rates indicate a greater degree of openness with industrialization, which is consistent with the trend observed in many other countries. We, however, failed to find a similar underlying process of social mobility with that of more industrialized countries when fitting the core

－ 61 －

model of social fluidity to the Taiwan data. In fact, our analysis revealed that the fit of the core model was rather poor. We argue that this poor fit has to do with the implicit assumption of stable class structures by the core model. Although the focus on relative class mobility, or social fluidity, has taken into account changes in the relative share of each social class over time, rarely considered is the possibility that the relative prestige, and thus advantages, of a social class may also alter as the broader social context changes. In Taiwan, the rapid economic changes have reshaped the relative difficulties of entering various social classes from fathers’ to children’s generations. As the threshold to a given social class increases or decreases, the desirability and social influences of this class may change accordingly. Because of Taiwan’s condensed trajectory of industrialization, the relative prestige and desirability of a social class could alter substantially within merely one generation of time. To capture the patterns of social fluidity in Taiwan, we have allowed the relative statuses of a few social classes to differ between fathers and children. The fact that fathers and children face asymmetrical class hierarchies, as well as class advantages and disadvantages, makes the Taiwan case unique and interesting. In addition, our analysis has also indicated asymmetry in the patterns and trends of social fluidity between men and women. Not only is it necessary to model women’s patterns of social fluidity somewhat differently from men’s, but the trend of increasing social fluidity can be observed only among women. As our cohort analysis has shown, for every younger cohort of women, their chances of moving away from their class origin have increased. Different cohorts of men, however, have faced similar chances with respect to their relative class mobility. Our later analysis that takes into account the roles of education suggests that women’s enhanced likelihood of moving across class boundaries largely results from the loosening link between class origin and educational attainment for this gender group. That is to say, with the educational expansion and declined family size, education became no longer a prestige preserved for girls from the upper class. Such changes in educational opportunities have benefited women far more than men, as Taiwanese parents from all social classes have always viewed educating sons as a necessity. As a result, the class influences on women’s educational attainment have decreased over time, leading to a wider range of occupational options for women from lower class backgrounds. By contrast, because class influences on sons’ educational attainment have remained relatively constant, the relative occupational opportunities for men from a given class background have been unchanged across different cohorts.

Overall, our findings provide little support for the argument of long-term stability and cross-national similarity in the underlying process of intergenerational class mobility. While previous research based on Western industrialized countries often find the pattern of social fluidity to be stable over time, the Taiwan case generally reflects anything but stability. We argue that the rapid social economic changes largely account for the differences in Taiwan’s experience from those observed in Western industrialized countries. By the same token, studies in other rapidly industrializing countries are likely to reveal interesting differences from the more stable trends observed in countries whose industrialization stretched over a much longer period of time. Findings from this research thus reveal the need for mobility research to pay more attention to newly industrializing countries.

－ 62 －

References Breen, Richard, and Ruud Luijkx. 2004. "Social Mobility in Europe between 1970 and 2000." Pp. 37-76 in Social Mobility in Europe, edited by Richard Breen. Oxford: Oxford University Press. Brinton, Mary C. 2001. "Married Women's Labor in East Asian Economies." Pp. 1-37 in Women's Working Lives in East Asia, edited by Mary C. Brinton. Stanford, California: Stanford University Press. Cumings, Bruce. 1987. "The Origins and Development of the Northeast Asian Political Economy: Industrial Sectors, Product Cycles, and Political Consequences." Pp. 44-83 in The Political Economy of the New Asian Industrialism, edited by Frederic C. Deyo. Ithaca: Cornell University Press. Erikson, Robert, and John H. Goldthorpe. 1992. The Constant Flux: A Study of Class Mobility in Industrial Societies. Oxford: Clarendon Press. Featherman, David L., Lancaster F. Jones, and Robert M. Hauser. 1975. "Assumptions of social mobility research in the U.S.: The case of occupational status." Social Science Research 4:329-360. Hamilton, Gary G., and Nicole Woolsey Biggart. 1988. "Market, Culture, and Authority: A Comparative Analysis of Management and Organization in the Far East." American Journal of Sociology 94:52-89. Huang, Fung-Mey. 2001. "Education, Earnings, and Fertility in Taiwan." Pp. 279-299 in Population Change and Economic Development in East Asia: Challenges Met, Opportunities Seized, edited by Andrew Mason. Stanford, CA: Stanford University Press. Ishida, Hiroshi, Walter Muller, and John M. Ridge. 1995. "Class Origin, Class Destination, and Education: A Cross-National Study of Ten Industrial Nations." The American Journal of Sociology 101:145-193. Lu, Yu-Hsia. 2001. "The "Boss's Wife" and Taiwanese Small Family Business." in Women's Working Lives in East Asia, edited by Mary C. Brinton. Stanford, CA: Stanford University Press. Parish, William L., and Robert J. Willis. 1993. "Daughters, Education, and Family Budgets: Taiwan Experiences." The Journal of Human Resources 28:863-898. Tsay, Ruey-Ming. 1997. "Leaving the Farmland: Class Structure Transformation and Social Mobility in Taiwan." Pp. 15-55 in Taiwanese Society in the 1990s, edited by Ly-Yun Chang, Yu-Hsia Lu, and Fu-Chang Wang. Taipei: Institute of Sociology, Academia Sinica. —. 1999. "A Structural Analysis of Social Mobility in Taiwan, The United States, and

－ 63 －

Japan." Taiwanese Journal of Sociology 22:83-125. Woo, Jennie Hay. 1991. "Education and economic growth in Taiwan: A case of successful planning." World Development 19:1029-1044. Xie, Yu. 1992. "The Log-Multiplicative Layer Effect Model for Comparing Mobility Tables." American Sociological Review 57:380-395. Yu, Wei-hsin. 2001. "Taking Informality into Account: Women's Work in the Formal and Informal Sectors in Taiwan." Pp. 233-262 in Women's Working Lives in East Asia, edited by Mary C. Brinton. Stanford, CA: Stanford University Press. —. 2004. "Gender, Family, and Form of Labor Force Participation: Woman and Nonstandard Employment in Japan and Taiwan." Pp. 27-59 in Old Challenges, New Strategies? Women, Work, and Family in Contemporary Asia, edited by Leng Leng Thang and Wei-hsin Yu. Leiden and Boston: Brill Academic Publishers. —. 2005. "Changes in Women's Postmarital Employment in Japan and Taiwan." Demography 42:693-717. —. Forthcoming. Gendered Trajectories: Women, Work, and Social Change in Japan and Taiwan. Stanford, CA: Stanford University Press. Yu, Wei-hsin, and Kuo-hsien Su. 2004. "On One's Own: Self-Employment Activity in Taiwan." Pp. 388-425 in The Reemergence of Self-employment: A Comparative Study of Self-Employment Dynamics and Social Inequality, edited by Richard Arum and Walter Müller. Princeton, NJ: Princeton University Press. —. 2006. "Gender, Sibship Structure, and Educational Inequality in Taiwan: Son Preference Revisited." Journal of Marriage and Family 68:1057-1068.

－ 64 －

Figure 1. Percentages of Taiwanese Workers in Services, Manufacturing, and Agricultural Industries, 1951-2000

Taiwan

0%

10%

20%

30%

40%

50%

60%

70%

80%

90%

100%

1951 1955 1959 1963 1967 1971 1975 1979 1983 1987 1991 1995 1999

Year

Perc

enta

ge ServicesManufacturing Agriculture

Source: Directorate-General of Budget, Accounting and Statistics, Executive Yuan, R.O.C., Yearbook of Manpower

Survey Statistics, Taiwan Area, Republic of China, various years.

－ 65 －

Tabl

e 1.

Cla

ss D

istri

butio

ns b

y B

irth

Coh

ort a

nd G

ende

r

Men

W

omen

C

ohor

t 19

28-4

5 19

46-5

519

56-6

519

66-8

6To

tal

1928

-45

1946

-55

1956

-65

1966

-86

Tota

l N

1,

780

2,

928

3,43

8 1,

321

9467

82

1

1,84

1 2,

494

974

6,

130

Cla

ss D

estin

atio

n

I+

II

18.9

26

.4

29.2

36

.5

27.4

9.

6

18.8

23

.0

35.9

22

.0

IIIa

b 4.

7

6.7

7.

5

9.4

7.

0

11.5

18

.1

27.3

34

.1

23.5

Iv

ab

20.9

27

.7

24.4

16

.1

23.6

16

.0

19.4

16

.1

9.9

16

.1

IVc+

VII

b 28

.8

10.0

5.

6

2.8

10

.9

35.1

9.

5

3.7

1.

1

9.2

V

+VI

11.6

14

.6

19.2

20

.9

16.6

11

.6

13.9

14

.1

8.0

12

.7

VII

a 15

.2

14.7

14

.2

14.4

14

.6

16.3

20

.3

15.8

11

.0

16.4

Cla

ss O

rigi

n

I+

II

9.7

13

.6

12.8

14

.8

12.7

8.

4

13.4

15

.3

17.3

14

.1

IIIa

b 4.

0

5.9

6.

2

5.7

5.

6

3.8

7.

0

5.3

5.

2

5.6

Iv

ab

15.0

16

.9

19.9

27

.4

19.1

12

.4

16.7

22

.1

28.5

20

.2

IVc+

VII

b 59

.9

46.6

37

.6

20.8

42

.2

65.7

46

.3

34.8

20

.2

40.1

V

+VI

4.6

7.

8

11.7

12

.9

9.4

4.

1

7.7

10

.2

12.4

9.

0

VII

a 6.

8

9.2

11

.8

18.4

11

.0

5.6

8.

9

12.4

16

.3

11.0

Tota

l Mob

ility

Rat

e 0.

59

0.72

0.

75

0.75

0.

71

0.59

0.

75

0.80

0.

82

0.76

In

dex

of D

issi

mila

rity

0.31

0.

37

0.32

0.

33

0.31

0.

31

0.37

0.

37

0.48

0.

35

－ 66 －

Table 2. Outflow Rates by Birth Cohorts

Outflow Rates to Class Destination: Men Women I+II IIIab IVab IVc+VIIb V+VI VIIa I+II IIIab IVab IVc+VIIb V+VI VIIaFrom Class Origin: I+II 1928-45 50.9 7.5 19.7 4.6 6.4 11.0 36.2 34.8 11.6 4.3 4.3 8.7 1946-55 57.1 9.8 19.5 1.0 5.8 6.8 51.8 22.7 10.9 1.6 3.6 9.3 1956-65 54.9 9.8 19.4 0.7 7.5 7.7 44.4 32.0 14.2 0.3 4.7 4.5 1966-86 65.6 8.2 10.8 0.5 8.7 6.2 49.4 38.7 6.5 0.0 2.4 3.0 IIIab 1928-45 33.8 19.7 26.8 4.2 5.6 9.9 19.4 32.3 12.9 6.5 16.1 12.9 1946-55 35.1 12.1 24.7 1.7 12.1 14.4 31.0 29.5 19.4 0.8 6.2 13.2 1956-65 41.0 13.2 19.3 0.5 14.6 11.3 28.2 45.0 9.9 0.0 9.9 6.9 1966-86 44.0 12.0 13.3 2.7 13.3 14.7 29.4 51.0 5.9 0.0 5.9 7.8 IVab 1928-45 26.6 6.0 37.5 4.5 11.2 14.2 23.5 20.6 19.6 11.8 15.7 8.8 1946-55 29.1 6.5 41.7 2.2 9.7 10.7 20.8 22.7 24.4 1.9 10.4 19.8 1956-65 31.7 9.6 31.4 1.2 14.3 11.7 26.9 29.9 19.1 1.5 11.4 11.3 1966-86 34.3 12.4 21.5 0.6 17.7 13.5 39.9 36.0 9.0 0.0 6.8 8.3 IVc+VIIb 1928-45 11.8 2.4 16.1 44.0 11.5 14.1 3.0 5.0 15.4 48.2 10.0 18.4 1946-55 14.7 5.4 26.6 19.3 16.7 17.4 7.9 12.1 20.0 17.9 19.1 23.0 1956-65 19.9 4.4 24.6 12.6 23.5 15.0 13.5 19.9 16.4 8.3 18.8 23.2 1966-86 20.7 6.2 17.8 10.5 30.2 14.5 24.4 22.3 17.3 4.1 12.7 19.3 V+VI 1928-45 20.7 9.8 19.5 3.7 23.2 23.2 20.6 14.7 20.6 8.8 20.6 14.7 1946-55 31.4 7.0 21.4 1.3 24.5 14.4 17.0 29.8 19.9 3.5 10.6 19.1 1956-65 28.0 6.9 21.8 1.2 26.1 15.9 22.4 28.7 16.5 1.6 15.7 15.0 1966-86 32.2 9.9 13.5 0.6 23.4 20.5 37.2 34.7 8.3 1.7 7.4 10.7 VIIa 1928-45 8.3 5.0 25.6 14.0 15.7 31.4 2.2 15.2 19.6 17.4 21.7 23.9 1946-55 25.7 5.2 26.8 3.3 19.0 20.1 14.1 15.3 19.0 3.1 17.8 30.7 1956-65 21.4 8.4 22.7 3.0 22.2 22.4 14.9 28.8 14.9 2.3 17.8 21.4 1966-86 35.0 8.2 12.8 0.8 25.5 17.7 30.2 34.6 8.2 0.6 11.3 15.1

－ 67 －

Table 3. Inflow Rates by Birth Cohort

Inflow Rates from Class Origin Men Women I+II IIIab IVab IVc+VIIb V+VI VIIa I+II IIIab IVab IVc+VIIb V+VI VIIaTo Class Desitination: I+II 1928-45 26.2 7.1 21.1 37.5 5.1 3.0 31.6 7.6 30.4 20.3 8.9 1.3 1946-55 29.5 7.9 18.6 25.8 9.3 8.9 37.0 11.6 18.5 19.4 6.9 6.6 1956-65 24.0 8.7 21.6 25.7 11.3 8.7 29.4 6.4 25.8 20.4 9.9 8.0 1966-75 26.6 6.8 25.7 11.8 11.4 17.6 23.7 4.3 31.7 13.7 12.9 13.7 IIIab 1928-45 15.7 16.9 19.3 31.3 9.6 7.2 25.5 10.6 22.3 28.7 5.3 7.4 1946-55 20.0 10.8 16.4 37.4 8.2 7.2 16.8 11.4 21.0 30.8 12.6 7.5 1956-65 16.8 10.9 25.8 22.3 10.9 13.3 17.9 8.7 24.2 25.4 10.7 13.1 1966-75 12.9 7.3 36.3 13.7 13.7 16.1 19.6 7.8 30.1 13.3 12.7 16.6 IVab 1928-45 9.1 5.1 26.9 46.2 4.3 8.3 6.1 3.1 15.3 63.4 5.3 6.9 1946-55 9.6 5.3 25.4 44.8 6.0 8.9 7.6 7.0 21.0 47.9 7.8 8.7 1956-65 10.1 4.9 25.6 37.9 10.5 11.0 13.4 3.2 26.1 35.3 10.4 11.4 1966-75 9.9 4.7 36.8 23.1 10.8 14.6 11.5 3.1 26.0 35.4 10.4 13.5 IVc+VIIb 1928-45 1.6 0.6 2.3 91.6 0.6 3.3 1.0 0.7 4.2 90.3 1.0 2.8 1946-55 1.4 1.0 3.8 89.8 1.0 3.1 2.3 0.6 3.4 87.9 2.9 2.9 1956-65 1.6 0.5 4.2 84.9 2.6 6.3 1.1 0.0 8.7 78.3 4.3 7.6 1966-75 2.7 5.4 5.4 78.4 2.7 5.4 0.0 0.0 0.0 72.7 18.2 9.1 V+VI 1928-45 5.3 1.9 14.6 59.7 9.2 9.2 3.2 5.3 16.8 56.8 7.4 10.5 1946-55 5.4 4.9 11.3 53.3 13.1 12.0 3.5 3.1 12.5 63.7 5.9 11.3 1956-65 5.0 4.7 14.8 46.0 15.9 13.6 5.1 3.7 17.9 46.3 11.4 15.6 1966-75 6.2 3.6 23.2 30.1 14.5 22.5 5.1 3.8 24.4 32.1 11.5 23.1 VIIa 1928-45 7.0 2.6 14.0 55.4 7.0 14.0 4.5 3.0 6.7 73.9 3.7 8.2 1946-55 6.3 5.8 12.4 55.2 7.7 12.6 6.1 4.5 16.3 52.4 7.2 13.4 1956-65 7.0 4.9 16.4 39.8 13.1 18.7 4.3 2.3 15.8 51.1 9.7 16.8 1966-75 6.3 5.8 25.8 21.1 18.4 22.6 4.7 3.7 21.5 35.5 12.1 22.4

－ 68 －

Table 4. Percentage Distributions of Educational Levels by Cohort 1928-45 1946-55 1956-65 1966-75 Total Men Low 72.36 48.22 32.08 14.99 42.26 Medium 13.6 24.52 35.66 37.32 28.30 High 14.04 27.25 32.26 47.69 29.44 N 1,780 2,928 3,438 1,321 9,467 Women Low 87.21 60.56 39.45 12.53 47.91 Medium 6.94 20.21 32.92 39.73 26.70 High 5.85 19.23 27.63 47.74 25.38

N 821 1,841 2,494 974 6,130 Note: "Low:" up to middle school; "medium:" completion of high school; "high:" some tertiary education.

－ 69 －

Table 5. Goodness of Fit Statistics for Taiwanese Mobility Models across Cohorts df G2 p △G2 rG2 BIC ID G2(s)Men (N=9467) Cond. Independence 100 1823.7 .000 908.2 .162 536.8 CSF (full association) 75 84.3 .216 .954 -602.3 .031 77.4 Unidiff (full association) 72 81.5 .209 2.9 .955 -577.7 .030 74.4 CSF+Original Core 90 215.2 .000 .882 -608.8 .050 121.7 CSF+Taiwan variant 90 103.0 .165 112.2 .944 -721.0 .035 93.3 Unidiff + Taiwan variant 87 99.6 .167 3.3 .945 -696.9 .033 90.2 unidiff parameters Φ 1928-1945 1.00 Φ 1946-1955 1.02 Φ 1956-1965 0.91 Φ 1966-1975 0.92 Women (N=6130) Cond. Independence 100 1227.3 .000 355.2 .165 546.0 CSF (full association) 75 91.8 .091 .925 -562.3 .043 81.6 Unidiff (full association) 72 84.5 .149 7.3 .931 -543.4 .041 77.0 CSF+Original Core 90 210.2 .000 .829 -574.7 .065 137.6 CSF+Taiwan variant 90 106.0 .120 104.2 .914 -678.9 .045 96.3 Unidiff + Taiwan variant 87 97.4 .210 8.6 .921 -661.3 .043 91.1 unidiff parameters Φ 1928-1945 1.00 Φ 1946-1955 0.93 Φ 1956-1965 0.80 Φ 1966-1975 0.73

Note: �G2 is the differences of G2 from the preceding model

－ 70 －

Table 6. The Taiwan Version of Class Hierarchy Father Son/daughter Upper Stratum Service (I+II) Service (I+II)

Routine nonmanual (III) Routine nonmanual (III)Self-employed (IVab)

Middle Stratum

Skilled manual (V+VI) Self-employed (IVab)

Unskilled manual (VIIa) Skilled manual (V+VI) Farming (IVc+VIIb)

Lower Stratum

Unskilled manual (VIIa)Farming (IVc+VIIb)

－ 71 －

－ 72 －

Table 7. Effect Parameters of the Core and Taiwan Models Men (N=9467) Women (N=6130)

Original Core

Taiwan version

Original Core

Taiwan version

DIG(I+II) 1.17 *** 0.59 *** 1.10 *** 0.84 ***

(.11) (.10) (.15) (.15) DIG(III) 0.92 *** 0.78 *** 1.08 *** 0.99 ***

(.14) (.14) (.13) (.13) DIG(IVab) 0.58 *** 0.49 *** 0.09 0.21 (.07) (.08) (.09) (.12) DIG(IVc+VIIb) 2.28 *** 2.10 *** 2.17 *** 1.87 ***

(.11) (.11) (.14) (.15) DIG(V+VI) 0.33 *** 0.61 *** -0.22 0.07 (.09) (.09) (.14) (.14) DIG(VIIa) 0.30 ** 0.03 0.08 0.06 (.10) (.09) (.12) (.11) Hierarchy (HI1) 0.01 -0.24 *** (.04) (.04) Hierarchy (HI2) -0.37 *** -0.57 *** (.07) (.09)

Positive Affinity (AF2A) 0.61 *** 0.50 *** 0.99 *** 0.77 ***

(.08) (.08) (.09) (.10)

Positive Affinity (AF2B) 0.14 *** 0.21 ***

(.04) (.05) National Variant of the Affinity Effect

Hierarchy (HI1) - Taiwan -0.22 *** -0.21 ***

(.04) (.05)

Hierarchy (HI2) - Taiwan -0.42 *** -0.57 ***

(.06) (.08)

Positive Affinity (AF2B_TW) - Taiwan 0.25 *** 0.34 ***

(.04)

(.06)

Note: Standard errors are in the parentheses. * p < .05 **p<.01 ***p<.001;

－ 73 －

Tabl

e 8.

Orig

in b

y Ed

ucat

ion

by C

ohor

t Tab

le

dfχ2

p G

2p

△G

2B

IC

ID

G2 (s

)M

en (N

=946

7)

C

ond.

Inde

pend

ence

4015

13.7

.000

1517

.5.0

00

1151

.3

.151

414.

4 C

SF (f

ull a

ssoc

iatio

n)

30

40.3

.098

40.4

.098

-2

34.3

.0

2032

.6

Uni

diff

(ful

l ass

ocia

tion)

2736

.6.1

0237

.1.0

933.

2 -2

10.1

.0

1929

.6

CSF

usi

ng e

ffec

t mat

rices

33

75.8

.102

75.2

.093

-2

26.9

.0

2943

.7

Uni

diff

usi

ng e

ffec

t mat

rices

30

71.9

.000

71.9

.000

3.3

-202

.7

.027

40.6

unid

iff p

aram

eter

s

Φ

192

8-19

45

1.00

Φ

194

6-19

55

1.07

Φ

195

6-19

65

1.02

Φ

196

6-19

75

0.87

W

omen

(N=6

130)

Con

d. In

depe

nden

ce

40

1448

.9.0

0014

12.8

.000

10

63.9

.1

7958

3.1

CSF

(ful

l ass

ocia

tion)

3066

.5.0

0065

.0.0

00

-196

.6

.032

43.9

U

nidi

ff (f

ull a

ssoc

iatio

n)

27

56.9

.001

54.4

.001

10.7

-1

81.1

.0

2737

.8

CSF

usi

ng e

ffec

t mat

rices

33

89.1

.000

86.8

.000

-2

01.0

.0

3954

.3

Uni

diff

usi

ng e

ffec

t mat

rices

30

78.1

.000

74.4

.000

12.5

-1

87.3

.0

3447

.6

un

idiff

par

amet

ers

Φ 1

928-

1945

1.

00

Φ

194

6-19

55

0.88

Φ

195

6-19

65

0.73

Φ

196

6-19

75

0.76

Not

e: �

G2 is

the

diff

eren

ces o

f G2 fr

om th

e pr

eced

ing

mod

el.

－ 74 －

Table 9. Estimates of Origin-Education Effect Parameters Men (N=9467) Women (N=6130) SO1 0.67 *** 0.83 *** (.08) (.10) SO2 -0.60 *** -0.58 ** (.15) (.18) MO1 -0.46 *** -0.57 *** (.06) (.08) MO2P 0.35 ** 0.51 *** (.12) (.15) MO2F 1.39 *** 1.81 *** (.12) (.15) MO2S 0.62 *** 0.68 *** (.14) (.17) MO2U 0.91 *** 1.25 *** (.13) (.16) Note: Standard errors are in the parentheses. * p < .05 **p<.01 ***p<.001;

Table 10. Education-Destination Association by Gender and Birth Cohort

df N G2 p △G2 BIC ID G2(s)

Men CSF (full ED association) 230 9470 252.1 .152 -1853.815 .047 235.6 CSF with ED effect matrix 233 9470 353.6 .000 -1779.737 .058 263.6 Unidiff (full ED association) 227 9470 242.9 .223 9.1 -1835.458 .046 231.0 Unidiff with ED Effect matrix 230 9470 345.0 .000 8.6 -1760.853 .057 259.1 unidiff parameters Φ 1928-1945 1.00 Φ 1946-1955 0.87 Φ 1956-1965 0.84 Φ 1966-1975 0.94 Women CSF (full ED association) 230 6139 227.7 .531 -1778.504 .054 229.1 CSF with ED effect matrix 233 6139 281.7 .016 -1750.622 .062 252.3 Unidiff (full ED association) 227 6139 221.6 .590 6.1 -1758.449 .052 224.8 Unidiff with ED Effect matrix 230 6139 273.4 .026 8.3 -1732.779 .059 247.2 unidiff parameters Φ 1928-1945 1.00 Φ 1946-1955 0.91 Φ 1956-1965 0.80 Φ 1966-1975 0.77

－ 75 －

Table 11. Origin-Destination Association Controlling for Education and Cohort df N G2 p △G2 BIC ID G2(s)

Men CSF (full OD association) 275 9470 286.0 .312 -2231.89 .052 277.8 CSF with OD effect matrix 290 9470 314.7 .153 -2340.57 .055 296.2 Unidiff (full OD association) 272 9470 283.8 .299 2.2 -2206.65 .052 275.0 Unidiff with OD effect matrix 287 9470 312.7 .142 1.9 -2315.03 .055 293.5 unidiff parameters Φ 1928-1945 1.00 Φ 1946-1955 0.96 Φ 1956-1965 0.87 Φ 1966-1975 0.99 Women CSF (full OD association) 275 6182 240.2 .936 -2160.45 .063 261.2 CSF with OD effect matrix 289 6139 259.2 .903 -2270.33 .066 277.2 Unidiff (full OD association) 272 6139 243.8 .890 -3.6 -2128.75 .063 260.8 Unidiff with OD Effect matrix 287 6139 257.1 .897 2.1 -2246.23 .066 275.2 unidiff parameters Φ 1928-1945 1.00 Φ 1946-1955 0.83 Φ 1956-1965 0.78 Φ 1966-1975 0.87

－ 76 －

Table 12. Estimates of Origin-destination Effect Parameters after Controlling for OE and ED associations

Men (N=9470) Women (N=6139) Original Core Original Core

DIG(I+II) 0.16 0.02 (.11) (.17) DIG(III) 0.50 *** 0.57 *** (.15) (.14) DIG(IVab) 0.56 *** 0.19 (.08) (.12) DIG(IVc+VIIb) 1.95 *** 1.56 *** (.11) (.15) DIG(V+VI) 0.51 *** 0.05 (.09) (.15) DIG(VIIa) 0.12 0.10 (.09) (.11) Hierarchy (HI1) - Taiwan -0.14 *** -0.13 * (.04) (.05) Hierarchy (HI2) - Taiwan -0.21 ** -0.25 ** (.06) (.09) Positive Affinity (AF2A) 0.10 0.15 (.09) (.11) Positive Affinity - Taiwan(AF2B_TW) 0.14 *** 0.11 (.04) (.07)

Note: Standard errors are in the parentheses. * p < .05 **p<.01 ***p<.001

－ 77 －

App

endi

x 1.

Eff

ect M

atrix

for t

he N

atio

nal V

aria

nt o

f the

Cor

e M

odel

H

iera

rchy

1 E

ffec

t (H

I1)

Hie

rarc

hy 2

Eff

ect (

HI2

)

I+

II

III

IVab

IV

c+V

IIb

V+V

IV

IIa

I+

IIII

IIV

abIV

c+V

IIb

V+V

IV

IIa

I+II

0

1 1

1 1

1 I+

II

0 0

0 1

1 1

III

1 0

0 1

1 1

III

0 0

0 0

0 0

IVab

1

0 0

1 1

1 IV

ab

0 0

0 0

0 0

IVc+

VII

b 1

1 1

0 0

0 IV

c+V

IIb

1 0

0 0

0 0

V+V

I 1

0 0

1 1

1 V

+VI

0 0

0 0

0 0

VII

a 1

1 1

0 0

0 V

IIa

1 0

0 0

0 0

Inhe

ritan

ce E

ffec

t (D

IG)

I+II

II

I IV

ab

IVc+

VII

bV

+VI

VII

a

I+II

1

0 0

0 0

0

III

0 2

0 0

0 0

IV

ab

0 0

3 0

0 0

IV

c+V

IIb

0 0

0 4

0 0

V

+VI

0 0

0 0

5 0

V

IIa

0 0

0 0

0 6

Posi

tive

Aff

inity

(AF2

A)-

-TW

I+II

II

I IV

ab

IVc+

VII

bV

+VI

VII

a

I+II

0

1 0

0 0

0

III

1 0

0 0

0 0

IV

ab

0 0

0 0

0 0

IV

c+V

IIb

0 0

0 0

0 0

V

+VI

0 0

0 0

0 0

V

IIa

0 0

0 0

0 0

Po

sitiv

e A

ffin

ity E

ffec

t (A

F2B

)--M

an

Posi

tive

Aff

inity

Eff

ect (

AF2

B)-

-Wom

en

I+II

II

I IV

ab

IVc+

VII

bV

+VI

VII

a

I+II

III

IVab

IVc+

VII

bV

+VI

VII

aI+

II

0 0

0 0

0 0

I+II

0

00

0 0

0 II

I 0

0 0

0 0

0 II

I 0

00

0 0

0 IV

ab

1 1

0 0

0 0

IVab

1

10

0 0

0 IV

c+V

IIb

0 0

1 0

1 0

IVc+

VII

b0

01

0 1

1 V

+VI

0 0

0 0

0 1

V+V

I 0

00

0 0

0 V

IIa

0 0

0 0

0 0

VII

a 0

00

0 0

0

－ 78 －

Trends in Intergenerational Social Mobility across

Four Cohorts in South Korea*

Hyunjoon Park (University of Pennsylvania)

and

Jongchun Cha (Sungkyunkwan University)

Abstract Considerable educational expansion and dramatic labor market changes in Korea during the last few decades make the Korean case interesting to examine changes in intergenerational social mobility. Applying log-linear and log-multiplicative models to intergenerational mobility tables constructed for four birth cohorts (born before 1944, 1945-54, 1955-64, and 1965 or later), we investigate trends across cohorts in the association between origin and destination classes. We also explore the role of educational attainment in mediating the origin-destination association. The analysis of three-way tables (origin by destination by cohort) indicates remarkable stability in social fluidity despite the extraordinary degree of educational expansion and occupational changes. The analysis of the association between origin class and educational attainment also suggests no significant change across cohorts in the effect of origin class on education. Moreover, the analysis of four-way tables (origin by destination by education by cohort) shows that both education-destination association conditional on origin, and origin-destination association conditional on education have not changed substantially across four cohorts.

1. Introduction Previous studies of intergenerational social mobility in Korea have documented some interesting features of the mobility pattern. Associated with the extraordinary speed of industrial transformation during the last few decades, Korea shows a considerably high degree of difference between distributions of class origin and destination. In particular, the extent to which the share of farmers among total labor forces has declined over a generation is indeed remarkable. Meanwhile, we have seen rapid growth in the share of blue- and white-collar workers. Korea is distinctive not only in its features associated with absolute mobility but also with relative mobility, which is the net association between class origin and destination after the distributions of class structure are taken into account. Some studies particularly based on data of early 1990s provide some evidence of a greater degree of social fluidity in Korea compared to western European countries and the United States (Cha 1991; Park 2004).

Although features of the Korean mobility regime in the early 1990s as compared to other western countries (represented in the core model of social fluidity by Erikson * This paper was presented at the project meeting, “Social Mobility in Late-Industrializing Nations,” held at the University of Tokyo, December 15-16, 2007.

－ 79 －

and Goldthorpe 1992) are well documented, little research has systematically examined time trends in social mobility, especially for long-term trends. The paucity of research for long-term trends in social mobility has been mainly due to the lack of data that provide key information for social mobility research over the long period. However, along with several data collections during the last 10-15 years, there are now more opportunities for addressing the interesting question of changes over time in social mobility.

The Korean case is particularly useful for investigating trends over time in social fluidity given that the society has experienced tremendous changes in educational and occupational distributions during the recent periods, which is in sharp contrast to many western countries where recent educational and occupational changes have occurred in much limited ranges. In 2002, for instance, the proportion of people who had attained at least upper secondary education was only 31 percent among 55 to 64-year-olds in Korea, while it was 95 percent among those aged 25-34. The difference between younger and older age groups in the statistic reveals the magnitude of educational expansion in Korea. Note that the corresponding differences between the two age groups were only 3 percent in the United States, 14 percent in the United Kingdom, and 14 percent in Sweden (OECD 2004). The same story is told when the focus is on educational expansion at the level of college education. Two competing hypotheses have directed many studies of social mobility, mostly focusing on developed western countries. A perspective highlights the pivotal role of the industrialization of an economy and the modernization of the society. Most commonly, industrialization or modernization hypothesis is viewed as suggesting that along with industrialization, individual’s educational attainment rather than her origin class should become more important for her own occupational attainment (Treiman 1970; Treiman and Yip 1989). In regard to trends in social mobility, therefore, the hypothesis leads us to expect increasing social fluidity as the society experiences industrialization, often indicated by economic growth. Comparing 21 countries that differed in their levels of economic development, Treiman and Yip (1989) found that educational attainment had a stronger effect on occupational attainment in more industrialized countries.

A contrasting perspective, however, emphasizes the constant social fluidity in modern societies despite substantial changes during the last decades in economy and other aspects of society. Finding a considerable similarity in relative mobility rates across a variety of western countries, numerous studies have extended the argument for temporal stability in social mobility that is not affected by industrialization or even other major macro factors (Featherman, Jones, and Hauser 1975; Erikson and Goldthorpe 1992). In none of the previous studies that examined temporal changes in social mobility, the Korean case has been systematically investigated yet. The considerable magnitude of social change in Korea including industrialization and educational expansion that occurred recently makes Korea an excellent case for testing competing hypotheses on trends over time in social fluidity. A major aim of the current study is to assess how social fluidity has changed over time in Korea. In particular, we focus on how the dramatic expansion of education has affected social fluidity.

2. Macro-Level Changes In the current paper, we investigate temporal changes in the association between class

－ 80 －

origin and destination among Korean men across four birth cohorts constructed from five national representative surveys conducted in 1988, 1990, 1995, 2000, and 2005. A significant change in Korea during the last few decades, relevant to intergenerational social mobility, is the changes in occupational structure. Korea has experienced remarkably rapid industrialization during the last few decades, transforming it society from predominantly agricultural to industrial, and moreover from industrial to post-industrial based upon information technology. Examining the change in occupation structure associated with the evolution of industrialization is particularly useful for understanding the change in marginal distributions of class origin and destination as we will present later. Figure 1 and 2 present the occupational distribution by year for male employed and female employed, respectively. Most evident from the two figures is the drastic decline in agricultural sector among both male and female employed. On the opposite side is the increase in the share of those having a professional or managerial job especially among men. Although the proportion of women with a managerial job is still negligible in 2000, the share of those with a clerical job has been increasing significantly.

Along with the industrialization, Korea achieved remarkable economic growth. In the initial period of industrialization between 1963 and 1971, the annual growth rate of GNP (Gross National Product) reached 9.5 percent. During the second period of industrialization from 1973 to 1979, the government adopted a new strategy geared to promoting heavy and chemical industries (HCI) like steel, automobiles, and shipbuilding. During this period, economic growth continued at an 8.5 percent annual rate. Since the middle of the 1980s the service and information sectors have become much more important (Seong 2001). The remarkable economic achievement is reflected in a striking increase in per capita GDP from $1437 in 1963 to $10,874 in 1995.

However, Korea has recently been experiencing economic and social transformation of unprecedented degree since the economic crisis which began in late 1997. As seen in Figure 3, the Korean economy experienced negative growth in 1998. However, the economic crisis has far-reaching influences on the Korean society as a whole beyond economic contraction. For instance, a large-scale unemployment was accompanied by the economic crisis. During the earlier of rapid industrialization between late 1960s and 1980s, Korea maintained relatively low level of unemployment. As can be seen in Figure 4, unemployment rate among men continued to decline maintaining less than 3 percent until 1997. Unemployment rate among women were fairly stable in the range of 2 to 3 percent from 1970 to 1997. In 1998, when Korean economy was turbulent by economic crisis that had started from the end of 1997, unemployment rate soared to 7.6 percent and 5.6 percent for men and women, respectively. Although this level of unemployment was relatively fair in international comparison (for instance, the unemployment rate in 1998 was 19 percent, 10.6 percent, 6.5 percent in Spain Greece and Sweden, respectively), it was the unprecedented level of unemployment ever in the history of Korean economy. Although, since the peak in 1998 the unemployment rate has been declining toward the level before the economic crisis, the growing instability of employment associated with relative easiness of massive lay-off and rising proportion of non-regular workers is a serious social concern. Along with the growing job insecurity, there has been deteriorating equality of income distribution since the economic crisis (Lee 2002). Before the economic crisis, Korea has managed to maintain a relatively low level of income inequality in international standard. The combination of rapid economic growth and low income inequality was a major feature among the economies of ‘East Asian Miracle’ including

－ 81 －

the Korean economy (World Bank 1993). But since the economic crisis, income inequality has apparently been rising along with the increase of poverty. As in many other countries, inequality in wealth distribution, specially associated with land ownership, has been more apparent than inequality in income distribution. The sharply rising inequality in wealth distribution since the economic crisis has also contributed to the widening economic gap between low and high strata (Lee 2002).

In the context of the economic and social changes associated with the economic crisis, especially the rising employment insecurity and growing economic inequality, we may expect a declining fluidity in the Korean mobility regime after the economic crisis. On the other hand, however, educational expansion that had started in the early 1980s has become more apparent, showing an extraordinary degree during the 1990s. Although the educational expansion occurred in most industrial societies more or less during the 20th century, the extent of Korea’s educational expansion is so dramatic that few other countries have achieved a comparable level (Park 2004). Within a generation, the overall level of education among Korean population has rapidly risen, now being the country with the highest proportion of young people with college degrees in the world. Figure 5 presents advancements from lower secondary (middle school) to upper secondary (high school) education and from upper secondary education to tertiary education (including both four-year universities and two-year junior colleges) by year separately for females and males. For the transition from lower secondary to upper secondary education, the advancement rate continuously increased to reach 95 percent in 1990 for both females and males. The advancement rate from upper secondary to tertiary education shows rapid increase since 1990, followed by modest increase during the 1980s for both females and males. The increase during the 1990s in the advancement rate to tertiary education is particularly apparent. Between 1990 and 2002, the proportion of students in upper secondary education who proceeded to tertiary education increased more than two times.

In contemporarily societies, educational qualification is probably the most important determinant of social position. Numerous studies have investigated the role that educational attainment plays in mediating the effect of class origin on destination in a variety of society (Breen 2004). Evidence shows that the effect of class origin on educational attainment has declined over time in some countries, which also contributed to the decrease in the overall association between class origin and destination. In such countries as the United States (Hout 1988) or France (Vallet 2004), educational expansion contributed to the weakening association between class origin and destination by increasing the proportion of population with college degrees among which the association between class origin and destination is substantially weaker compared to those with low education.

In short, economic change and educational expansion that occurred in the 1990s in Korea make this period particularly interesting to researchers of social mobility. We are interested in examining the extent to which the association between class origin and destination has changed over time. We assess how the relative openness of the Korean mobility regime, which earlier studies found in comparison to other countries, has evolved over time in the context of growing economic inequality and job insecurity. We examine how the educational expansion has contributed to temporal stability or change in intergenerational social mobility. In examining trends in social fluidity, we take a cohort perspective rather than a period perspective. Comparing mobility tables, which cover the entire population regardless of the age and career stage given a survey year, across different years is

－ 82 －

difficult to identify the impact of specific historical changes that might affect the specific groups of population (Müller and Pollak 2004). As seen above, for instance, educational expansion in Korea occurred in a large scale mostly since 1980s, which suggests that only very recent cohorts have been affected by educational expansion. Moreover, economic crisis in 1997 might have different impacts among different age groups. Therefore, by lumping different cohorts into a mobility table, the period perspective of social mobility has limitations in identifying sources of changes in social fluidity. The problem is more apparent in our case where the five different cross-sectional surveys cover only the limited span of 17 years from 1988 to 2005.

3. Data and Class Classification In this study, we use five cross-sectional surveys to construct mobility tables. Those are Regional Conflict Study (RCS, total sample 2,011)) conducted in 1988 and three sets of Social Inequality and Justice (SIJ) conducted in 1990 (total sample 1,974), 1995 (total sample 1,865), and 2000 (total sample 1,858), and Social Stratification and Mobility survey in 2005 (total sample 2,080). As national representative surveys, they contain information on father’s and respondent’s occupation, employment status (i.e., whether self-employed or not), and educational attainment. Father’s occupation indicates the job the father had when the respondent was adolescent, while the respondent’s occupation indicates the current one.

Pooling the five surveys, we construct four birth cohorts to examine trends in social fluidity: those born before 1944, 1945 – 54, 1955 – 64, and 1965 or later. In constructing the four cohorts, we include only those aged 25 and 64 for each survey considering that most people would have completed the process of acquiring educational qualifications by age 25. We further limit our analyses to men. Literature has highlighted the tenuous attachment to labor market among Korean women (Brinton 2001). Even in recent years, only a half of Korean women at working ages actually are in labor force. Given this selectivity of those women who are in labor force, we need to be careful about classifying women’s class position based upon their own job. We basically adopt the EGP class schema (Erikson and Goldthorpe 1992) to classify origin and destination classes. But, given the negligible number of agricultural labor workers (VIIb) in Korean farming, we combine it with the classes of farmers (IVc), leading to six categories of class origin and destination: service class (I+II), routine non-manual class (III), non-farm self-employed (IVa+b), the farming class (IVc+VIIb), skilled manual workers (V+VI), and unskilled manual workers (VIIa). We classify father’s and respondent’s educational attainment into three categories: less than upper secondary completion, upper secondary (high school) completion, and tertiary completion (including both two-year junior colleges and four-year universities).

4. Class Structure and Absolute Mobility Table 1 shows the distributions of origin and destination classes across the four birth cohorts. Note that the distributions of classes in Table 1 were on the basis of weighted data. Most evident is the significant difference in class structure between origins and

－ 83 －

destinations given a cohort. The class of farmers and farm workers (IVc, VIIb) occupied the majority of origin classes even in the most recent cohort: 36 percent of the cohort members had fathers who were farmers or farm workers (Class IVc+VIIb; hereafter farmers for simplicity). The share of the class of farmers was significantly reduced among destination classes. For example, the proportion of farmers declined from 35 percent among the oldest cohort to only 3 percent among the youngest cohort. Concurrent with the rapid decline of the class of farmers, there has been the rise of white-collar classes. Both service class (Class I+II) and the routine non-manual class (III) increased from 10 percent to about 25 percent from the oldest to the youngest cohort. Notable is the substantial size of the self-employed in the Korean labor market. Even of the youngest cohort, 17 percent were self-employed. Note that the proportion of the youngest cohort might be underestimated because the members of the youngest cohort (born 1965 or later) are relatively younger (the most recent survey used in this analysis was conducted in 2005). The significant change in class structure between origins and destinations is reflected in the comparably high level of the index of dissimilarity shown in the bottom of the table. The index of dissimilarity between origin and destination classes indicates the proportion of cases that would be relocated in order to make the two distributions of origin and destination identical (Hout 1983). For instance, 43 percent of the oldest cohort should be reclassified in order to have the identical distributions of origin and destination classes. Although the index slightly declined for the youngest cohort, it was still 40 percent. The value of 40 can be considered fairly high in comparison to the index of other countries. For instance, among 11 European countries included in the recent cross-national comparison of social mobility, the highest value of the index for the 1990s data was 30.5 in Israel (Breen and Luijkx 2004; Table 3.5). In short, the comparison between class structures of origins and destinations highlights extraordinarily rapid changes between father’s and son’s generations in Korea.

In the bottom of the table, reported are total mobility rates, which indicate the percentages of people whose own classes differ from their origin classes. Of the oldest cohort, 60 percent attained classes that differed from classes of their fathers. The rate increased to 75 percent for the second cohort of 1945 – 54 and then remained constant for the two younger cohorts.

Table 2 presents outflow mobility rates, which show the extent of moving out of an origin class to a specific destine class. Among the members of the oldest cohort whose father’s class was service class, 44 percent attained service class themselves. The outflow rate from service class as origin to service class as destination increased to 56 percent of the youngest cohort. Except for the outflow rate from the farming class, the outflow rate to service class increased across the four cohorts for all the origins, reflecting the increase of the size of service class. At the same time, outflows to the farming class decreased substantially across the four cohorts. Inflow rates to service class in Table 3 indicate that the extent to which sons of service originated from fathers of service class increased from 20 percent of the oldest cohort to 32 percent of the youngest cohort, while the extent to which service class members were recruited from farmers as origin class declined significantly.

Showing the distributions of educational attainment for each cohort, Table 4 reveals the magnitude of educational expansion during the last few decades in Korea, as already discussed above. Compared to 62 percent of the oldest cohort who did not graduate from high schools, only 4 percent of the youngest cohort did so. On the other

－ 84 －

hand, the proportion of cohort members who completed tertiary education increased five times between the oldest and the youngest cohorts.

5. Trends in Social Fluidity

To examine changes over time in social fluidity, which is represented by the association between origin and destination classes after differences in class distribution are taken into account, we apply to three-way tables (origin by destination by cohort) a set of log-linear and log-multiplicative models. The results are presented in Table 5. As a conditional independence model, Model 1 assumes no association between origin and destination classes conditional on cohort. Model 2 postulates temporal invariance in the association between origin and destination classes (i.e., no three-way interaction among origin, destination, and cohort) without imposing a specific structure on the two-way interaction between origins and destinations. Often called as ‘Constant Social Fluidity Model,’ this model fits the data much better than the conditional independence model. In fact, the model is acceptable by the conventional statistical criterion (p = 0.5). In order to assess the extent to which the strength of the origin-destination association has changed across cohorts, we compare Model 2 and 2-1. Model 2-1 allows the strength of the association between origins and destinations in the kth table to be estimated by the log-multiplicative product of the overall association parameter common to all tables and a kth table-specific deviation parameter, φk (Xie 1992; log-multiplicative effect model or uniform difference model). The greater the value of a table specific-deviation parameter, the greater the strength of association between origin and destination classes. 1 This model does not improve on the constant fluidity model (Model 2) as indicated by the less negative BIC statistic (-536 vs. -556). In other words, evidence indicates stable social fluidity across the four cohorts. The examination of the φk parameters derived from the uniform difference model also confirms the constant social fluidity. Although Model 2 is the most preferable among the models estimated so far, it is not parsimonious using 25 more degrees of freedom than the conditional independence model. Therefore, it is necessary to find a more parsimonious model that specifies the pattern of the origin-destination association with a limited number of parameters. Among various models, the ‘core mode’ of social fluidity developed by Erikson and Goldthorpe (1992) has been widely applied to mobility tables of many industrial countries to detect the fluidity pattern. The core model basically captures the fluidity pattern prevailing in a mobility table using eight matrices, which are designed to represent the difficulty of moving across classes in different levels of hierarchy, the propensity of immobility, the likelihood of moving between agricultural and non-agricultural sectors, and the tendency of moving to specific classes.

Because the original code model is based upon the seven-class schema, however, the eight effect matrices should be modified in order for the core model to be fitted to our mobility table that has six categories of origin and destination classes. Ishida et al. (1995) provided modifications of the original core model to be applied to a mobility table based on the six-class schema that combined class of farmers (IVc) and class of farm workers

1 For identification of parameter, the deviation parameter for the first table is fixed to be 1. We estimate the uniform difference model using LEM (Vermunt 1997).

－ 85 －

(VIIb) into one category. The modified core model for the 6 by 6 table has five effect matrices (10 parameters) to capture the fluidity pattern as shown in the Appendix A. The first matrix, IN, captures the propensity of class inheritance separately for each class using six parameters. The second matrix, HI 1, represents the tendency of hierarchical movement. In the core mode, three levels of hierarchy are distinguished: the service class (I+II) occupies the top level, while the farming class (IVc) as an origin and the class of unskilled manual workers consist of the lowest level of hierarchy. The remaining classes are at the intermediate level (note that the farming class as a destination is at the intermediate level). The matrix, HI 1, captures any movements across the level of hierarchy. The third matrix, HI 2 is another component that pertains to hierarchical movement. Compared to HI 1, however, HI 2 distinguishes long-range movements across two levels of hierarchy from others. The positive affinity effect, AF2A, represents movement between the service class (I+II) and the routine non-manual class (III). Finally, another affinity matrix, AF2B, identifies mobility between the service class (I+II) and the class of self-employed (IVa+b), mobility between the self-employed and the farming classes, mobility between the classes of skilled (V+VI) and unskilled (VIIa) manual workers, and non-reciprocal mobility from the origin of farming class to the destination of unskilled manual workers. Model 3 assumes both the pattern and the strength of social fluidity described by the five effect matrices of the core model to be constant across four cohorts. In other words, Model 3 posits the constant social fluidity similar to Model 2 but describes the pattern of social fluidity more parsimoniously than Model 2. The more negative BIC statistic (-654) of Model 3 indicates that this model is preferable to Model 2 (-556). Model 3 is even acceptable by the conventional statistical criterion (p = 0.16). The significant improvement of the fit of Model 3 upon Model 2 indicates that the pattern of social fluidity represented by the core model describes well the Korean mobility regime. Model 3-1 postulates the pattern of social fluidity represented by the core model to be constant across four cohorts but the overall strength of association between origin and destination classes to vary across cohorts by the deviation parameter φk. This model does not improve upon the constant core model as indicated by the less negative BIC statistic (-633 vs. -654). The comparison based upon the core model is consistent with the comparison between the constant social fluidity model and the uniform difference model based upon the full two-way interaction model (Model 2 vs. Model 2-1). The pattern of the deviation parameters across cohorts confirms constant social fluidity. In the bottom of the table, we present the estimates of parameters for each effect matrix of the core model applied to 6 by 6 tables. The estimates presented are those from Model 3 that assume the constant fluidity across cohorts. First, among the estimates of immobility parameters in the IN matrix, the propensity for immobility of service class is 2.7 times (e0.98 = 2.7) what it would have been in the absence of the immobility effect. Likewise for the farming class, class immobility is 5.6 times (e1.72 = 5.6) what it would have been in the absence of the immobility effect. The next level of the propensity of immobility is among the classes of routine non-manual workers and skilled manual workers, followed by the propensities among self-employed and unskilled labor workers. The effect HI 1 captures the barriers to mobility across any different levels of hierarchy, while the effect HI 2 pertains specifically to long-range mobility across two levels of hierarchy. As expected, the two coefficients associated with HI 1 and HI 2, are negative, indicating the relative difficulty to move across different levels of hierarchy. However, note that the effect of HI 1 is significant only at 0.1 level (t = -1.9). Finally,

－ 86 －

there is evidence of the positive affinity in movement between the service class (I+II) and the routine non-manual class (III). Another positive affinity effect, AF2B, which assumes affinity in mobility between the service class (I+II) and the class of self-employed (IVa+b), between the self-employed and the farming classes, between the classes of skilled (V+VI) and unskilled (VIIa) manual workers, and non-reciprocal mobility from the origin of farming class to the destination of unskilled manual workers, is not significant.

6. Education and Social Mobility Following the idea of the status attainment model, we can decompose the intergeneration mobility into the direct effect from origin to destination classes and the indirect effect of origin through its influence on the respondent’s education, which in turn affect his own class. In other words, the gross effect of origin on destination consists of the sum of 1) the partial effect of origin on destination, controlling for education and 2) the effect of origin on destination via education (this is the product of the path from origin to education and the path from education to destination). Therefore, the overall change or stability of the association between origin and destination classes reflects the complicated combination of the change or stability of each path in the triangle structure of origin, education, and destination (Vallet 2004). For instance, the decline in the gross association between origin class and destination class may be caused by the decline either in the direct effect of origin on destination controlling for education, in the effect of origin on education, or in the effect of education on destination. However, even if any of the three path coefficients has not changed over time, the gross effect of origin on destination can decline when educational expansion increases the number of people with high qualifications for whom the association between origin and destination is very weak (compositional effect; Hout 1988, Vallet 2004). 1) Trends in the Effect of Origin Class on Education

In order better to understand the trends in gross social mobility and the mediating role of education, we first address how the relationship between origin classes and respondents’ educational attainment has changed over time. Table 6 presents the results of log-linear and log-multiplicative models fitted to three-way tables by origin class by educational attainment by cohort. Model 1 assumes no association between origin classes and educational attainment. Model 2 postulates the association between origin classes and educational attainment as full two-way interactions, which is assumed to be constant across cohorts. Model 2-1 specifies the association between origin classes and educational attainment to vary across cohorts by the deviation parameter φk. Comparison between Model 2 and Model 2-1 on the basis of BIC statistics indicates the better fit of Model 2, suggesting no significant change in the association across four cohorts. Although the deviation parameters seem to have slightly declined across cohorts, the degree of decline is not large enough to be significant.

Next, we fit another set of models that detect the pattern of the association between origin classes and education using an effect matrix proposed by Ishida et al. (1995) (Appendix B). Model 3 assumes the association, as specified by the effect matrix, to be constant across four cohorts, while Model 3-1 specifies the association to vary

－ 87 －

across cohorts by the deviation parameter φk. Similar to the results for the full two-way interaction models, the comparison between Model 3 and Model 3-1 indicates that the model of constant fluidity is preferred to the model of varying association. In short, regardless of whether the pattern of the association is detected by the full two-way interaction model or the effect matrix, the conclusion is consistent, indicating that the effects of origin classes on educational attainment has remained constant across cohorts. 2) The Four-Way Table: Origin by Destination by Education by Cohort

Once we find the origin-education association to have been constant, we now move to examine how the effect of educational attainment on classes of destination has changed over time, conditional on origin classes. Because the relationship between education and destination to be examined is conditional on origin, we analyze four-way tables by origin by destination by education by cohort. The analyses of four-way tables also allow us to assess trends in the relationship between origin and destination classes conditional on education. In Table 7, Model 1 reproduces the observed three-way (origin- education-cohort) margin exactly and allows differences across cohorts in the marginal distribution of destination. Model 2 adds the association between origin and destination classes to Model 1, while Model 3 adds the association between education and destination. Model 4 includes both the origin-destination association and the education-destination association, showing the best fit among four models estimated so far (note the most negative BIC value).

In order to test trends in the partial association between origin and destination, Model 5 allows the association to vary across cohorts by the deviation parameter φk. The comparison of the BIC statistics between Model 5 and Model 4 indicates the better fit of Model 4, suggesting no significant change across cohorts in the partial association between origin and destination. Similarly, Model 6 specifies the partial association between education and destination to vary across cohorts the deviation parameter φk. The poorer fit of Model 6 than Model 4 indicates that similar to the pattern for the partial association between origin and destination, the partial association between education and destination is constant across cohorts. Along with the result in Table 6, which showed the constant effect of origin on education, the results in Table 7 reveal little change across cohorts in any of the three paths in the origin-education-destination triangle.

Finally, Model 7 tests whether the origin-destination association varies across educational levels by allowing the association to differ across cohorts, while Model 8 specifies the trends by the deviation parameter φk. As described above, the gross effect of origin on destination classes can still change over time even if none of the three paths in the origin-education-destination triangle changes over time. It is possible if the origin-destination association is much weaker for population with higher education for which significant educational expansion occurred during the last few decades. Comparisons to Model 4, however, indicate that this compositional effect caused by expansion of higher education is not significant enough to reject the hypothesis of constant fluidity.

Note that Model 4 through 6 did not specify the origin-destination association but allowed full two-way interactions. In order to assess the robustness of the conclusion, we conducted additional analyses that specified the association between origin and destination by the core model as in the Appendix A. The comparisons of Model 4-1 through Model 6-1 to their corresponding models with full two-way interactions show

－ 88 －

their better fits, suggesting that the core model describes the Korean mobility regime better than the less parsimonious full two-way models. Notable, however, is that similar to the result on the basis of full two-way interactions, the result indicates no significant trends across cohorts in the origin-destination association and in the education-destination association.

In the bottom of Table 7, presented are estimates of parameters for effect matrices of the core model as obtained from the best model, M4-1. Note that the estimates indicate the association between origin and destination classes after education was taken into account. Therefore, comparing those estimates of effect matrices to estimates from the model without controlling for education is useful to address the extent to which education mediates the association between origin and destination classes. In Table 5, we already presented the estimates of effect matrices of the core model without taking into account education, which can be compared to the estimates in Table 7. Figure 1 displays the two groups of coefficients to facilitate comparisons. Almost half of the inheritance effect for service class (I+II) and routine non-manual class (III) is accounted by education, while education explains little of the inheritance effect for self-employed and two manual working classes. Controlling for education also reduces the effect HI 2 that indicates the difficulty of long-range mobility between service class and unskilled manual workers (VIIa) and of upward mobility from farm origin to service class. The affinity matrix, AF2A, which represents movement between the service class and the routine non-manual class, is also substantially reduced with education taken into account.

7. Conclusion The results of our analysis indicate remarkable stability of the Korean mobility regime across four birth cohorts (before 1944, 1945-54, 1955-64, and 1965 or later), despite the extraordinary degree of educational expansion and occupational changes during the last decades. Considering the important role of educational attainment for social mobility, it is puzzling that the dramatic expansion of education especially since the 1980s has not led to significant changes in social fluidity. The constant fluidity is particularly interesting as compared to the recent studies showing a significant increase of social fluidity in some western countries where the degree of educational expansion among recent cohorts is much limited than the degree in Korea (Breen 2004). In order to better understand the factors that might contribute to the constant fluidity, we examined in detail the relationships among origin, education, and destination. The analysis of the association between origin class and educational attainment showed no significant change across cohorts in the effect of origin class on education. Moreover, the analysis of four-way tables (origin by destination by education by cohort) indicated that both the partial association between education and destination conditional on origin, and the partial association between origin-destination conditional on education changed litter across four cohorts. In other words, none of the three paths in the origin-education-destination triangle showed a significant change across cohorts. Studying changes in the origin-destination association in the United States, Hout (1988) found the compositional effect, which contributed to the overall decline in the gross association between origin and destination classes, despite no change in any of the three paths in the origin-education-destination triangle. The compositional effect was caused by educational expansion that increased the proportion of the college educated for

－ 89 －

whom origin classes had weaker effects than their counterparts with low education. Our analysis, however, did not find the similar compositional effect in Korea where educational expansion occurred even in a greater degree that in the United States. Although there is some evidence that the origin-destination association is relatively weaker among the college educated than low educated in Korea as well, the variation in the effect of origin on destination by educational levels did not lead to the decline in the overall origin-destination association. The remarkable stability in social mobility in Korea, despite a considerable degree of educational and occupational changes during the last few decades, poses an interesting challenge to social mobility research. How could the Korean mobility regime change so little when the society has gone tremendous transformation? The stability in social fluidity in Korea is particularly contrasted to the significant increase of social fluidity since the late 1970s in several European countries (Breen 2004). Our analysis with the relatively small sample size, especially of the youngest cohort, might not well detect changes in social fluidity across cohorts. However, the careful examination of the parameters from the models across cohorts seems to support the conclusion of the stability in social fluidity.

References Breen, Richard. 2004. “The Comparative Study of Social Mobility.” Pp.1-16 in Social Mobility in Europe , edited by Richard Breen. Oxford: Oxford University Press. Breen, Richard, and Ruud Luijkx. 2004. "Social Mobility in Europe between 1970 and 2000." Pp. 37-76 in Social Mobility in Europe, edited by Richard Breen. Oxford: Oxford University Press. Cha, Jongchun. 1991. “Association Models for Intergenerational Social Mobility among Korean Men.” in Korean Journal of Sociology 25: 155-176 (in Korean). Cho, Donggi and Hyuikyung Cho. 2002. Knowledge Information and Change in Occupational Structure. Research Report. Korea Information Society Development Institute. (in Korean). Erikson, Robert. and John H. Goldthorpe. 1992. The Constant Flux: A Study of Class Mobility in Industrial Societies. Oxford: Clarendon Press. Featherman, David L., Frank L. Jones, and Robert M. Hauser. 1975. “Assumptions of Social Mobility Research in the United States: The Case of Occupational Status.” Social Science Research 4: 329-360. Hout, Michael. 1988. “Expanding Universalism, Less Structured Mobility: The American Occupational Structure.” American Journal of Sociology 93: 1358-1400. Ishida, Hiroshi, Walter Műller, and John M. Ridge. 1995. “Class Origin, Class Destination, and Education: A Cross-National study of Ten Industrial Nations.” American Journal of Sociology 101: 145-194.

－ 90 －

Lee. Jung-Woo. 2002. “The Economic Crisis and Economic Inequality in Korea.” Korea Journal 42: 178-203. Müller, Walter and R. Pollak. 2004. “Social Mobility in West Germany: The Long Arms of History Discovered?” Pp.77-111 in Social Mobility in Europe, edited by Richard Breen. Oxford: Oxford University Press. Organization for Economic Cooperation and Development (OECD). 2004. Education in a Glance. Paris: OECD. Park, Hyunjoon. 2004. “Intergenerational Social Mobility among Korean Men in Comparative Perspective.” Research in Social Stratification and Mobility 20: 227-253. Seong, Somi. 2001. “Korea’s Industrial Development, Globalization, and Prospects for Regional and Global Integration.” Working paper 2001-03. Korea Development Institute. Treiman, Donald D. 1970. “Industrialization and Social Stratification.” Pp. 207-234 in Social Stratification: Research and Theory for the 1970s , edited by Edward O. Laumann. Indianapolis: Bobbs-Merrill. Treiman, Donald D., and Kam-Bol Yip. 1989. “Educational and Occupational Attainment in 21 Countries.” Pp.373-394 in Cross-National Research in Sociology, edited by Melvin L. Kohn. Newbury Park: Sage. Vallet, Louis-André. 2004. “Change in Intergenerational Class Mobility in France from the 1970s to the 1990s and its Explanation: An Analysis Following the CASMIN Approach.” Pp.115-147 in Social Mobility in Europe, edited by Richard Breen.Oxford: Oxford University Press. World Bank. 1993. “Growth, Equity, and Economic Change.” Pp. 27-77 in The East Asian Miracle: Economic Growth and Public Policy. Washington,DC: Oxford University Press. Xie, Yu. 1992. "The Log-Multiplicative Layer Effect Model for Comparing Mobility Tables.” American Sociological Review 57:380-95.

－ 91 －

Figure 1. Occupational distribution by year: men

0%

20%

40%

60%

80%

100%

1970 1975 1980 1985 1990 1995 2000

Agricultural

Production

Service

Sales

Clerical

Managerial

Professional

Figure 2: Occupational distribution by year: women

0%

20%

40%

60%

80%

100%

1970 1975 1980 1985 1990 1995 2000

Agricultural

Production

Service

Sales

Clerical

Managerial

Professional

Source: Cho and Cho (2002)

－ 92 －

Figure 3. Annual growth rate of Gross Domestic Product

-10

-5

0

5

10

15

1960 1965 1970 1975 1980 1985 1990 1995 2000 2005

%

Source: Seong (2001)

－ 93 －

Figure 4: Unemployment rates by year

0

1

2

3

4

5

6

7

8

1970 1972 1974 1976 1978 1980 1982 1984 1986 1988 1990 1992 1994 1996 1998 2000 2002

Men Women

Source: ILO LABORSTA

－ 94 －

Figure 5. Advancement rates by gender and year

0

10

20

30

40

50

60

70

80

90

100

1965 1970 1975 1980 1985 1990 1995 2000 2005

from upper secondary to tertiary: males f rom upper secondary to tertiary: females

f rom low er secondary to upper secondary: males f rom low er secondary to upper secondary: females

Source: KEDI Online Statistics

－ 95 －

Figure 6. Coefficients of Effect Matrices of the Core Model

-1.0

-0.5

0.0

0.5

1.0

1.5

2.0

Dig (I+

II)

Dig (II

I)

Dig (IV

ab)

Dig (IV

c+VIIb

)

Dig (V

+VI)

Dig (V

IIa)

HI1

HI2

AF2A

AF2B

gross effect (without education controlled)

net effect (with education controlled)

－ 96 －

Table 1. Distributions (%) of Origins and Destinations by Cohort (weighted)1944 or before 1945-54 1955-64 1965 or later

Father's Class Class I+II 4.1 7.3 8.6 13.6 Class III 5.6 8.0 8.9 11.1 Class IVa+b 7.2 13.6 18.0 22.6 Class IVc+VIIb 78.6 64.2 53.2 36.2 Class V+VI 2.0 2.4 5.7 8.4 Class VIIa 2.6 4.4 5.7 8.1

Respondent's Class Class I+II 9.6 15.4 14.7 23.8 Class III 9.9 10.4 18.6 25.2 Class IVa+b 19.7 28.7 24.1 16.5 Class IVc+VIIb 35.4 13.8 8.0 2.7 Class V+VI 8.6 16.8 20.6 17.5 Class VIIa 16.8 14.9 14.1 14.3

Dissimilarity between origins and destination 43.15 50.5 45.2 39.6Total mobility rate 59.3 75.0 77.2 76.5

Unweighted N 1231 1264 1371 620

－ 97 －

Table 2. Outflow Rates to DestinationDestination Class I+II Class III Class IVa+b Class IVc+VIIb Class V+VI Class VIIa

Origin Class I+II

1944 or before 44.4 11.1 22.2 11.1 4.4 6.71945-54 42.1 9.5 24.2 6.3 11.6 6.31955-64 41.1 26.6 16.9 0.8 4.8 9.7

1965 or later 56.2 23.6 9.0 0.0 6.7 4.5 Class III

1944 or before 20.0 21.7 15.0 13.3 18.3 11.71945-54 35.0 21.4 18.4 4.9 15.5 4.91955-64 20.5 33.9 22.0 2.4 12.6 8.7

1965 or later 31.5 39.7 8.2 1.4 11.0 8.2 Class IVa+b

1944 or before 14.5 7.9 38.2 7.9 11.8 19.71945-54 17.1 11.4 41.7 2.9 17.7 9.11955-64 17.6 16.4 32.8 1.1 20.2 11.8

1965 or later 25.7 23.6 23.6 0.0 12.8 14.2 Class IVc+VIIb

1944 or before 6.6 9.0 17.8 42.3 7.5 16.71945-54 10.2 9.2 27.8 19.3 16.2 17.31955-64 8.7 15.4 23.0 13.6 22.7 16.7

1965 or later 9.7 19.3 21.8 6.7 24.8 17.6 Class V+VI

1944 or before 10.5 5.3 15.8 15.8 31.6 21.11945-54 16.1 9.7 35.5 3.2 29.0 6.51955-64 18.1 15.7 20.5 1.2 37.3 7.2

1965 or later 22.2 33.3 9.3 1.9 20.4 13.0 Class VIIa

1944 or before 7.4 14.8 33.3 3.7 3.7 37.01945-54 5.2 6.9 24.1 3.4 27.6 32.81955-64 9.6 22.9 24.1 3.6 20.5 19.3

1965 or later 18.5 29.6 5.6 0.0 22.2 24.1

－ 98 －

－ 99 －

Table 3. Inflow Rates from OriginDestination Class I+II Class III Class IVa+b Class IVc+VIIb Class V+VI Class VIIa

Origin Class I+II

1944 or before 19.4 4.8 4.8 1.3 2.2 1.71945-54 20.2 6.7 6.2 3.4 5.1 3.11955-64 23.9 12.2 6.0 0.9 2.0 5.9

1965 or later 32.1 12.7 7.3 0.0 5.2 4.3 Class III

1944 or before 11.7 12.4 4.3 2.1 12.0 3.91945-54 18.2 16.4 5.1 2.8 7.4 2.61955-64 12.2 15.9 8.0 2.6 5.4 5.4

1965 or later 14.7 17.6 5.5 5.6 7.0 6.5 Class IVa+b

1944 or before 10.7 5.7 13.8 1.6 9.8 8.31945-54 15.2 14.9 19.8 2.8 14.3 8.41955-64 21.6 15.9 24.6 2.6 17.7 15.1

1965 or later 24.4 21.2 32.1 0.0 16.5 22.6 Cl

93.9 68.5 78.389.3 61.8 74.990.5 58.9 62.988.9 51.3 45.2

Cl0.8 6.5 2.20.6 4.1 1.00.9 10.4 2.95.6 9.6 7.5

Cl0.3 1.1 5.61.1 7.4 9.92.6 5.7 7.80.0 10.4 14.0

ass IVc+VIIb1944 or before 54.4 72.4 71.4

1945-54 42.4 56.7 62.11955-64 31.5 44.1 50.9

1965 or later 14.7 27.9 47.7ass V+VI

1944 or before 1.9 1.0 1.41945-54 2.5 2.2 3.01955-64 7.0 4.8 4.9

1965 or later 7.7 10.9 4.6ass VIIa

1944 or before 1.9 3.8 4.31945-54 1.5 3.0 3.81955-64 3.8 7.0 5.7

1965 or later 6.4 9.7 2.8

Table 4. Distributions (%) of Respondent's Education by Cohort1944 or before 1945-54 1955-64 1965 or later

Respondent's Education Less than High School 61.6 41.7 19.0 4.4 High School 27.3 38.8 48.4 41.5 Tertiary 11.1 19.5 32.6 54.0

T

able

5. R

esul

ts o

f Fi

tting

Mod

els

to T

hree

-Way

Tab

les

by O

rigi

n C

lass

, Des

tinat

ion

Cla

ss, a

nd C

ohor

t

L2

dfp

I.D.

BIC

Pan

el A

M

1: C

ondi

tiona

l Ind

epen

denc

e M

odel

790.

425

100

0.00

00.

145

-50.

5

M

2: C

onst

ant F

luid

ity M

odel

(fu

ll tw

o-w

ay in

tera

ctio

ns)

74.3

1475

0.50

10.

035

-556

.3

M2-

1: U

nifo

rm D

iffe

renc

e M

odel

with

Ful

l Tw

o-W

ay In

tera

ctio

ns69

.612

720.

558

0.03

3-5

35.8

P

aram

eter

s(1

944

or b

efor

e)(1

945-

54)

(195

5-64

)(1

965

or la

ter)

1.00

00.

942

1.05

31.

275

M

3: C

onst

ant F

luid

ity M

odel

with

Mat

rice

s10

3.32

090

0.15

90.

038

-653

.5

M3-

1: U

nifo

rm D

iffe

renc

e M

odel

with

Mat

rice

s98

.916

870.

209

0.03

7-6

32.6

P

aram

eter

s(1

944

or b

efor

e)(1

945-

54)

(195

5-64

)(1

965

or la

ter)

1.00

00.

954

1.05

91.

277

Pan

el B

Eff

ect M

atri

x Pa

ram

eter

(fr

om M

odel

3 in

Pan

el A

)E

stim

ates

SE

Dig

(I+

II)

0.98

3(0

.225

)*

Dig

(II

I)0.

974

(0.1

42)*

D

ig (

IVab

)0.

442

(0.1

24)*

D

ig (

IVc+

VII

b)1.

719

(0.1

38)*

D

ig (

V+V

I)0.

519

(0.1

97)*

D

ig (

VII

a)0.

394

(0.1

92)*

H

I1-0

.124

(0.0

65)

H

I2-0

.695

(0.1

43)*

A

F2A

0.59

5(0

.121

)*

AF2

B0.

034

(0.0

81)

* st

atis

tical

ly s

igni

fica

nt a

t lea

st 0

.05

leve

l.

－ 10

0 －

Tabl

e 6.

Res

ults

of F

ittin

g M

odel

s to

Thr

ee-W

ay T

ables

by O

rigin

Cla

ss, R

espo

nden

t's E

duca

tion,

and

Coh

ort

L2df

pI.D

.BI

CM

1: C

ondi

tiona

l Ind

epen

denc

e M

odel

(OT

ET)

638.

206

400.

000

0.13

430

1.9

M2:

Con

stan

t Ass

ocia

tion

(full t

wo-w

ay in

tera

ctio

ns: O

T ET

OE)

36.9

2730

0.17

90.

024

-215

.3M

2-1:

Uni

form

Diff

eren

ce M

odel

with

Ful

l Two

-Way

Inte

ract

ions

35.6

3627

0.12

40.

030

-191

.4

Par

amet

ers

(194

4 or

bef

ore)

(194

5-54

)(1

955-

64)

(196

5 or

late

r)1.

000

0.89

30.

895

0.87

4

M3:

Con

stan

t Ass

ocia

tion

with

Mat

rices

54.4

2933

0.01

10.

031

-223

.1M

3-1:

Uni

form

Diff

eren

ce M

odel

with

Mat

rices

52.0

7230

0.00

80.

030

-200

.2

Par

amet

ers

(194

4 or

bef

ore)

(194

5-54

)(1

955-

64)

(196

5 or

late

r)1.

000

0.87

10.

867

0.80

9

－ 10

1 －

Table 7. T

Pan M1: M2: M3: M4: M M6: M7: M8:

M4- M5- M6-

PanEffe

－ 102 －

he Results of Analyses of Four-Way Tables (Origin by Destination by Education by Cohort)

L2 df p I.D. BICel A

OET DT 2172.594 340 0.000 0.264 -689.6 OET DT OD 1464.454 315 0.000 0.205 -1187.3 OET DT ED 643.863 330 0.000 0.117 -2134.2 OET DT OD ED 294.656 305 0.654 0.069 -2272.9

5: OET DT ODTuni ED 291.135 302 0.663 0.068 -2251.2OET DT OD EDTuni 293.711 302 0.623 0.080 -2248.6 OET DT ODE 239.267 255 0.752 0.064 -1907.4OET DT ODEuni ED 284.227 303 0.774 0.069 -2266.5

1: OET DT Matrices ED 322.309 320 0.453 0.073 -2371.51: OET DT MatricesTuni ED 317.976 317 0.474 0.072 -2350.61: OET DT Matrices EDTuni 321.361 317 0.421 0.072 -2347.2

el Bct Matrix Parameter (from M4-1) Estimates SE

Dig ( Dig ( Dig ( Dig ( Dig ( Dig ( HI1 HI2 AF AF

I+II) 0.562 (0.241)*III) 0.521 (0.150)*IVab) 0.556 (0.125)*IVc+VIIb) 1.392 (0.138)*V+VI) 0.551 (0.200)*VIIa) 0.391 (0.194)*

-0.044 (0.066)-0.359 (0.145)*

2A 0.075 (0.132)2B -0.020 (0.081)

App

endi

x A

. Cor

e So

cial

Flu

idity

Mod

el fo

r 6 b

y 6

Tabl

esIn

herit

ance

Eff

ect M

atrix

(IN

)I+

IIIII

IVab

IVc+

VIIb

V+V

IV

IIaI+

II2

11

11

1III

13

11

11

IVab

11

41

11

IVc+

VIIb

11

15

11

V+V

I1

11

16

1V

IIa1

11

11

7

Hie

rarc

hy E

ffec

t Mat

rix (H

I1)

Hie

rarc

hy E

ffec

t Mat

rix (H

I2)

I+II

IIIIV

abIV

c+V

IIbV

+VI

VIIa

I+II

IIIIV

abIV

c+V

IIbV

+VI

I+II

12

22

22

I+II

11

11

1III

21

11

12

III1

11

11

IVab

21

11

12

IVab

11

11

1IV

c+V

IIb2

22

22

1IV

c+V

IIb2

11

11

V+V

I2

11

11

2V

+VI

11

11

1V

IIa2

22

22

1V

IIa2

11

11

Posi

tive

Aff

inity

A E

ffec

t Mat

rix (A

F2A

)Po

sitiv

e A

ffin

ity B

Eff

ect M

atrix

(AF2

B)

I+II

IIIIV

abIV

c+V

IIbV

+VI

VIIa

I+II

IIIIV

abIV

c+V

IIbV

+VI

I+II

12

11

11

I+II

11

21

1III

21

11

11

III1

11

11

IVab

11

11

11

IVab

21

12

1IV

c+V

IIb1

11

11

1IV

c+V

IIb1

12

11

V+V

I1

11

11

1V

+VI

11

11

1V

IIa1

11

11

1V

IIa1

11

12

Sour

ce: I

shid

a et

al.

(199

5)

－ 10

3 －

Appendix B. Matrix for Association between Origin and Education

Class Origin High Medium LowI+II 2 1 3III 1 1 1IVab 1 1 5IVc+VIIb 4 1 6V+VI 4 1 7VIIa 4 1 8Source: Ishida et al. (1995)

Education

－ 104 －

Trends in Class Mobility and Education in China*

Xiaogang Wu

(Hong Kong University of Science & Technology)

Abstract This paper studies the trends in intergenerational class mobility pattern in the People’s

Republic of China in a comparative perspective, based on an analysis of five cohorts from two national representative surveys conducted in 1996 and 2005 respectively. In the context of rapid economic industrialization and dramatic social changes since 1949, the absolute mobility rates in China have increased over time, particularly for younger cohorts. The relative mobility patterns were largely stable for both men and women, despite the initial implementation and subsequent dismantling of socialist egalitarian programs. While education played an important role in intergenerational social reproduction, the effect of origin on educational attainment does not change over time; neither does the effect of education on destination. Despite the constant social fluidity, the rural-urban divide also led to certain distinctive features of Chinese mobility patterns that deviate from the core mobility models.

Key words: intergenerational mobility, education, China, socialism, industrialization.

1. Introduction One of the main research interests for scholars in social stratification and mobility is intergenerational class mobility, which focuses on the association between individual’s origins and destinations. As industrialization has gradually altered the mechanism of social stratification from ascription to achievement, Lipset and Zetterberg (1959) predicted that countries with similar degrees of development would have similar observed mobility rates. This constant absolute mobility thesis, however, subsequently received little empirical support. Featherman, Jones and Hauser (1975) suggested that, while observed mobility rates might differ from country to country because of macro- level changes in occupational structures, their relative mobility chances are essentially invariant once the distributions of origin and destination are controlled. The “FJH” or the constant relative mobility thesis was subsequently expanded to include a wider spectrum of industrial societies including both capitalist and socialist regimes and was refined as the “constant flux” thesis (Grusky and Hauser 1984; Erikson and Goldthorpe 1992).

While such invariance thesis has dominated comparative studies of social mobility for decades, a number of studies in recent years have demonstrated that there are significant cross-national variations in both absolute and relative mobility patterns, which * This paper was presented at the International Conference on “East Asian Comparative Research,” November 24-25, 2007 in Taipei, Taiwan, and various meetings of the Project on Social Mobility in Late-industrializing Nations at the University of Tokyo, Japan. Direct all correspondence to Xiaogang Wu, Social Science Division, Hong Kong University of Science & Technology, Clear Water Bay, Kowloon, Hong Kong SAR (Email: sowu@ust.hk).

－ 105 －

cannot be accounted for by simple typologies (such as capitalism vs. socialism, developed vs. developing) but are rather related to country-specific historical developments (e.g., Ganzeboom, Luijkx, and Treiman 1989; Wong 1990, 1992). Under the new log-multiplicative framework, Xie (1992:391) has identified the cross-national variations in circulation mobility and suggested “what might be constant is the ‘pattern of circulation mobility’” (Xie 1992). Ishida, Goldthorpe, and Erickson (1991) argued that the cross- national difference could be better understood as variations on the common or “core” pattern of social fluidity, rather than constituting a distinctive pattern of mobility. In this sense, the core model of social fluidity originally developed in western developed countries seems to fit well in Eastern Europe under a different political system and in Japan of a different cultural tradition.

In a comparative typology, China occupies a unique position. As a major socialist country, China provides an indispensable case to examine how the state intervention shaped social mobility patterns; as the most populous developing country, China also serves as an intriguing example to investigate the impact of rapid economic industrialization on social mobility. Moreover, the industrialization in China under different historical periods may have different implications on social mobility. Under the socialist industrialization program, the Chinese government has installed a unique institution to control the inflow of peasants into cities and created a structural barrier for moving out of agriculture, thus the industrialization focusing on heavy industry may not have created occupational opportunities as many as those countries under different development paths. Indeed, the rural-urban institutional divide, according to Wu and Treiman (2007), has contributed to high immobility (in agriculture) on the one hand, and unusual “openness” in the urban population among which individuals from rural origins are highly selective based on the criteria imposed by the government (Wu and Treiman 2004). The economic reform since 1978 has led to a spontaneous industrialization and geographic migration which gradually converged to the trajectory experienced by other developing countries.

This chapter studies the trend in intergenerational class mobility (social mobility) pattern in the People’s Republic of China in a comparative perspective, based on an analysis of five cohorts from two national representative surveys conducted in 1996 and 2005. In the following, I first describe the context of economic development and social changes in China, then examine the trends in the absolute mobility rates and in relative mobility over time. I adopt the class schema and models commonly used in comparative mobility research and demonstrate to what extent China deviates from the commonality. In this respect, the Chinese distinctiveness is to be identified empirically rather than to be presumed conceptually. Finally, I investigate the effect of origin on education and the effect of education on destination in intergenerational social reproduction.

2. Economic Growth and Social Change in China Since 1949 In the second half of the 20th century, China has been undergoing dramatic economic and social changes. Shortly after the Communist Party took the power in 1949, the Chinese government initiated ambitious industrialization programs in the early 1950s. The socialist-style economic development model China adopted from the Soviet Union put focus on capital intensive heavy industry, which could not absorb the oversupply labor. China’s socialist industrialization program was made possible by the hukou system and

－ 106 －

restricted migration, which allowed the government to exploit the agricultural sector and sacrifice the interests of rural residents for those of urban residents. To ensure food grain needed for urban industrial growth, the government relied on a system of “unified purchase”(tonggou) to forcibly procure farm produce at low prices. At the same time, consumer products allocated free-of-charge or at low prices to urban residents as welfare benefits of their work units were sold at high prices in rural areas. The government’s discriminatory policy resulted in a substantial gap in income and living standards between rural and urban residents. Permanent urban residents also enjoyed many other welfare benefits delivered by the state, such as free or subsidized food grain, free or low-rent apartments, and retirement and medical insurance. The government also guaranteed every eligible urban resident a permanent job, but accepted no such responsibility for rural residents. Since the available resources are fairly limited given the level of economic development in China at that time, with the hukou system the government was able to confine the majority of people in the bottom so as to make redistribution favorable to urban residents. As a result, the hukou differentiated 70 percent of population in rural areas from 30 percent of population in urban areas, creating a fundamental structure of inequality in state socialist China. Changing from rural hukou to urban hukou was very difficult (Wu and Treiman 2004), most through receiving higher education. This process bears important implications for intergenerational social mobility in China (Wu and Treiman 2007).

As Table 1 shows, while the primary sector’s share of total GDP decreased from 50.5 percent in 1952 to 32.4 percent in 1975, shortly after which the economic reform started, the workers employed in the primary sector decreased from 83.5 percent only to 77.2 during the same period. Occupational structural mobility lagged far behind the transformation of economic structures. Industrialization under socialism may not have created many mobility opportunities as expected to occur in other developed and developing countries.

Also since the early 1950s, private ownership was eliminated in China and the economy was organized in a new system. The distribution of income was controlled by the state. The overall income inequality, measure by the Gini index, had been declining from 0.561 in 1953 to 0.286 in 1975, as shown in Table 2. Intergenerational transmission of class status via economic capital seemed to be weakened. Empirical analyses of data collected from China have lent some support to this claim. For example, Parish (1981, 1984) reported that the effect of parental status on children’s educational and occupational attainment, although positive in China as in other societies, declined sharply after 1966 as a result of the Cultural Revolution. Blau and Ruan (1990) found that transmission of occupational status was much less pronounced than in the urban United States. In particular, a father’s occupational status did not improve his son’s achievement. Similar results have been reported by Lin and Bian (1991) for another Tianjin sample and by Lin and Xie (1988) for a Beijing sample. Hence, many scholars concluded that Maoist destratification policies promoting social equality were successful (e.g., Parish 1984).

The market-oriented reform since 1978 has been not only driving economic growth and industrialization more rapidly than before, but also changing the way of economic growth with different implications on mobility and inequality. Most literature on Chinese social stratification and mobility are focused on examining the impact of market reforms on the change of income inequality, which have been increasing dramatically since the mid-1990s. For example, Gini index has increased from 0.3 in the mid -1980s to 0.45 in 2003. The increasing inequality caused by economic reforms has

－ 107 －

profound implications for intergenerational and intra-generational class mobility that need to be explored.

First, the abolition of the collective farming system, starting in 1978, freed peasants from seeking jobs in the industrial and service sectors either at local communities or in cities. Rural industries boomed in the 1980s and created a spontaneous industrialization in China comparable to other developing countries. Migration into cities has been surging, particularly since the 1990s (Liang and Ma 2004). Despite the fact that the hukou system is still in place and rural migrants were discriminated in cities without permanent residential status, they were offered more opportunities in industrial and service sectors and likely to achieve socioeconomic mobility. These opportunities were unavailable in the pre-reform period when the migration and employment were strictly controlled by the government rather than the market.

Second, after a long-time suppression, self-employment and private economic activities had been legitimized as integral parts of socialism since 1982, and new social propertied classes are emerging from different social background (Wu 2006). As a result, the economic development and institutional transition in China have brought about significant consequences in social mobility.

A recent report by the Chinese Academy of Social Sciences showed that more opportunities have become available and mobility rates have increased. The intergenerational mobility rates increased from 41.4 percent (32.4 percent upward) prior to 1980 to 54.4 percent (40.9 percent upward) after 1980. The overall job mobility rate was only 12.3 percent prior to 1979, but increased to 30.3 percent between 1980 and 1989, and further to 54.2 percent between 1990 and 2001. The upward mobility from the first job to the current job was only 7.4 percent prior to 1979, but increased to 18.2 percent in the 1980s, and further to 30.5 percent in the 1990s (Lu 2004, p 13).

On the other hand, in terms of relative mobility, the increase in income inequality and the emergence of propertied class may suggest that the market transition may increase the intergenerational transmission of class status previously reduced by the state intervention, as Gerber and Hout (2004) found in the post-Soviet Russia. In light of the economic and social changes in China since 1949, the following analysis aims to examine the trend of intergenerational social mobility, and the role of education in mediating between class origin and class destinations.

3. The Data and Class Schema Because multi-wave national representative survey data for China are rarely available, we rely on the cohort analysis of social mobility pattern based on the pooled data from the Chinese Life and Social Changes survey (1996) and the Chinese General Social Survey (2005), two national probability samples of adults aged 20-69 from all regions in China (except Tibet in 1996 and Tibet and Qinghai province in 2005). The surveys are effectively a national probability sample of the Chinese population, since the population of Tibet and neighboring Qinghai is so small that it is extremely unlikely an individual was selected.

The surveys covered both rural and urban population, and the questionnaires included a broad range of topics and solicited information about both the respondents and their families. Information on respondents’ occupations, education, household registration status (hukou), and similar information about the respondent’s father when the respondent

－ 108 －

was at age 14, is used in the analyses. Rural and urban sample are combined, with appropriate weights, to represent the general population of China.

One should be aware that cohort analysis of occupational class mobility to address temporal trend is only the second choice, since cohort effect is likely complicated with life course effect. Because most Chinese by age 30 have established a stable career, we restrict the sample aged between 30-64 years old and the cohort variations could be approximately seen as reflecting the historical trends.

I divide all men and women into five cohorts: 1932-1940, 1941-1949, 1950-1958, 1959-1967, and 1968-1976. Missing cases are listwise deleted. To ensure enough sample cases (for the 1996 survey), I impute the respondent’s last occupation if his/her current occupation is missing (retired or retreated from the labor force). If father’s occupation when the respondent was at age 14 was missing, we impute it with father’s occupation in 1948, 1963, 1970, 1979, or 1989, whichever is closer to the year when the respondent was 14 years old. Given the high rate of labor force participation, only 92 women and 12 men have no occupation (accounting for 3.58 percent and 0.45 percent of sampled women and men respectively). In the end, 10,842 individuals are obtained, including 3860 in 1996 survey and 6982 in the 2005 survey, for analyses.

The most important variable in this analysis is occupation, for both the respondent and his/her father. To facilitate the comparative analysis of sectoral barriers to mobility (Featherman and Hauser 1978; Goldthorpe 1987), we recode occupations into the EGP scheme (Erikson, Goldthorpe1992; Ganzeboom et al. 1992; Ganzeboom and Treiman 1996). The EGP class scheme has several versions. Originally developed by Erikson et al. (1979), the full text version includes 11 classes, and many comparative studies have used the 7-class version as follows:

I+II Service class (professional and managers) III Routine nonmanual workers Iva+IVb Petty Bourgeoisie IVc Self-employed farmers and landowners; V+VI Skilled workers VIIa Non-skilled workers VIIb Agricultural workers

Since self-employed farmers/landowners and agricultural workers are hardly distinguishable in China as well as in other East Asian countries, I use the collapsed version of 6 classes, which combines IVc and VIIb (e.g. Ishida, Muller, and Ridge 1995). We code both the respondent’s current occupation and his father’s occupation when the respondent was age 14 into 6 classes. The following is mainly based on the analysis of the 6x6 square tables among 5 birth cohorts for men and for women.

While the specific ranking order of 6 classes is not presumed, in many countries IVc (self-employed farmers and landowners) is placed between petty bourgeoisie (IVa+IVb) and skilled workers (V+VI). In the 6-class version that combines (IVc+VIIb), I let farmers stay in the same place in fitting the core models for comparative analysis, and demonstrate how deviates from the models derived from most industrial countries.

It is well-known that education is both the primary determinant of occupational mobility and the main vehicle of intergenerational status transmission. Respondent’s education is recoded into three levels, namely high level (tertiary education), middle level

－ 109 －

(senior high school), and low level (junior high school or below). I investigate how the effect of family class background on education and the effect of education on occupational class have changed across five cohorts.

4. The Absolute Mobility Rates What proportion of respondents kept the same class as their fathers and what proportion experience occupational mobility? The total mobility rates are presented in the bottom of Table 3. Compared to other countries which have experienced industrialization, the absolute mobility rates are lower in China than in most western industrial societies as well as other developing societies, only between 40 percent and 50 percent for men and 30 percent to 40 percent for women. Due to its socialist industrialization program that discouraged the mobility out of agricultural sector on the basis of the household registration system, the rate of inheritance among farmers are particularly high, and farmer is the largest class category for both fathers’ and children’s generations. This explains the lower total mobility rate of the Chinese working population.

There are variations in absolute mobility rate among five birth cohorts, presenting a U-shape curve. The total mobility rate is low for the two cohorts in the middle and high for two youngest cohorts. This pattern can be linked to China’s economic development models. The oldest cohort enjoyed relative high mobility opportunity because the socialist state had not fully established the institutions to control labor mobility and allocated life chances in the course of economic development. Indeed, in the early 1950s, rural farmers could still move freely into cities and looked for industrial jobs without restrictions and many newly established state industrial enterprises indeed had recruited millions of rural laborers during the first five-year plan (1952-1957) (Meisner 1999).

A dramatic increase in (nominal) industrial growth and urban inflow pushed China’s urban population to about 20 percent in 1960, all time high in the pre-reform era (National Bureau of Statistics 1999). The government soon realized that China’s grain production capacity was unable to sustain such a huge urban population, especially given the decline in agricultural production during the Great Leap Forward (1958-1959). Thus, beginning in 1959 the government expanded and rigorously enforced its use of the hukou system as a tool to control migration. About 18 million urban workers were sent back to their home villages between 1961 and 1963 (Chan 1994, p39), and more than 20 million university and middle school students from urban areas were sent down to rural and border regions during the Cultural Revolution (1966-1976), to help reduce both urban unemployment and school crowding (Bernstein 1977; Zhou and Hou 1999). The mobility of the two middle cohorts were indeed subject to the full-fledged socialist planning system that rigid control the mobility opportunities.

Economic reforms since 1978 had weakened the system and brought about structural changes in mobility opportunities. Farmers were allowed to either work in local non-agricultural industrial sector (i.e., Township or Village Owned Enterprises, Walder 2002) or to migrate into cities for more economic opportunities. Social mobility rates for men and women of the two youngest cohorts increased as the market-oriented economic reform proceeded.

Table 3 also presents the distribution of class origins and class destinations for different cohorts. The variations in the distributions of class origins and destinations across different cohorts reflect some profound institutional transitions undertaking in the

－ 110 －

history of the People’s Republic of China. First, the share of service, routine non-manual, and farm classes had been stable

for the first three cohorts, and the share of farm class indeed increased slightly. Despite the fact that the GDP share of industrial and service sector increased from 50 percent in 1952 to 70 percent in 1980 (Table 1), the class structure did not change very much until the economic reform started. The most pronounced change is decline in the share of petty bourgeoisie in class origins and the increasing share in class destinations, particularly for the youngest cohorts. This pattern has resulted from the Chinese government policy shift towards the private ownership over the past half century. For those born in 1968-1976, 8 percent are self-employed or small business owners, whereas the share of farmers in this cohort is 33.4 percent, much lower than older cohorts. Such structural change associated with the market-driven economic growth and industrialization could have yielded profound impact on class structure and class mobility in China.

The patterns across cohorts described above hold true for both men and women, although a gender gap in total mobility rate (shown in the bottom of the tables) can be observed. For all five cohorts, men enjoyed high mobility rates than did women. A close examination of the tables shows that the inheritance rate of farm class is much higher among women than among men. That is to say, a farmer’s son has more chance to move out of agriculture than a farmer’s daughter. As rural China is still a patriarchic society influenced by traditional culture, parents generally favor sons over daughters in allocating resources for them to get access to education and job opportunities.

How do class destinations differ from their origins? Table 4 presents the outflow rates by cohorts and gender to affirm the patterns observed in Table 3 with more details. For those from farm origins, the rate of outflow into a non-farm class was indeed highest for the oldest cohorts born before 1949, which is clearly at odds with the continuing growth of China’s industrial sector since the 1950s. Specifically, the outflow of individuals from farm origins to skilled and unskilled working classes, which is a typical mobility contour in developing countries in the course of industrialization, had not happened in large scale. For instance, for men of farm origin born in 1932-1940, none had became non-skilled manual workers and 22.1 percent became skill manual workers, which are lower than other countries for which we have data. The rates are even lower for the younger cohorts: only 11.3 percent for those born in 1941-1949, 6.1 percent for those born in 1950-1958, 5.1 percent for those born in 1959-1967, and 4.4 percent for those born in 1968-1976, had became non-skilled manual workers; and only 21.5 percent, 9.0 percent, 10.6 percent, and 6.8 percent of the respective cohort became skill manual workers. About 60 percent stayed in the farm class as their father, and such rates are quite stable across cohorts except for the youngest one. Similar patterns can be observed for women.

Interestingly, the outflow rate from farm to service class, a long-distance mobility, is indeed higher than that to working class. Wu and Treiman (2004, 2007) attributed this pattern to the Chinese household registration, which blocked the mobility for the majority of those born in rural agricultural backgrounds from moving out of farming but did allow a small percentage of them, who succeeded in the strict selective process with the criteria imposed by the government, to move into cities and work there. Because higher education is a main basis for the change of registration status, those able to overcome the structural barriers and move into cities typically ended up in service class (professionals and managers).

Table 4 also shows another abnormal pattern of reverse mobility from non-farm

－ 111 －

class to farm class, which is rarely seen in other countries. From men from service class, 13.2 percent of those born in 1932-1940, 13.0 percent of those born in 1941-1949, 9.2 percent of those born in 1950-1958, 6.8 percent of those born in 1959-1967, and 10.3 percent of those born in 1968-1976 have become a farmer. This pattern has been reported by Cheng and Dai (1995) but misinterpreted as a result of the Maoist sent-down policy implemented during the Cultural Revolution from 1966 to 1976. Wu and Treiman (2005) argued that this conclusion is not sound, since most youths and intellectuals “sent down” during the Maoist era had returned to the cities well before 1996 (Zhou and Hou 1999). They demonstrated that the high rate of downward mobility into farm class is only limited to those with rural hukou status at age 14 (for which I do not have information here). The hukou system blocked opportunities for the rural majority. Unless they changed hukou status, the prospects of the children of peasants are tenuous even when the father had left agriculture and worked in rural industry or services (e.g., as a village cadre or accountant).

Table 5 shows the inflow rate into class destinations by cohort and gender. Again, I pay special attention to the service class and farm class. Over 90 percent of farmers came from farm backgrounds for all cohorts and the percentages did not declined across cohort. However, for the service class, few came from farm origins. On the other hand, the percentage of professionals and managers who were the children of a professional or manager remained quite high for all cohorts. For men, the percentages are 19.1, 9.4, 26.1, 31.2 and 22.3 for the five cohorts; for women, the percentages are 17.5, 21.9, 37.6, 35.7, and 37.3 for the five cohorts, respectively. Notably, the percentages of petty bourgeoisie who came from the same family backgrounds decline sharply across cohorts, suggesting that three decades of socialist experimentation has successfully eliminated the transmission of family advantages through economic capital on the one hand, but high social reproduction rates among the farm and service classes. The latter is not consistent with the claims on the societal openness under state socialism.

The above description of mobility tables reveals some idiosyncratic features of Chinese class structure in the course of its socialist industrialization from 1950s to 1970s and the market transition in the 1980s and 1990s. However, the pattern observed here might be related to the change in the marginal distribution of both class origins and class destinations. For this reason, I now turn to analyze the trend of relative mobility rates with log-linear models.

5. The Relative Mobility Rates In the analysis of relative mobility rates, I employ the standard 6-category EGP class schema and ignore the unique features of Chinese social structure associated with its state socialist economy. For comparative analysis, I fit core models of social mobility originally proposed in CASMIN project by Erikson and Goldthorpe (1992). The core models have been proved to fit the mobility patterns for most countries, including both capitalist and socialist, European and non-European (e.g., Japan), developed and developing. The exercise in this paper is to test whether the model fits Chinese mobility pattern, given the unique social institutions and development path described above.

Our key interest is to examine whether relative mobility chances have changed in China in the course of its economic industrialization in the past 50 years or so (i.e., for those born between 1932-1976). I employ a series of loglinear models and compare their

－ 112 －

goodness of fit. The first model is Conditional Independence model, which presumes no association between origin and destination, namely,

(1) τττττη DT

jt

OT

it

T

t

D

j

O

iijtF =

where Fijt refers to the expected frequency in cell (i, j, t) of origin by destination by cohort mobility table, η is a scale term, τi

O is the main effect of class origin, τjD is the main effect

is the main effect of class destination, τtT is the main effect of cohort, and τit

OT and τjtDT

represent the effect of two way interaction between origin and cohort and between destination and cohort. There is no two-way interaction between origin and destination in this model.

The second model is Constant Social Fluidity (CSF) model, presuming a two way interaction between origin and destination, which does not vary by cohort:

(2) ττττττη OD

ij

DT

jt

OT

it

T

t

D

j

O

iijtF =

Note that the association between class origin and destination may be represented by the model of core social fluidity, which are composed of different effects to capture different aspect of mobility. Erikson and Goldthorpe (1992) argued and demonstrated that these effects operated in the mobility tables for any industrial countries. Here I employed the modified version of effect matrix from Ishida, Muller and Ridge (1995) for 6 by 6 mobility table mention earlier in this chapter.

First of all, there is an inheritance effect, referring to the propensity of individuals to stay in the same class of their fathers’. The second effect includes two hierarchy effects (HI1 and HI2). The third matrix captures the two positive effect affinity effects indicating relatively easy flow between two particular classes (for details can be found in Erikson and Goldthorpe 1992, Ishida, Muller and Ridge 1995). Thus, the log- linear CSF with effect matrix can be represented as

ττττττττττη BAF

ij

AAF

ij

HI

ij

HI

ij

DIGk

ij

DT

jt

OT

it

T

t

D

j

O

iijtF 2221= (3),

and the effects do not vary by birth cohorts. Finally, I allow the association between origin and destination to vary uniformly

across cohorts (T), using the model proposed by Xie (1992). Despite the fact that the core model captures the mobility pattern in all industrial societies, it may not fit the Chinese case, as I have already observed some idiosyncratic patterns in the mobility tables. Therefore, I also fit the model with additional designed matrix to capture the higher exchange rates between farm and service classes. I will empirically examine whether the core model can capture the social mobility pattern under Chinese socialism.

Table 6 presents the fit statistics of Constant Social Fluidity Model and other related models for intergenerational mobility tables for men and women, respectively. The conditional independence model assumes no association between class origins and destination, which yields very poor fit (G2 =1429.55, df=125, and p=.000). Even after taking into account of the large sample size, the BIC (357.52) statistic indicates the model still does not fit the data well. The constant social fluidity models, however, yields relatively better fit (G2=139.93, df=100, BIC=-717.88).

I then estimate the core model with effect matrix (Appendix 1), with the model fit significantly improved. According to BIC statistic, the CSF model with effect matrix fits the data better than the CSF model without the effect matrix (BIC=-786.24), with only 5.4 percent of cases misclassified.

Because of the rural-urban divide and high mobility rate from farmer class to

－ 113 －

service class and to petty bourgeoisie class, and high mobility from service class to farmer class (for those who unable to change the hukou status, see Wu and Treiman 2007) , I add an additional matrix to capture the positive affinity specific to institutional contexts in China (AF2C). The CSF with modified matrix fits the data even better, according to the BIC statistic (-794.76 for men and -803.38 for women).

To test whether the association between origin and destination varies by cohort, I further fit a log-multiplicative model proposed by Xie (1992) and Erikson and Goldthorpe (1992), with full two-way interaction and with effect matrix. According to BIC statistics, both models do not significantly improve the fit over the CSF core model. Hence, the CSF model with effect matrix (China variant) is preferred for both men and women.

Table 6 also presents the estimated parameters for unidiff models. Over the five birth cohorts, the parameters fluctuate, first increasing and then decreasing, but the variation is not statistically significant. In any case, CSF model with effect matrix (China variant) is the preferred model and there is no significant change in the association between class origins and destinations across different cohort, both for both men and for women.

Notwithstanding the core model is the best fitted model, Chinese mobility pattern still shows some idiosyncratic features. In Table 7a, I present the estimated effect-parameters represented in Figure 1, for men and women separately. Similar to what has been observed in other societies, the inheritance parameter for each class origin and destination is statistically significant, and the inheritance rates among farm class and service class are particularly high.

In addition to inherence parameters, there are two hierarchy effects, HI1 and HI2, to capture any movement of occupational hierarchy. First, HI1 divides all six classes into three hierarchical levels by separating service class (I+II) on the top and the non-skilled working class (VIIa) at the bottom of the hierarchy. HI2 specifies additional asymmetry effect involving the farm class - the change in the status of farming - between two generations. In western industrialized societies, farming class in the father’s generation is mostly composed of peasants based on largely subsistence agriculture while farming is more commercialized and market oriented in the children’s generation (Erikson and Goldthorpe 1992). Other two positive affinity effects, AF2A and AF2B, are meant to capture the easy mobility between particular classes.

In general, HI1 and HI2 are expected to be negative for most countries which have been studied so far. However, in China, they are unexpectedly both positive and HI1 is also statistically significant. This suggests the general pattern of core model of social fluidity do not hold in China. It is believed that the rural-urban institutional divide and selective mobility, as well as state discriminating policy against petty bourgeoisie are mainly responsible for the patterns. Indeed, as Wu and Treiman’s (2007) analysis shows, the farmer and petty bourgeoisie are the two lowest ranked and least desired occupational classes. On the other hand, the long-ranged mobility from agricultural class to directly professional/managerial class, because of the selective mobility from rural to urban status on the basis of educational credential, is also a prominent feature that renders a weak occupational association across generations among those who were occupationally mobile.

To consider this unique pattern of social mobility, I add an additional matrix effect to recognize the positive affinity between farmer and professional/managerial classes and between farmers and petty bourgeoisie classes. Results show that the

－ 114 －

modified model fits the data the best and the effect is positive and statistically significant for both men and women. After taking into account of the effect, HI1, HI2, AF2A, and AF2b become insignificant for men, and HI2 effect even becomes negative. For women, none of the HI1, HI2 and AF2B effects is statistically significant.

In summary, the analyses above show that the Chinese mobility pattern does not fit the core model and the origin— destination association differs from that in other countries. Despite the idiosyncratic features, the constant social fluidity model still holds.

6. The Role of Education in Class Mobility As education plays an important role in class mobility and reproduction, I now turn to analyze the association of class origins and education and between education and class destinations. I code education into three levels: high, medium and low. The high level includes both tertiary and specialized secondary education; the medium level includes senior high school, and the low level includes junior high school or below. Table 7 presents the distribution of educational levels across cohort for men and women separately. As the table shows, men are more educated than women for all cohorts. The educational expansion in China occurred largely at the medium level, and the expansion of high education occurred only recently.

I examine the trend of association between class origin and education and education and class destination through the cohort analysis. In the upper panel of Table 8, I present the fit statistics of different log-linear model for class origin by education by cohort for men. Model 1 is the conditional independence model, suggesting that class origin does not have effect on educational attainment, which yields a poor fit according to both the p value and BIC statistic (G2=522.28, d.f=50, BIC=93.54). The conditional constant social fluidity model (Model 2) assumes the existing association between origin and education does not change over time. Following Ishida, Muller, and Ridge (1995), I also represent the associations with different effect matrix to save degree of freedom. For the three-way table of class origin by education and by cohort, the equation can be written as

ττττττττττη UMO

ik

SMO

ik

FMO

ik

PMO

ik

MO

ik

SO

ik

SO

ik

T

t

E

k

O

iijtF 2222121= (4). where Fijt or refers to the expected frequency in cell (i, j, t) of origin by destination by cohort mobility table, η is a scale term, τi

O is the main effect of class origin, τkE is the

main effect of education attainment, and τtT is the main effect of cohort. And the rest of

parameters represent the effect matrix in Appendix 2. To inspect the temporal trends in the association between social origin and

education change, I fit uniform difference models with and without matrices effects. The constant association model and the log- multiplicative layer model with or without effect matrix all fit the data well, but the latter does not significantly improve the model fit, with additional 3 degree of freedom consumed. Based on the BIC statistics, I select the constant fluidity model with effect matrices as the preferred model (G2 =88.67, d.f.=43, BIC=-280.11).

I replicate the above models for women in the lower panels of Table 8. Similarly, for women, the constant association model is preferred. Therefore, the impact of class origin on education does not vary by cohort, despite of the 3-decade long experiment of state socialist egalitarian policies in China.

－ 115 －

Table 8a presents the estimated parameter for the effect matrix for the two-way table between origin and education (since the association does not vary by cohort). As we can see, unlike what have been reported from other countries, only 2 out of the 7 effects are statistically significant (individuals from petty bourgeoisie and farm origin are more likely to achieve low education. The results suggest the impact of state policy discriminating against farmers and self-employed business owners under state socialism. Other than that, family origin seems to have no effects on education attainment in China.

To what extent such a pattern mediates the association between origin and destination? I examine first the effect of education on destination and then the origin-destination association after controlling for the origin-education and education-destination association. In Table 9 I present the fit statistics of various models for the four–way tables (origin by destination by education by cohort) for men and women separately.

Model 1 allows a full two-way interaction between education and class destination, which yields an acceptable fit (BIC=-2114.66). I also represent the associations with different effect matrices (see Appendix 3) to save degree of freedom (Model 2). I then let the association parameters vary by cohort (Models 3 and 4). Comparing the BIC statistic for the four model, Model 1 yields the best fit for men (BIC=-2114.66) and for women (BIC=-2207.96), suggesting that class destination is associated with education qualification, and the association does not change across five birth cohort.

Despite the fact that Model 2 is not preferred, the estimated matrices effects are presented to show how education affects class destination. For example, the positive and large coefficients for HQ1, LQ2P, and LQ2 suggests that, for both men and women, those with high education are more likely to be in service class (I+II), and those with low education are more likely to become petty bourgeoisie and farm classes. The negative coefficients for HQ2 suggest that people with higher education are less likely to become farmers, skilled and unskilled workers.

After taking into account of the mediating effect of education, I re-estimate the association between class origin and class destinations with full two-way interaction (Model 5) and with effect matrices (Model 6). I let the association in respective models vary with cohort (Model 7 and Model 8). The fit statistics are presented in the lower panel of Table 9. Based on BIC criteria, Model 6 is preferred for both men and women, which means that the origin-destination association is constant over time, even after controlling for the origin-education association and education-destination, and their interaction with cohort. The estimated matrices effects are presented in Table 9a. Hence, the thesis of constant social fluidity is supported.

－ 116 －

7. Summary and Conclusion

To sum up, this paper studies the intergenerational social mobility pattern in the People’s Republic of China in a comparative perspective, based on an analysis of five cohorts from two national representative surveys conducted in 1996 and 2005, respectively. In the context of rapid economic industrialization and massive social changes since 1949, the absolute mobility rates in China increased over time, but the level is much lower than those in other countries. The relative mobility patterns were stable for both men and women, despite the initial implementation and subsequent dismantling of socialist egalitarian programs. Education played an important role in intergenerational social reproduction: the effect of origin on educational attainment does not change over time, and neither does the effect of education on destination.

Despite the similarity, the analysis also has revealed some unique features of Chinese mobility pattern different from the core models of social mobility. First, the occupational hierarchy presumed in the core models does not hold in China. In particular, farmer and petty bourgeoisie seem to be the lowest and second lowest ranked classes because of transmission of private property (land or capital) have been blocked in China under socialism. Second, under the household registration system (hukou) that the Chinese government installed to control rural-urban migration, for a long period of time, only those able to receive tertiary school education were allowed to move into cities, who typically end up in service class, hence, opposite to the core models, the long-ranged mobility from agricultural class directly to professional/managerial class is most likely to happen, and downward mobility from (rural) professional/managerial class are also very likely to happen if the respondents were unable to change their hukou status (see analysis in Wu and Treiman [2007]). Finally, positive affinity between farmer and petty bourgeois suggests that people from farmer backgrounds may also be likely to become self-employed in rural areas (if unable to achieve mobility through receiving education), but the positive affinity between farmer and un-skilled workers seem to be insignificant. This is at odds with the pattern of occupational mobility in most countries, industrialized and industrializing, in which farmers moved to become unskilled manual workers from agriculture. It is suggested that the pattern might be the results of the socialist industrialization dominated by the state for the most period studies, in contrast to market –driven industrial growth.

While I refer to the unique institutional arrangements and development paths to understand the idiosyncratic features of Chinese mobility pattern, under a comparative framework, I have no measure of the institutional variables (such as hukou origin) to examine their explanatory power. Even though log-linear analyses do not show significant variations in relative mobility rates (more like a U shape or inverse U shape), it might be resulted from a mixture of cohorts subject to different mobility regimes. I hope future research will be able to address these questions with more recent and detailed data on relevant information.

References Bernstein, Thomas P. 1977. Up to the Mountains and Down to the Villages: The Transfer of Youth from Urban to Rural China. New Haven, CT: Yale University Press.

－ 117 －

Bian, Yanjie. 1994. Work and Inequality in Urban China. New York: SUNY Press. Blau, Peter M., and Danqing Ruan. 1990. “Inequality of Opportunity in Urban China and America.” Research in Social Stratification and Mobility 9:3 -32. Breen, Richard (ed.). 2004. Social Mobility in Europe. Oxford: Oxford University Press. Chan, Kam Wing. 1994. Cities with Invisible Walls: Reinterpreting Urbanization in Post- 1949 China. Hong Kong: Oxford University Press. Cheng, Yuan, and Jianzhong Dai. 1995. “Inter-generational Mobility in Modern China.” European Sociological Review 11:17-36. Erikson, Robert, and John H. Goldthorpe. 1992. The Constant Flux: A Study of Class Mobility in Industrial Societies. Oxford: Clarendon. Featherman, David L., and Robert M. Hauser. 1978. Opportunity and Change. New York: Academic Press. Featherman, David L., F. Lancaster Jones, and Robert M. Hauser. 1975. “Assumptions of Social Mobility Research in the United States: The Case of Occupational Status.” Social Science Research 4:339-60. Ganzeboom, Harry B. G., and Donald J. Treiman. 1996. “Internationally Comparable Measures of Occupational Status for the 1988 International Standard Classification of Occupations.” Social Science Research 25:201-39. Ganzeboom, Harry B. G., Ruud Liujkx, and Donald J. Treiman. 1989. “International Class Mobility in Comparative Perspective.” Research in Social Stratification and Mobility 8:3-84. Gerber, Theodore P. and Michael Hout. 2004. “Tightening Up: Declining Class Mobility during Russia’s Market Transition.” American Sociological Review 69:677-703. Goldthorpe, John 1987. Social Mobility and Class Structure in Modern Britain. Second edition. Oxford: Oxford University Press. Grusky, David B., and Robert M. Hauser. 1984. “Comparative Social Mobility Revisited: Models of Convergence and Divergence in 16 Countries.” American Sociology Review 49:19-38. Ishida, Hiroshi, John Goldthorpe and Robert Erikson. 1991. “Intergenerational Class Mobility in Post-war Japan.” American Journal of Sociology 96: 954-992. Ishida, Hiroshi, John Ridge and Walter Muller. 1995. “Class Origin, Class Destination, and Education: A Cross-national Study of Industrial Nations.” American Journal of

－ 118 －

Sociology 101:145-193. Liang, Zai, and Zhongdong Ma. 2004. “Chinese Floating Population: New Evidence from the 2000 Census” Population and Development Review 30:467-88. Lin, Nan, and Yanjie Bian. 1991. “Getting Ahead in Urban China.” American Journal of Sociology 97:657-88. Lin, Nan, and Wen Xie. 1988. “Occupational Prestige in Urban China.” American Journal of Sociology 93:793-832. Lipset, Seymour Martin, and Reinhard Bendix. 1959. Social Mobility in Industrial Society. Berkeley: University of California Press. Lu, Yilong. 2003. Huji Zhidu:Kongzhi yu Shehui Chabie (The Household Registration System: Control and Social Differentiation). Beijing: Shangwu Press. (In Chinese) Meisner, Maurice. 1999. Mao’s China and After: a History of People’s Republic. New York: Free Press. Parish, William L.1981. “Egalitarianism in Chinese Society.” Problems of Communism 29:37-53. _____. 1984. “De-stratification in China.” Pp. 84-120 in Class and Social Stratification in Post-Revolution China, edited by James L. Watson. New York: Cambridge University Press. Treiman, Donald J. 1998. Life Histories and Social Change in Contemporary China: Provisional Codebook. Los Angeles: UCLA Institute for Social Science Research. Walder, Andrew G. 2002. “Markets and Income Inequality in Rural China: Political Advantages in an Expanding Economy.” American Sociological Review 67:231-53. Wong, Raymond S.K. 1990. “Understanding Cross-National Variation in Occupational Mobility.” American Sociological Review 55:560-573. _____. “Vertical and Non-vertical Effects in Class Mobility: Cross-National Variations.” American Sociological Review 57:396-410. Wu, Xiaogang, and Donald J. Treiman. 2004. “The Household Registration System and Social Stratification in China 1955-1996.” Demography 41:363-84. _____. 2007. “Inequality and Equality under a State Socialist Regime: Occupational Mobility in Contemporary China.” American Journal of Sociology 113:415-45. Xie, Yu. 1992. “The Log-Multiplicative Layer Effect Model for Comparing Mobility Tables” American Sociological Review 57:380-395

－ 119 －

Zhou, Xueguang, and Liren Hou. 1999. “Children of the Cultural Revolution: the State and the Life Course in the People’s Republic of China.” American Sociological Review 64:12-36.

－ 120 －

－ 121 －

Table 1.Change in Economic Structure and Employment Structure in China, 1952-2005

Economic Structure Employment Structure Year Total GDP

(RMB Yuan) Primary Secondary Tertiary Employment Primary Secondary Tertiary

1952 679.0 50.5 20.9 28.6 20729 83.5 7.4 9.11955 910.0 46.3 24.4 29.3 22328 83.3 8.6 8.21957 1068.0 40.3 29.7 30.1 23771 81.2 9.0 9.81960 1457.1 23.4 44.5 32.1 25880 65.2 15.9 18.41962 1149.3 39.4 31.3 29.3 25910 82.1 7.9 9.91965 1716.1 37.9 35.1 27.0 28670 81.6 8.4 10.01970 2252.7 35.2 40.5 24.3 34432 80.8 10.2 9.01975 2997.3 32.4 45.7 21.9 38186 77.2 13.5 9.31980 4517.8 30.1 48.5 21.4 42361 68.7 18.2 13.11985 8964.4 28.4 43.1 28.5 49873 62.4 20.8 16.81990 18547.9 27.0 41.6 31.3 64749 60.1 21.4 18.51995 58478.1 20.5 48.8 30.7 68065 52.2 23.0 24.82000 89442.2 16.4 50.2 33.4 72085 50.0 22.5 27.52005 183956.1 12.6 47.5 39.9 75825 44.8 23.8 31.4Data Source: Comprehensive Statistical Data and Materials on 50 Years of New China, Complied by Comprehensive Statistics of National Bureau of Statistics. China Statistical Press.

Table 2. Trends in Income Inequality (Gini Index) in China, 1953-2003 Year GDP Per capita (RMB Yuan) Gini Index 1953 142 0.561 1964 208 0.328 1970 275 0.299 1975 327 0.286 1980 460 0.295 1985 853 0.300 1990 1,634 0.340 1995 4,854 0.290 2000 7,858 0.390 2003 14,040 0.449

Sources: http://www.wider.unu.edu/wiid/wiid- introduction-2004-1.htmComprehensive Statistical Data and Materials on 50 Years of New China,

Tabl

e 3.

Per

cent

age

Dis

tribu

tion

of C

lass

Orig

in a

nd C

lass

Des

tinat

ion

and

Tota

l Mob

ility

Rat

es b

y B

irth

Coh

orts

and

Gen

der

Me

omn

n

We

Cla

ss O

rigin

:

1932

-40

1941

-49

1950

-58

1959

-67

1968

-76

19

32-4

019

41-4

919

50-5

819

59-6

719

68-7

6I+

II

6.3

4.5

7.6

11.8

10.7

5.

36.

29.

510

.09.

8II

I 1.

1 2.

63.

34.

75.

7

2.1

2.1

3.6

4.9

5.4

IVab

9.

6 2.

41.

41.

20.

3

8.3

2.2

1.0

1.0

1.3

IVc+

VII

b 76

.9

80.7

74.2

65.3

64.2

79

.080

.073

.269

.668

.4V

+VI

4.1

7.5

6.8

10.3

10.5

3.

47.

07.

38.

78.

0V

IIa

2.0

2.4

6.7

6.8

8.7

2.

02.

85.

55.

97.

1

Cla

ss D

estin

atio

n:

I+II

20

.7

16.2

12.6

12.0

13.5

11

.18.

17.

26.

87.

8II

I 2.

1 5.

55.

57.

69.

7

3.4

8.3

7.2

9.8

17.4

IVab

4.

0 3.

76.

710

.18.

0

4.7

2.9

3.5

7.8

8.5

IVc+

VII

b 53

.1

57.3

53.1

41.4

33.4

62

.863

.863

.657

.247

.2V

+VI

12.7

10

.012

.416

.316

.6

8.1

10.4

10.2

10.2

13.4

VII

a 7.

3 7.

49.

812

.718

.9

9.9

6.4

8.4

8.3

6.1

To

tal M

obili

ty R

ates

45

.4

41.8

41.0

51.8

60.8

39

.136

.034

.740

.147

.5In

dex

of d

issi

mila

rity

29.4

23

.421

.223

.930

.8

19.8

1611

.915

.624

.4W

eigh

ted

sam

ple

size

31

6 10

6715

1714

8491

9

273

915

1490

1705

1165

－ 12

2 －

Tabl

e 4:

Out

flow

Rat

es to

Cla

ss D

estin

atio

n by

Coh

ort a

nd G

ende

r

O

utflo

w R

ates

to C

lass

Des

tinat

ion

Men

Wom

en

From

Cla

ss O

rigin

I+

II

III

IVab

IVc+

VII

bV

+VI

VII

aI+

II

III

IVab

IVc+

VII

bV

+VI

VII

aI+

II

19

32-4

0 63

.2

71.4

37

.013

.249

.911

.536

.4

68.5

23.4

5.6

28.8

22.0

19

41-4

9 34

.1

9.8

21.9

13.0

39.7

17.6

28.6

11

.017

.94.

920

.014

.3

1950

-58

43.1

22

.4

16.1

9.2

14.8

8.2

28.5

25

.33.

42.

514

.412

.5

1959

-67

31.7

26

.7

8.6

6.8

11.9

18.0

24.2

8.

012

.43.

112

.510

.1

1968

-76

28.2

15

.7

21.5

10.3

15.1

15.9

29.7

16

.716

.33.

06.

017

.5II

I

1932

-40

0 0

02.

30

15.3

3.3

8.6

6.4

3.0

08.

8

1941

-49

9.2

32.6

0

4.3

6.2

13.5

18.3

30

.15.

75.

621

.417

.0

1950

-58

9.6

22.6

5.

03.

011

.313

.515

.6

32.3

1.9

3.4

15.0

17.9

19

59-6

7 6.

8 17

.8

3.6

5.7

10.8

16.8

21.9

28

.25.

73.

926

.219

.1

1968

-76

17.5

13

.4

06.

313

.617

.921

.8

40.5

42.0

11.7

31.6

22.9

IVab

1932

-40

0 0

10.7

3.5

3.7

7.7

0 0

6.4

4.3

18.6

4.4

19

41-4

9 2.

0 5.

9 24

.63.

22.

14.

70

00

3.1

0.8

16.6

19

50-5

8 8.

0 11

.7

23.7

5.3

7.6

14.5

3.3

5.2

22.1

3.1

4.8

2.5

19

59-6

7 8.

6 9.

2 49

.110

.28.

58.

07.

4 14

.512

.17.

94.

55.

6

1968

-76

4.1

15.6

18

.08.

96.

13.

06.

3 10

.115

.07.

615

.110

.1IV

c+V

IIb

19

32-4

0 31

.9

0 18

.962

.922

.10

15.7

10

.236

.x72

.631

.720

.9

1941

-49

28.1

29

.4

26.9

65.3

21.5

11.3

32.9

22

.930

.474

.810

.49.

7

1950

-58

17.8

6.

5 19

.167

.69.

06.

126

.8

18.3

36.1

79.0

15.6

20.1

19

59-6

7 15

.3

8.8

6.5

57.7

10.6

5.1

24.0

17

.849

.474

.48.

116

.7

1968

-76

8.6

15.6

0

47.5

6.8

4.4

13.7

15

.87.

363

.46.

115

.0V

+VI

19

32-4

0 4.

9 14

.3

15.1

11.3

24.2

57.9

23.1

0

12.8

6.6

10.5

13.2

19

41-4

9 12

.0

21.3

17

.57.

922

.517

.814

.0

24.2

28.3

7.2

24.8

32.9

19

50-5

8 13

.2

13,7

12

.78.

735

.627

.68.

4 17

.825

.26.

630

.925

.9

1959

-67

23.2

26

.0

17.8

10.8

34.0

23.5

12.1

21

.011

.76.

228

.518

.7

1968

-76

19.5

12

.1

23.7

14.7

19.9

26.1

17.8

12

.310

.610

.431

.816

.9V

IIa

19

32-4

0 0

14.3

18

.36.

80

7.7

21.5

12

.814

.97.

910

.530

.8

1941

-49

14.6

1.

0 9.

06.

28.

035

.26.

2 11

.918

.14.

422

.69.

6

1950

-58

8.4

23.1

23

.56.

221

.730

.317

.5

1.2

11.4

5.5

19.3

20.5

19

59-6

7 14

.4

11.6

14

.48.

924

.228

.610

.4

10.5

8.7

4.5

20.2

29.9

19

68-7

6 22

.1

26.6

36

.812

.338

.632

.710

.7

4.5

8.8

3.9

9.4

17.7

－ 12

3 －

－ 12

4 －

Tabl

e 5:

Inflo

w R

ates

to C

lass

Des

tinat

ion

by C

ohor

t and

Gen

der

In

flow

Rat

es to

Cla

ss D

estin

atio

n

M

en

W

omen

Fr

om C

lass

Orig

in

I+II

II

I IV

abIV

c+V

IIb

V+V

IV

IIa

I+

II

III

IVab

IVc+

VII

bV

+VI

VII

aI+

II

19

32-4

0 19

.1

0 0

3.8

2.4

0

17.5

5.

10

1.3

15.2

11.6

19

41-4

9 9.

4 7.

5 2.

52.

25.

48.

9

21.9

13

.60

3.2

8.4

6.0

19

50-5

8 26

.1

13.4

9.

12.

68.

26.

5

37.6

20

.69.

04.

07.

819

.9

1959

-67

31.2

10

.5

10.0

4.4

16.8

13.4

35

.7

22.5

9.5

4.2

11.9

12.6

19

68-7

6 22

.3

19.3

5.

42.

712

.612

.5

37.3

12

.67.

32.

813

.017

.2II

I

19

32-4

0 3.

7 0

00

1.2

2.1

12

.7

5.1

00.

30

2.7

19

41-4

9 1.

6 15

.3

4.2

1.3

5.5

0.4

2.

9 7.

70

0.8

4.9

3.9

19

50-5

8 5.

8 13

.5

5.7

0.4

3.6

7.7

12

.5

16.0

5.4

1.0

6.2

0.5

19

59-6

7 10

.6

11.0

4.

31.

07.

54.

4

5.7

14.0

9.0

1.5

10.0

6.2

19

68-7

6 6.

6 7.

8 11

.02.

74.

18.

3

11.6

12

.86.

41.

85.

04.

0IV

ab

19

32-4

0 17

.2

0 25

.73.

411

.424

.2

17.5

15

.411

.34.

813

.012

.5

1941

-49

3.3

0 16

.21.

14.

23.

0

4.9

1.5

01.

06.

06.

2

1950

-58

1.8

1.3

4.9

0.5

1.4

3.3

0.

5 0.

36.

10.

52.

41.

3

1959

-67

0.9

0.6

5.8

0.2

1.3

1.4

1.

7 0.

51.

50.

81.

11.

0

1968

-76

0.5

0 0.

70

0.4

0.6

2.

7 3.

12.

30.

21.

01.

8IV

c+V

IIb

19

32-4

0 49

.0

85.5

66

.891

.168

.171

.6

39.7

69

.273

.491

.264

.363

.4

1941

-49

64.9

63

.1

69.8

92.1

63.9

68.2

48

.2

53.6

82.6

93.4

55.3

55.1

19

50-5

8 54

.0

41.3

58

.394

.652

.247

.0

25.3

34

.265

.290

.947

.447

.9

1959

-67

37.0

48

.6

65.9

91.0

43.3

21.4

32

.0

28.1

70.8

90.5

42.0

37.7

19

68-7

6 48

.7

42.1

71

.691

.356

.742

.0

26.3

47

.161

.491

.953

.043

.9V

+VI

19

32-4

0 9.

9 0

3.8

1.7

7.8

0

8.7

013

.41.

74.

33.

6

1941

-49

18.3

8.

4 4.

32.

816

.88.

1

17.3

18

.01.

81.

116

.724

.7

1950

-58

8.0

14.2

7.

71.

219

.715

.1

14.7

15

.310

.31.

822

.317

.0

1959

-67

10.2

14

.5

8.7

2.6

21.4

19.6

16

.1

23.5

5.0

1.2

24.4

21.4

19

68-7

6 11

.8

14.8

8.

02.

112

.621

.5

6.2

14.9

14.2

1.0

19.0

12.4

VII

a

19

32-4

0 1.

1 14

.6

3.8

09.

12.

1

4.0

5.1

1.9

0.7

3.3

6.3

19

41-4

9 2.

6 5.

9 3.

10.

54.

211

.4

4.9

5.6

15.6

0.4

8.7

4.1

19

50-5

8 4.

3 16

.4

14.3

0.8

14.8

20.5

9.

5 13

.64.

01.

813

.913

.4

1959

-67

10.2

14

.9

5.4

0.8

9.8

15.3

8.

7 11

.54.

21.

710

.721

.2

1968

-76

10.2

16

.1

3.2

1.2

13.6

15.0

16

.0

9.6

8.4

2.3

9.0

20.7

Table 6. Fit Statistics of the Constant Social Fluidity (CSF) Model and Other Models to Mobility Tables with Origin by Destination by Cohort G2

P D.F. Δ% P value BIC

Men 1.Conditional Independence 1429.55 125 19.9 0 357.522.Constant Social Fluidity (CSF) 139.93 100 4.5 0.005 -717.883. CSF with Effect Matrices 200.24 115 5.4 0.000 -786.24 CSF with Effect Matrices (China variant) 182.75 114 4.9 0.000 -794.763. Uniform Difference (Unidiff) 123.14 96 3.7 0.032 -700.355. Unidiff with Effect Matrices 180.73 111 4.6 0.000 -771.44 Unidiff with Effect Matrices (China variant) 160.18 110 4.1 0.001 -783.03 Parameters (1932-40) (1941-49) (1950-58) (1959-67) (1968-76) 1.000 0.932 1.455 1.284 1.135 Women 1. Conditional Independence 1821.24 125 22.4 0.000 743.232. Constant Social Fluidity (CSF) 124.55 100 3.5 0.049 -737.853. CSF with Effect Matrices 228.35 115 4.9 0.000 -802.29 CSF with Effect Matrices (China variant) 179.39 114 4.2 0.000 -803.384. Uniform Difference (Unidiff) 119.16 96 3.3 0.055 -763.42

5. Unidiff with Effect Matrices 224.96 111 4.8 0.000 -732.81 Unidiff with Effect Matrices (China variant) 173.18 110 4.0 0.000 -775.11 (1932-40) (1941-49) (1950-58) (1959-67) (1968-76) Parameters 1.000 1.222 1.340 1.387 1.260Δ is the index of dissimilarity, which indicates the percentage of cases misclassified.

－ 125 －

Table 6a. Estimated Parameters of Chinese Mobility Table for CSF Models with Effect Matrix Effect Metrics Parameter Men Women Core matrices China variant Core matrices China variant DIG (I+II) 1.125***

(0.183) 1.195***

(0.184) 1.436***

(0.204) 1.422***

(0.207) DIG(III) 0.680***

(0.201) 0.621**

(0.202) 0.734***

(0.167) 0.602***

(0.168) DIG(IVab) 1.581***

(0.273) 1.627***

(0.274) 0.304

(0.377) 0.451

(0.387) DIG(IVc+VIIb) 2.123***

(0.096) 2.484***

(0.120) 2.317***

(0.087) 2.748***

(0.106) DIG(V+VI) 0.493***

(0.129) 0.300**

(0.134) 0.526***

(0.127) 0.248

(0.133) DIG(VIIa) 1.057***

(0.168) 0.599***

(0.194) 0.969***

(0.188) 0.545***

(0.201) HI1 0.256***

(0.063) 0.145

(0.067) 0.191***

(0.069) 0.010

(0.073) HI2 0.108

(0.108) -0.212 (0.126)

0.083 (0.124)

0.141 (0.125)

AF2A 0.103 (0.140)

0.151 (0.141)

0.356*** (0.122)

0.484*** (0.127)

AF2B 0.318*** (0.078)

0.137 (0.087)

0.364*** (0.081)

0.174 (0.090)

AF2C China Variant - 0.520*** (0.104)

- 0.817*** (0.113)

－ 126 －

Table 7. Percentage Distributions of Education by Cohort and Gender 1932-1940 1941-1949 1950-1958 1959-1967 1968-1976Men High 5.8 6.8 5.7 10.7 25.2Medium 8.0 9.0 17.7 23.3 13.6Low 86.2 84.2 76.6 66.1 61.2N 315 1067 1515 1480 919 Women High 3.2 3.9 3.3 6.7 15.5Medium 3.8 5.1 11.2 17.0 10.1Low 93.0 91.0 85.5 76.4 74.4N(sample size) 272 915 1489 1704 1165

Table 8. Fit Statistics of the Constant Social Fluidity (CSF) Model and Other Models for the Origin by Education by Cohort Table

G2 D.F. Δ% P value BICMen 1.Conditional Independence 522.28 50 11.1 0.000 93.542.Constant Social Fluidity 75.69 40 3.0 0.001 -267.363.Uniform Difference 61.95 36 2.2 0.005 -246.80 1932-40 1941-49 1950-58 1959-67 1968-76 Parameter 1.000 0.737 0.541 0.690 0.8904. Constant Fluidity with Effect Matrices 88.67 43 3.5 0.000 -280.115. Uniform Difference with Effect Matrices 72.98 39 2.6 0.001 -261.50 (1932-40) (1941-49) (1950-58) (1959-67) (1968-76) Parameters 1.000 0.743 0.536 0.663 0.904 Women 1.Conditional Independence 706.02 50 12.2 0.000 274.892.Constant Social Fluidity 38.91 40 1.6 0.000 -305.993.Uniform Difference 28.41 36 1.6 0.361 -271.99 Parameter 1932-40 1941-49 1950-58 1959-67 1968-76 1.000 0.840 0.873 0.863 0.848 4. Constant Fluidity with Effect Matrices 40.04 43 1.6 0.600 -330.725. Uniform Difference with Effect Matrices 39.37 39 1.6 0.454 -296.91 (1932-40) (1941-49) (1950-58) (1959-67) (1968-76) Parameters 1.000 0.818 0.861 0.849 0.831Δ is the index of dissimilarity, which indicates the percentage of cases misclassified.

－ 127 －

Table 8a. Effect Matrix Parameters Men Women SO1 0.390

(0.228) 0.461

(0.238)

SO2 -0.273(0.199)

0.026 (0.192)

MO1 -0.231(0.200)

-0.250 (0.212)

MO2P 0.590(0.281)

*** 0.784 (0.306)

**

MO2F 1.160(0.170)

*** 1.792 (0.167)

***

MO2S -0.054 (0.191)

-0.061 (0.188)

MO2U -0.009(0.200)

0.113 (0.199)

－ 128 －

Table 9. Fit Statistics of Models for the Origin by Destination by Education by Cohort Tables G2

D.F. Δ% P value BICMen 1.OET, ODT, ED (full two-way interaction) 374.95 290 6.6 0.001 -2114.662. OET, ODT, ED Matrices 436.93 293 7.3 0.000 -2078.433. OET, ODT, ED (Unidiff T) 437.81 286 6.3 0.000 -2088.464. OET, OT, OD, ED Matrices (Unidiff T) 428.68 289 7.0 0.000 -2052.34Parameters (1932-40) (1941-49) (1950-58) (1959-67) (1968-76) 1.000 0.849 0.717 0.882 0.837 5. OET,EDT, OD (full two-way interaction) 402.68 350 6.8 0.027 -2602.026. OET,EDT, OD Matrices 432.19 364 7.1 0.008 -2692.707. OET, EDT, OD (Unidiff T) 380.64 346 6.2 0.097 -2589.728. OET, EDT, OD Matrices (Unidiff T) 404.87 360 6.5 0.051 -2685.68 Parameters (1932-40) (1941-49) (1950-58) (1959-67) (1968-76) 1.000 1.073 1.737 1.488 1.141Women 1. OET, ODT, ED (full two-way interaction)

294.89 290 5.1 0.409 -2207.96

2. OET, ODT, ED Matrices 331.08 293 5.5 0.062 -2197.673. OET, ODT, ED (Unidiff T) 293.29 286 7.0 0.371 -2175.044. OET, OT, OD, ED Matrices (Unidiff T) 325.89 289 5.4 0.067 -2168.33Parameters (1932-40) (1941-49) (1950-58) (1959-67) (1968-76) 1.000 0.998 0.890 0.943 1.097 5. OET,EDT, OD (full two-way interaction) 362.57 350 5.7 0.310 -2658.116. OET,EDT, OD Matrices 416.19 364 6.4 0.031 -2725.327. OET, EDT, OD (Unidiff T) 354.60 346 5.6 0.363 -2631.568. OET, EDT, OD Matrices (Unidiff T) 409.09 360 6.4 0.038 -2697.90 Parameters (1932-40) (1941-49) (1950-58) (1959-67) (1968-76) 1.000 1.275 1.382 1.401 1.145Δ is the index of dissimilarity, which indicates the percentage of cases misclassified

－ 129 －

Table 9a. Effects Matrices Parameters in Four-way Tables from Selected Models

Men Women ED Matrices (Model 2) HQ1 1.103 0.954 HQ2 -0.719 -0.921 LQ1 -0.652 -0.916 LQ2P 1.186 1.429 LQ2F 1.736 2.334 LQ2S 0.709 0.480 LQ2U 0.874 0.429 OD Matrices (Model 2) DIG (I+II) 1.019 0.127 DIG(III) 0.469 0.570 DIG(IVab) 1.555 0.324 DIG(IVc+VIIb) 2.345 2.375 DIG(V+VI) 0.318 0.364 DIG(VIIa) 0.779 0.712 HI1 0.164 0.083 HI2 -0.067 0.046 AF2A 0.013 0.318 AF2B 0.146 0.227 AF2C China Variant 0.586 0.412

－ 130 －

Appendix 1. Effect Matrices in Core Social Fluidity Model

Inheritance 1 0 0 0 0 0 0 2 0 0 0 0 0 0 3 0 0 0 0 0 0 4 0 0 0 0 0 0 5 0 0 0 0 0 0 6

HI1

0 7 7 7 7 7 7 0 0 0 0 7 7 0 0 0 0 7 7 7 7 7 7 0 7 0 0 0 0 7 7 7 7 7 7 0

HI2

0 0 0 0 0 8 0 0 0 0 0 0 0 0 0 0 0 0 8 0 0 0 0 0 0 0 0 0 0 0 8 0 0 0 0 0

AF2A

0 9 0 0 0 0 9 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0

AF2B

0 0 10 0 0 0 0 0 0 0 0 0 10 0 0 10 0 0 0 0 10 0 0 10 0 0 0 0 0 10 0 0 0 0 10 0

China AF2C

0 0 0 11 0 0 0 0 0 0 0 0 0 0 0 0 0 0 11 0 11 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0

－ 131 －

Appendix 2. Effect Martrics for Association between Class Origin and Education Education/Qualification Class Origin High Medium Low I+II SO1 - SO2 III - - - IVa+Ivb - - MO2F Ivc+VIIb MO1 - MO2F V+VI MO1 - MO2S VIIa MO1 - MO2U

Appendix 3. Effect Martrics for Association between Education and Class Destination Class Destination Education I+II III IVa+IVb IVc+VIIb V+VI VIIa High HQ1 - - HQ2 HQ2 HQ2 Medium - - - - - - Low LQ1 - LQ2P LQ2F LQ2S LQ2U See Ishida, Muller, and Ridige (1995)

－ 132 －

Social Mobility Trends and Education in Brazil＊

Carlos A. Costa-Ribeiro

(University Research Institute of Rio de Janeiro - IUPERJ)

Abstract This paper examines change and stability in the patterns of intergenerational class mobility in Brazil from during the 20 th century and from 1973 to 1996. This historical period is marked by fast industrialization until 1982 and economic stagnation afterwards. Absolute mobility rates are presented and analyzed for both men and women. The analysis of relative mobility rates is separated in three components. First, the analysis focus on class origin to educational qualification association, and indicates decline in strength of association. Second, the analysis of the association between educational qualification and class destination also become weaker across the years. Third, the association between origin and destination classes also weakens across periods. In contrast, the association between origin and destination does not decline across birth cohorts. The core social fluidity model of the CASMIN project is used but does not fit the data. However, a modification of this model fits the data well and helps to understand the Brazilian mobility pattern. All the analysis are implemented for men and women. Key words: Intergenerational mobility, trends, social class, education

1. Introduction There are many reasons why Brazil is an interesting case for the study of intergenerational mobility. The country has experienced rapid economic growth and industrialization between 1960 and 1980, which is a recent period in terms of comparative research. During the 1980’s and 1990’s the country experienced economic stagnation and crisis. The rapid change and subsequent stagnation had effects on many aspects of the social structure (Albuquerque, 1993), but we are just beginning to know about the consequences of these periods for social mobility (but see; Ribeiro 2007; and Pastore and Silva, 2000) 1 . Although researchers have been showing that social fluidity remained constant or increased in countries in the northern hemisphere (Golthorpe and Erickson, 1993; Hout, 1988, Breen 2004), little is known about developing countries that passed though really significant industrial changes in recent years. Moreover, the impacts of economic crisis on social fluidity are studied only in a few cases (Gerber and Hout, 2004). The Brazilian case presents both trends, one of rapid development until 1982 and another of crisis afterwards. Therefore, the study of social fluidity in the country is highly relevant for mobility research. What is even more important is the fact that Brazil ＊This paper was presented at the meetings of the Project on Social Mobility in Late-industrializing Nations. I gratefully acknowledge the comments from the members of the project. 1 For studies following the status attainment tradition see Haller (1991; 1997; 1996; 1981); Bills and Haller (1985); and Pastore (1979). Early studies in social mobility in Brazil were done by Hutchinson (1957, 1958, and 1996). For a literature review about mobility studies in Brazil see Silva (1999).

－ 133 －

has good data sets covering the period of rapid change and stagnation. In this paper I use these data sets collected in 1973, 1982, 1988 and 1996. While the two first data sets represent the period of fast development, the two last cover the period of economic stagnation. In addition, the country is a good case for the study of the role of education on social mobility. Educational provision in Brazil has been remarkably low during most of the 20th century. The fact that only recently this situation began to change can have important implications for social mobility trends analyzed herein. Studies have indicated that the expansion of educational provision does not diminish inequality of educational opportunity, and have therefore little impact on the expansion of fluidity (Shavit, and Blonsfeld, 1993). In Brazil, this expansion was recent and occurred fast throughout the period of economic development and stagnation. The study of the consequences of this expansion to social mobility also makes Brazil an interesting case to study. In order to analyze the relationship between origin and destination class and the role played by education on social mobility I analyze the various links between origin, education and destination. Since we are applying a class structure perspective in this book, I use a model designed by Ishida, Muller and Ridge (1995) to untangle these relationships. For the origin by destination social fluidity analysis, I will apply the core social fluidity model (Erickson and Goldthorpe, 1993; and Ishida, Muller and Ridge, 1995). Although this model has been for underestimating the role played by hierarchy in the mobility process (Hauser and Hout, 1993), it continues to be a relevant tool for comparative mobility research. It has very clear principles that can be compared across societies in a relatively easy way. Therefore, it is important to test in what measure this model is useful to study late industrializing countries, as well as which are the main departures from the core pattern in these countries. In order to analyze the origin to education, and education to destination relations I will use a methodologically similar model designed to capture inclusion and exclusion mechanisms. This model is also very useful for comparative research and informative about the main characteristics linking class of origin to educational levels, on the one side, and educational qualification to class destination, on the other side. The combined analysis of the relations between origin, education and destination across time is useful to understand the social mobility trends of increasing social fluidity that are observed in Brazil 2 .

In this paper, I will analyze data for men and women. Since women have been entering massively in the labor market during the period studied herein, it is also import to describe trends on their mobility chances. The main objective of the chapter is, therefore, to present trends in social mobility and social fluidity in Brazil in order to test important hypothesis about the impacts of industrialization on social stratification. Since Brazil had experienced both economic expansion and stagnation on recent periods it is an essential case to discuss these theories (MORE on hypothesis and theories). Apart from this introduction the paper has 9 more sections. The second section describes the Brazilian historical context or industrialization process. Section 3 is dedicated to present the data and the variables used. Section 4 describes changes in the class structure for men and women, and absolute mobility rates. Section 5 analyzes trends in the association of class of origin and educational qualification for men and

2 This was first shown by Ribeiro and Scalon (2001). Previous studies using data from 1976 indicate constant fluidity (Silva and Roditi, 1988).

－ 134 －

women. Section 6 presents analyses for the linkage between education and class destination. Section 7 presents the Brazilian version of the core fluidity model and analyses trends in social fluidity. Section 8 presents analyses that try to account for the main reasons that led to the increase in social fluidity in Brazil. Section 9 is the conclusion.

2. Historical context: Brazilian Industrialization Process Two distinct periods can be observed in Brazilian society during the second half of the XX century. The fist, from 1940 to 1982, was characterized by economic growth and industrialization, while the second, from 1982 to mid 1990’s, was marked by economic stagnation and deep inflationary crisis. The social structure of the country changed dramatically during the first period, while the crisis and stagnation of the second period slowed the pace of changes.

From 1940 up to 1980 Brazil was the country with the fastest economic growth in the world (Adelman, 2001; Abreu e Verner, 1997). During this period the economy experienced major transformations, and the country changed from a predominantly rural and agricultural society to an urban and industrial one. Figure 1 shows the steady growth in GDP per capita between 1940 and 1982, and the stagnation afterwards. It also indicates that this expansion was specially fast between 1967 and the end of the 1970’s, a period known as the Brazilian “economic miracle”, in which the country presented indexes of economic growth similar to those experienced today by China. This growth until the beginning of the 1980’s was driven mostly by the fast industrialization of the country, a consequence of favorable international economic situation and also of the policy of import substitution, that is, an economic policy in which there were many State incentives to substitute imported goods by nationally produced ones. In contrast to other Latin American countries in which this type of policy was mainly used to substitute consumption goods, in Brazil there was also a significant investment in capital goods industry. It was during the 1960’s and 1970’s that Brazilian society really became an industrial one. This trend came to a halt in the beginning of the 1980’s, because the economic boom, largely financed by state incentives and investments, could not continue in the new international economic order (Abreu, 1990). The international debt crisis had a big impact leading the country to a period of crisis and stagnation. From 1982 to mid 1990’s the country passed through severe inflationary crisis. In contrast to the previous decade, known as the “economic miracle”, the 1980’s and 1990’s are defined as the “lost decades”. Many social scientists argue that the industrial or modern transition in Brazil was incomplete (Bacha and Klein, 1989; Schawartzman, 1982; Fernandes, 1974; Hasenbalg and Silva, 1988).

The fact that the transformation of the 1960’s and 1970’s was driven by an industrial expansion can be seen by the changes in the labor force (Faria, 1989, Lopes, 1971). The percentage of the labor force in the industrial sector grew from 13% in 1960 to 20% in 1970 and 25% in 1980; then this figure declined to 20% in 2000. In contrast, the percentage in the primary sector diminished constantly and fast from 1940 to 1980. In the middle of this period, 1960, the primary sector still included 54% of the labor force, while in 1980 it included 29%. Although this trend was fastest until 1980, it continued throughout the next two decades and in 2000 about 18% of the workforce was in primary sector. The rapid decline of the rural sector also means that most workers in the labor

－ 135 －

force are either rural migrants or offspring of rural workers. In fact, the mobility data analyzed in this chapter indicate that until 1996 at least 60% of all workers were sons or daughters of rural employees or farmers (Ribeiro, 2003 and 2007). It is important, therefore, to understand how different the rural society in Brazil is from that of European and other nations. Brazilian rural structure has always been characterized by huge inequality in access to land, that is, the majority of the land is owned by few proprietors (Leal, 1976). From 1940 to 1996 about 30% of the total farmland was owned by about 0,05% of the total landowners 3 . These are the large proprietors 4 , who used to employ large number of workers in the past but increasingly adopted industrial forms of production and diminished the need for labor. During the same period between 15% and 18% of the total farmland was divided among small farmers who account for more than 90% of all landowners. The majority of these small farmers are subsistence producers who almost always have very low living standards. In fact, rural workers and small farmers in Brazil are characterized as the poorest sector in the country. To have a rural origin really means to come from a particularly deprived family. It is important to highlight that this high level of inequality in farmland ownership is constant during the whole period from 1940 to 1996.

During the same period that industry was expanding and agriculture shrinking the service or third sector was growing. Although this increase occurred throughout the period of economic development and stagnation, it is marked by two distinct phases. Until 1970 the growth of the service sector was mostly driven by the participation of male workers, but afterwards it has been characterized by increasing entrance of women in the labor force. In spite of the lack of growth, during 1980’s and 1990’s the economy was able to incorporate large numbers of women, and unemployment fluctuated between 4% and 8%.

Although unemployment cannot be considered high, the division between modern and traditional sectors in the economy has been present since the early period of industrialization in the 1950’s. Industrialization did not translate in an elimination of traditional sectors in the labor market. Jobs in traditional industry (such as semi industrial clothing and food, and civil industry), and lower service sector (domestic service, and street vendors, for example) are a characteristic of Brazilian, as well as many other Latin American labor markets. According to some interpreters, since the Brazilian population was large the economic expansion of the 1960’s and 1970’s could happen without including the majority of the workers (Bacha, and Taylor, 1976; Barros, 2000), what means that income inequality, for example, looms high since the 1970’s. The argument is that the relatively small proportion participating in modern sectors of the economy (sophisticated services and modern industry) was already a large absolute number that could sustain the market for luxury services and products that was responsible for the economic miracle. Parallel to this modern sector, a traditional sector remained alive and active (Tavares and Serra, 1971). This traditional sector is not restricted, but includes what is usually called the informal sector. 5 Although unemployment was not too high during the 1980’s and 1990’s, the informal sector

3 See table in appendix of Ribeiro (2007). 4 The probability of selecting these large proprietors even in large population surveys as the ones used in this paper is extremely low. 5 The informal sector is not well mapped in the class schema used in this chapter, since it can be represented either by class IVab or VIIa.

－ 136 －

included more than 50% of the urban population during the 1990’s (Cardoso, 2002). In addition to labor market and the rural society, the characteristics of the

educational system are another important feature related to the mobility trends presented in this chapter. During the period of economic expansion and stagnation, there were also some important fluctuations in the educational system. Figure 2 presents trends of gross educational enrollment rates from 1960 to 2000. The first thing that should be noticed is that in 1971 there is a break in the lines for elementary and middle school, and new lines for primary and secondary education initiate. In 1971, there was a major educational reform in Brazil. Until this date the pre tertiary educational system was divided in two sections, one of four years of compulsory elementary education, and another of seven years of middle school education. The reform of 1971 expanded the years of compulsory education from 4 to 8, and divided the system into two parts one of primary education (8 years) and another of secondary education (3 or 4 years). This division is important to the analysis presented in this paper because at least 80% of all respondents in the four surveys were in their school years before the 1971 reform 6 . Therefore, for these people the compulsory education was only four years. When compared to other countries in this study, educational attainment in Brazil is particularly low (see also Castro, 1989; and Ribeiro and Klein, 1991). In fact, figure 2 indicates low rates of primary education attainment, since even in 1971 there were only 70% of the children enrolled in this educational level. This situation began to change only in the mid 1980s, when the federal government started to promote primary schools, but it was only at the end of the 1990’s that primary education was finally universalized. Educational expansion trends at the tertiary levels grew fastest from 1965 to 1975, and stagnated in the 1980’s when a growing secondary/tertiary gap started, which resulted in a bottleneck in the transition to postsecondary education (Ribeiro, 2007b).

3. Data and Variables

3.1. Data The data used in this study come from the Brazilian National Household Survey (“Pesquisa Nacional por Amostragem Domiciliar”, PNAD hereafter) collected by the Brazilian National Bureau of the Census (“Instituto Brasileiro de Geografia e Estatistica, IBGE). The PNAD are probability, stratified, multistage surveys of Brazilian households that have been collected almost every year since 1967, with exception of Census years. Initially it was not a national survey, but since 1973 it has been based on a national sample, with the exception of the rural area of the North region (the rainforest of Amazônia, which accounts for only 2.3% of the Brazilian population according to 2000 census). The questionnaire has a core part collected every year including information on household conditions and labor market participation of all residents 10 years or older, and a supplement addressing different questions in each year. In 1973, 1982, 1988 and 1996 the supplement applied to heads of household and spouses included questions on social mobility (father’s occupation when respondent was growing up, father’s education, and first job). These surveys, therefore, provide us with unique opportunity to conduct 6 This does not mean that most people have at least four years of schooling. In fact, educational provision, especially in the rural areas, has been extremely low during most of the 20th century.

－ 137 －

cross-temporal comparisons using virtually identical variables. The sampling is based on the previous Census (1970, 1980, and 1990) distribution

of municipalities and “census sectors” and follows a three step probabilistic procedure. First, counties are selected, then “census sectors” within counties were chosen, and finally households within sectors were selected. In this paper the age range is set to 30 to 64 years old and the usable sample sizes are: 30,742 men and 9,894 women in 1973; 53,506 men and 19,980 women in 1982; 29,966 men and 13,313 women in 1988; and 35,320 men and 22,975 women in 1996.

3.2. Variables:

I use information on current socioeconomic standing of the respondent (head of household and spouse) and retrospective information on parental characteristics to construct mobility tables cross-classifying respondents’ class of origin and class of destination. Class of origin measures the position of the respondent’s father when respondent was growing up. Class of destination is the respondent’s class position at time of survey. The class classification utilized is the six-fold CASMIN class schema used throughout this book, which distinguishes the following classes: Professionals and managers (I+II), clerical (IIIab), employers and self-employed (IVab), agricultural workers (IVc+VIIb), skilled manual (V+VI), and unskilled manual (VIIb). One particularity of the Brazilian data should be noticed. The classification into the CASMIN classes is made using only the occupational variable, since the variable for employment status is built in the occupational titles themselves (see Silva, 1988). For some occupational titles there are three alternative names one including “employer”, another “self-employed” and a third without specification (meaning employee). This is especially important for the definition of origin class, because there is only a father’s occupation variable but not a separate variable for his employment status, and thus to be consistent I did not use the separate variable for classifying the respondents occupation. The main consequence of this decision is that some of those self-employed that should be in class IVb are actually classified in class V+VI or VIIa. Using both respondents’ occupation and employment status variables to verify the extent in which this occurred for destination class, I observed that 14% (in each survey) of those who could be classified as IVb are in fact classified as V+VI or VIIa. I have done some sensitive analysis in order to check if the results change significantly when we reclassify those 14% into IVb for the destination class and the results did not change significantly either in relative or in absolute rate. 7 In addition to class of origin and destination, this study uses three other variables: gender, respondents’ education and birth cohort. All the analyses are done for men and women that were in the labor market in the moment of the surveys, all four surveys include information on men and women. Education is classified in four categories and collapsed into three categories for most of the analyses. Incomplete elementary includes all respondents who did not complete the four years of elementary

7 I have also done other reclassifications and the results remain consistent. For more information on this issue contact the author.

－ 138 －

education that were compulsory until 1971. Complete elementary include all those who completed elementary (4 years) or primary education (8 years). In the relative rates analyses these two categories are collapsed into one named: primary education or less. The two other categories are some secondary education and some tertiary education. I constructed the following eight birth cohorts: (1) 1909-1917, (2) 1918-1923, (3) 1924-1931, (4) 1932-1937, (5) 1938-1943, (6) 1944-1952, (7) 1953-1958, and (8) 1959-1966. With the exception of cohort 1 and 8, that are represented respectively in the first and last survey, all cohorts are represented with different ages in more than one survey (see table 1a in the appendix). For the cohort analysis I collapsed the cohorts across surveys, I also tested for age effects in each cohort across surveys. I present the cohort information for the absolute mobility rates. Since there is no change across cohort in social fluidity (the constant social fluidity model fits better then the unidiff model) and there is change in fluidity across the four surveys I will present in this chapter the period analysis. There are some evidences of a period effect in Brazil.

4. Class structure and absolute mobility

I begin with the examination of the trends in the distribution of class origin and class destination. Table 1 presents the distributions of class of origin, class of destination, the index of dissimilarity, the percentage immobility, and the percentage total mobility for men between 30 and 64 years old both across the eight birth cohorts and across the four survey years (1973, 1982, 1988 and 1996). Table 2 presents the same information for women between 30 and 64 years old. In the case of the data for women it is important to highlight that in the period considered the percentage of women participating in the labor marked increased significantly, they were: 37.4% in 1973, 37.8% in 1982, 44.8% in 1988, and 55.9% in 1996. Therefore, the increase in women participation in the labor market was faster after 1982 (see also, Carvalho and Wong, 1997). Since the trends for the cohort data go in the same direction as the trends in the four survey years, I will present below the interpretation for the later.

The most evident trend for both men and women is the rapid diminishment of the farming class. There is considerable reduction in its share in the class destination distributions. In 1973, 36.7% of the men and 35.4% of the women were in the farming class, in 1996 this percentages reduced respectively to 26.7% and 23.3%. The declining percentage in the rural classes is paralleled by a tendency of expansion in all urban classes. For men this tendency across the four survey years is more evident in the expansion of the working classes (V+VI and VIIa) and the service class (I+II), and for women in the expansion of the clerical class (IIIab) and the service class (I+II). These same tendencies for men and women are also observed across the birth cohorts, and look more dramatic here indicating that most farmers are older and members from the other classes are younger. Another distinctive characteristic of the Brazilian class structure is the concentration of workers in the manual working classes (rural and urban). In 1973, classes IVc+VIIb, V+VI and VIIa included 75% of all male workers and 76% of all female workers. In 1996, the share of these classes was 68% of all male workers and 63.5% of all female workers. 8 Another important difference is that men have higher

8 In another study I have used 16 class categories in order to make more distinctions within the working class. This expanded class schema proved to be very important to understand the patterns

－ 139 －

proportions then women in the small proprietors class (IVc), and women higher proportions then men in the clerical class (IIIab). The distribution of the class origins do not represent the class distribution in a previous period of time because the age of the fathers’ vary substantially, some fathers are also represented in the destination distribution, and fathers’ with more offspring (most likely farmers and farm workers) have higher probability of been represented in any survey sampling sons and daughters (Blau and Duncan, 1967). The origin class distribution must be interpreted as the origins that respondents have in each year. It is interesting to notice that the origin distribution of men and women are very similar, this is not a surprise since in both cases origin is defined by fathers’ occupation. In both cases the origin class distribution is marked by the high proportion of fathers in the rural class. In 1973, from each four men almost three (72%) came from rural origin and the same proportion was valid for women. There is almost no change from 1973 to 1982, thereafter there is a constant decline in the proportion of men and women with rural origins so that in 1996 about 60% of men and women had this origin. However, what is most striking in the Brazilian data is the disparity between origin and destination class distributions either for men or for women. A common way of summarizing this disparity is the index of dissimilarity, representing the percentage cases that should be reclassified to make origin and destination distributions equal. For both men and women this index increases a little from 1973 to 1982 then there is a constant decline until 1996. The increase from 1973 to 1982 is probably reflecting the fact that the rural urban transformation in Brazil was fastest during the 1960’s and 1970’s, and slowed after 1980. Among the nations studied in this book only South Korea presents an index of dissimilarity as high as the Brazilian one. Both countries passed through radical transformation from a rural to an urban social structure in recent times. Finally tables 1 and 2 present the rates of total mobility. This simple index, measuring the total percentage of men and women in a class position different than that of their fathers, indicates an increase in mobility from 1973 to 1982 and thereafter little change. In 1973 from all men 55.3% were mobile and from all women 57.5% were, in 1982 these numbers were respectively 58% for men and 68% for women. The rates of total mobility in Brazil are not higher then in other countries in this study or in other industrialized countries, as previous studies on social mobility in Brazil had stated (Pastore, 1981; and Pastore and Silva, 2000). However, these studies were not completely wrong because if we calculate the rates of upward and downward mobility using three hierarchical levels (classes I+II in the top, classes III, IVab, IVc and V+VI in the middle, and VIIa and VIIb in the bottom) it is possible to observe that there are 4.5 more times upward then downward mobility in Brazil (see also Silva, 1979). The only country with a higher ratio of upward to downward mobility is probably South Korea with a ratio of 6 times more upward mobility. All other countries in this book, the CASMIN project (Erickson and Goldthorpe, 1993), and Breen’s book (2004) have lower rations of upward to downward mobility (Ribeiro, 2007). Table 3 present selected outflow rates for men and women in Brazil in 1973, 1982, 1988 and 1996 and for men in the other countries studied in this volume. The outflow from class origin I+II to the same class destination (or immobility in class I+II) declined for Brazilian men from 55% in 1973 to 50% in 1996, for women there is no trend. The percentage immobility in the service class (I+II) in Brazil is similar to other and level of absolute and relative mobility rates. (Ribeiro, 2007).

－ 140 －

late industrialized countries with the exception of China that has extremely low rates of immobility in class I+II for obvious reasons since the country passed through a communist revolution (?). However, all the late industrializing countries in this volume have lower immobility in class I+II then the industrialized countries studied in the CASMIN project (Erickson and Goldthorpe, 1993). Another difference is that the late industrializing countries in this book have in general lower rates of mobility from the service class (I+II) to the working class (V/VI and VIIa) than countries in the CASMIN project. In Brazil, there is no clear trend either for men or for women in this mobility path. About one out of each five men was mobile from service class to working class, a percentage similar to that observed in Chile and Taiwan. Women with origin in the working class (V+VI+VIIa) have increasing chances of entering the service class (I+II) since the rate changed from 13% to 19% between 1973 and 1988, and remained constant until 1996. For men there is no clear trend, but the percentages are similar to those of Japan and Taiwan. Immobility in the working class declined significantly for women from 58% in 1973 to 48% in 1996. Again there is no trend for men since this immobility remained around 55% along the years, a percentage similar to that of Chile and not to different from that in the other countries, again with the exception of China. Immobility in the working class seems to be lower in the countries studied in this volume then in the industrialized countries included in the CASMIN project (Erickson and Goldthorpe, 1993). This immobility, combined with the low rates of self-recruitment, is a sign of lower demographic identity in the working class in late industrializing countries. Finally table 3 presents the outflows from the rural class (IVc+VIIb) to the service and working class. For men and women in Brazil the chances of mobility from the rural to the service class are remarkably low when compared to other countries (Korea and China seem also to have low rates). There is no temporal trend for women and a slight tendency for increasing mobility for men. These low chances of mobility from rural to service class are probably expressing the socioeconomic distance between these two classes that characterizes Brazilian society. On the other hand, as usual in many countries, mobility from rural to working classes (V/VI and VIIa) is very common in Brazil for both men and women and increased most from 1973 to 1982. From an inflow or recruitment perspective (table 4), 90% of men and 90% of women in working class (V/VI and VIIa) in Brazil have origins either in the working class or in the rural classes. From 1973 to 1996 the proportion of self-recruitment in the working class increased as the proportion with origins in the rural classes decreased. This proportion of 90% is larger in Brazil than in any other country in this volume. The percentage in the working class with origins in the rural class is similar in Brazil and Korea, the two countries in this volume that experienced the most dramatic rural to urban transition. This rapid transition implies, as I have indicated, that there is a huge proportion of sons and daughters of farmers and farm workers in the labor market. Therefore, it is not a total surprise that circa of 30% of the men and women in Brazil and men in Korea in the service class have origins in the rural classes. For the Brazilian case, there is a decline in this proportion from 1973 to 1996. The proportion in the service class with origins in the working class is also high and increased from 23% in 1973 to 29% in 1996 for men and from 16% to 25% for women. By an inflow perspective the service class in Brazil cannot be considered closed. When compared to European industrialized countries recruitment into the service class for the countries in this volume tend to be higher from

－ 141 －

the rural classes and lower from the working classes. Finally, table 5 presents the distribution of respondents’ education by gender and survey year. The four levels distinguished are incomplete elementary (3 or less years of schooling), complete elementary (4 to 7 years) or complete primary (8 years), secondary education and tertiary education. The two lower levels are combined for the relative mobility rate analysis because as in other chapters of this volume I apply the models designed by Ishida, Muller and Ridge (1995) for three fold educational divisions. The first aspect to be noticed in table 5 is the very low educational attainment of Brazilian men and women. From a comparative perspective Brazil is the country with the lowest educational attainment in this volume. Although there is considerable improvement across the years, the levels of educational attainment for the 30 to 64 years old population was still very low in 1996. As usual in many countries, the data indicates considerable improvement of women vis-à-vis men across the years, to the point that since 1982 women have higher rates of education attainment.

5. From Origin to Education: Relative Rates. In this section I follow the model proposed by Ishida, Muller and Ridge (1995) to represent in a simple way the relation of origin (O) to education (E) and education to destination (D). The idea is to unveil the complex relationship between origin, education and destination using a class structure perspective. They propose the estimation of the mechanisms of inclusion and exclusion using effect matrices to model the relation between origin (O) and education (E), as well as the link between education and destination (D). Although the mechanisms operate in different ways linking origin to education, and education to destination, the simple idea of inclusion and exclusion summarizes well both links. People with service class origins (I+II) have access to resources that facilitate educational attainment, while people with manual class origins do not. Similarly, people with higher educational qualifications posses some resources that facilitate entrance in higher classes positions, while people with lower educational credentials do not. The effect matrices for origin (O) to education (E) transition are designed to capture these inclusion and exclusion mechanisms (see design matrix in the appendix). First, people with origin in the service class have higher propensities to be included in tertiary education (effect SO1) and excluded from lower educational levels (effect SO2). Second, people with manual class origins (IVc+VIIb, V+VI and VIIa) do not have the cultural and economic capital necessary to attain higher educational levels being relatively excluded from these credentials (effect MO1). Finally, those with manual and petty bourgeoisie class origins have propensities for having lower educational positions, since they do not have the resources and the cultural capital to avoid this fate (effects MO2P, MO2F, MO2S, and MO2U). The model linking educational qualification (E) to class destination (D) follows the same dual principle. Highly qualified people tend to be included in service class (I+II) positions (effect HQ1), and to avoid entering manual positions (HQ2). While poorly qualified people tend to be excluded from service class (effect LQ1) and included in petty bourgeoisie, skilled manual, unskilled manual and farming classes (effect LQ2P, LQ2S, LQ2U, and LQ2F). In this paper I am interested not only in modeling these relationships among origin, education and destination, but also on analyzing the possible changes or

－ 142 －

continuities in the strength of the relevant associations across the years in Brazil for both men and women. In order to test if there is any trend in the origin by education association three specifications are used. First, I use the model of Constant Social Fluidity (CSF) to test the hypothesis that the sets of odds ratio pattern in the table are exactly the same across survey years. Although the distribution of class origin and educational qualifications may vary across years, the relative mobility rates are set to be exactly the same. In other words, the strength of the association does not vary across the years. A second model is the “uniform difference model” or, for short, UNIDIFF model (Xie, 1992), and it is used to test for any change in the strength of the association between origin class and education across survey years. The model represents the difference between two survey years in the pattern of association using a single uniform difference parameter. The comparison among these parameters is used to verify if there is any trend of increasing or decreasing association strength. I begin presenting the analysis for the origin by education by period table for men then I present similar analysis for women. Table 6 presents five models fit to the origin by education by period table for men. The first model (M1) is the independence model, used only as a baseline since it is well know that class origins always have an effect on educational attainment. As expected the model does not fit the data (L2 = 29913, df = 40, Bic = 29437). The second and the third models fit the origin (O) by education (E) by survey year (S) table using a full interaction pattern of association and hypothesizing respectively no change across time (M2) and time varying strength of the origin by education association (M3). 9 Model 3 (Bic = -205) fits better the data than model 2 (Bic=-172.9) indicating that the strength of the association between origin and destination declines significantly from 1973 to 1982, and in a slower pace in the following years. Models 4 and 5 also hypothesize no change and change in the strength of this association across time, but use the effect matrices modeling the inclusion and exclusion mechanisms described above. Again the model for time change in strength of association (M5, Bic = -233.9) fits better than the no change model (M4, Bic = -201). Therefore, I conclude that there is increasing fluidity in the origin by education transition in Brazil across the years studied. The pattern of association hypothesizing inclusion and exclusion mechanisms gives me some comparative possibilities. For the selected model (M5) the effect matrices parameters have the expected signs, what indicates that they work in the way hypothesized by the idea of inclusion and exclusion mechanisms. Men with service class origins (I+II) tend to be included in higher education (MO1) and to avoid lower education (MO2), and the parameters are similar in magnitude to those observed in industrialized countries (Ishida, Muller, and Ridge, 1995). Men with origins in the manual classes (IVc+VIIb, V+VI, and VIIa) tend to be excluded from higher education (MO1) and the magnitude of the parameter is similar to the one found for industrialized countries. Sons of skilled and unskilled manual workers (V+VI and VIIa) also have high propensity for attaining no education or lower education (MO2S and MO2U). The mechanism of inclusion of sons of small proprietors (IVab) on lower educational credentials group cannot be observed in Brazil, since the parameter (MO2P) is not significant, this is probably a fact that in Brazil small proprietors with employees (class

9 Since the Brazilian data sets have a very large number of cases it is difficult to judge the goodness-or-fit of models based on chi-square statistics. I will use, therefore, the Bic statistics proposed by Raftery (1986, 1995).

－ 143 －

IVa) have a socioeconomic situation comparable to that of some members from the service class (I+II) and therefore can avoid lower education for their offspring (Ribeiro, 2007). Finally, the link between origin in the rural classes (IVc+VIIb) and low educational credentials (MO2F) seems to be especially strong in Brazil when compared to industrialized countries. This is another evidence about the high levels of deprivation that characterize the rural classes in Brazil. Table 7 presents the origin by education by period analysis for women. The interpretation of the parameter estimates modeling the inclusion and exclusion mechanisms is exactly the same as that for men, so I will not repeat it here. However, there is one major difference, since for women it is the constant social fluidity model that fits better than the uniform difference, both when we use the full interaction pattern (M2 fits better than M3) or the matrix effects pattern of association (M4 fits better than M5). This means that there is no increase in fluidity in the linkage between class origin and educational qualification for women across the years from 1973 to 1996.

6. Educational Qualification and Class Destination: Relative Rates In this section I turn to the relation between educational qualification and class of destination in order to analyze the other side of inclusion and exclusion mechanisms. I first model the education to destination relationship across survey years using both a full interaction pattern and then the matrix effect pattern designed to capture inclusion and exclusion mechanisms described in the previous section. Once this models are fit and selected, I include origin in the analysis, that is to say I fit models to the education by destination by origin by survey year table. The idea is to compare the models without and with origin in order to see the impact of controlling for class of origin on the effects of educational qualification on class destination. For both analyses I use the models of constant social fluidity and uniform difference to capture possible trends across survey years. The analysis of the four-way table use the full interaction pattern for the OD and OE associations, and test both the full interaction and effect matrix patterns for the ED association.

Table 8 presents the analysis for the table crossing education (E) by destination (D) by survey year (Y) for men. The models of uniform change fit better than the CSF models either using a full interaction pattern (M2, Bic=-178, versus M1, Bic=42.3) or the matrix effect pattern of association (M4, Bic=-58.9, versus M3, Bic=160.7). The uniform difference parameter estimates for model 2 and 4 indicate that the strength of the association between educational qualification (E) and class destination (D) declined constantly from 1973 to 1996. Table 9 presents the same analysis for the four-way table (PEDO). All the models in table 9 fit the ODP and OEP interactions and let the DE association be constant or vary across period. As it happened with the models in table 8 the uniform difference models (M2, and M4) fit better than the constant social fluidity ones (M1 and M3), indicating again a decline in the association between education and destination. This declining association across the period is probably expressing the tendency for credentials obtained in the educational system to depreciate their value across the years from 1973 to 1996 (?).

It is also interesting to observe the impact of controlling for origin on the education to destination effect matrix pattern of association. In order to do that, I need to compare model 4 from table 8 with model 4 from table 9, what is presented in the last

－ 144 －

column of table 9 that displays the percentage reduction in the estimates of ED effect matrix parameters after controlling for class of origin. The main conclusion is similar to that reached for industrialized countries (Ishida, Muller and Ridge, 1995), the reduction are in general small, that is, educational credentials really help to allocate men to different class positions, in spite of their class origin. Also in a similar way to industrialized countries the only exception is on the propensity for people with low qualification to move into farming (LQ1F). Controlling for class origin reduces this propensity by one third (33%) in Brazil. As suggested by Ishida, Muller and Ridge (1995) part of this association is spurious because sons of farmers and farm workers tend to have low education and remain in the rural sector. Actually, there is a very small percentage of people from other origin class (either with low or with high education) entering in farming.

The other mechanisms of transition from education to qualification work in the expected way. High qualification (HQ1) grant to individuals the chance of entering in the service class (I+II) almost independently of their class of origin, since the reduction after controlling for origin is of only 6%, on the one hand, and help them to avoid manual working class destinations (HQ2) again with little effect from origin (reduction of 10%), on the other hand. In Brazil, the attainment of higher education is useful for both inclusion in the service class and exclusion from manual classes since both effects have similar magnitudes (HQ1 and HQ2). The main conclusion is that higher education is a very significant advantage in Brazil.

Men with low qualification have high propensities for entering the working classes (LQ2S and LQ2U). The effect of low qualification is even higher for entering in the unskilled working class (VIIa) than in the skilled working class (V+VI). Controlling for origin does not diminishes the propensity of people with poor educational qualifications to enter in the working class, what means that low qualification is really a disadvantage in Brazil even when compared to industrialized countries (the magnitude of the LQ2S and LQ2U in Brazil is among the highest when compared to industrialized countries). Finally, the propensity for those with lower education to enter in the petty bourgeoisie (LQ2P) in Brazil is not very high expressing that the inclusion into this class destination is not locked for those with higher education.

Tables 10 and 11 present the same type of education to destination analysis for women. In terms of patterns of association the results for women are very similar to those for men, but the difference in the magnitude of the effects between men and women indicate some interesting types of inequality. In comparison to men, women with high education have a lower propensity for avoiding working class destinations (HQ2), while those women with lower qualification have a higher propensity for being excluded from the service class (LQ1). This two patterns combined with the high propensity of highly educated women to enter service class (HQ1), when compared to the effects for men, suggest that women in the labor market have more difficulties than men to enter the service class and to avoid lower class positions, that is, when they have low education is even more difficult than it is for men to enter the service class and when they have high education it is more difficult, than it is for men, to avoid working class positions.

In addition, the propensity of women with low qualification to enter the skilled (LQ2S) and unskilled working class (LQ2U) is considerably higher than the equivalent propensities for men. The same is valid for the link between low qualification and the small proprietors class (LQ2P) and the farming class (LQ2F). The main conclusion from the comparison between men and women is that qualifications are even more important

－ 145 －

for women, since when they have higher educational qualifications they find relatively more difficulty than men to be excluded from lower class positions and included in higher classes positions in the labor market. One alternative explanation is that they are included in higher classes positions and excluded from lower ones via marriage market (Ribeiro, 2005 and 2007). The analysis about ED association indicates the high relevance of education to reach the top positions in the class structure and avoid the lower positions. Education is extremely important, and once achieved it is a powerful mean for eliminating the disadvantages of coming from a lower class origin. This power of education, however, is declining for men and women since the association between education and destination is getting less strong along the years in Brazil.

7. Origins and Destinations: Relative Mobility Rates To analyze trends in social fluidity from 1973 to 1996, I fit a series of log-linear and log-multiplicative models to the Brazilian mobility table crossing 6 classes of origin by 6 classes of destination by 4 survey years. I first analyze the data for men and then the data for women, and use the constant social fluidity and the uniform difference models discussed above. In addition, I used a specification of the UNIDIFF model that sets a linear or curvilinear trend to the uniform difference parameters. This is a very parsimonious specification since it uses only one degree of freedom more than the CSF. The comparison between the fit of CSF and UNIDIFF to the data help to judge if there is any trend in the strength of association between origin and destination. If the Unidiff fits better then the CSF, I consider either a linear or curvilinear effect model. Otherwise I consider the CSF model as the best fit. Finally, I analyze the origin by destination by period by education tables for men and women, because I want to understand the mediating role played by education on the origin by destination association. For this I apply the constant social fluidity and the uniform difference models to test for origin by destination trends in the tables including education. For these four-way analyses I use the full interaction pattern of association between OE and ED, instead of the matrix effects presented in the previous sections. The CSF and UNIDIFF models are designed to test for trends on the strength of association, and can be used with any pattern of association (Xie, 1992). For the origin by destination analysis I use three specifications for the pattern of association. First, I use a pattern of full interaction that specifies a parameter for each cell in the mobility table crossing origin by destination class. Second, I use a core social fluidity model first proposed by Goldthorpe and Erickson (1993) for a seven by seven mobility table, and later designed for the six by six mobility table by Ishida, Muller, and Ridge (1995). And, finally, I use a variant of the core model specially designed for Brazil (Ribeiro, 2007) 10 . The core model hypothesizes that the association between class of origin and destination can be represented by four types of effects in any industrial society. These effects are statistically implemented by five matrices. The first effect, represented in matrix one, is inheritance. There are six inheritance parameters, one for each class, and they are all designed to capture the propensity for individuals to stay in their class of origin rather than move to any other class of destination. Each class has a different 10 Scalon (1999) also estimate topological models, but uses a different class schema.

－ 146 －

propensity of inheritance since the social processes by which a class position is passed from fathers to sons or daughters tend to vary in each class. The second effect is hierarchy that is captured by two matrices (HI1 and HI2) designed to divide the six classes into three hierarchical levels. The service or professional and managerial class (I+II) is at the top of the hierarchy, and the non-skilled working class (VIIa) in the bottom. All other classes are in the intermediary hierarchical level. The farming class (IVc+VIIb) has a special position since it is in the lower level on the origin distribution and on the intermediary level on the destination distribution. This asymmetry was proposed based on the idea that the farming sector has been transformed in the process of industrialization, that is, on the fathers’ generation farming tent to be mainly a subsistence activity, while on the destination it turned into a market-oriented form of production. The third effect is the positive affinity designed to capture the proximity among certain classes of origin and destination. Positive affinity A is designed to capture the probable exchange between the professional and managerial class (I+II) and the routine non-manual class (IIIab), that is, this affinity represents the link within what could be named a “white-collar bloc.” Positive affinity B is designed to capture different social processes making groups of classes closer to each other. First, it expresses the propensity for relative mobility between the two propertied classes of farmers (IVc+VIIb) and small proprietors (IVab), because they probably have the possibility of transferring capital from one to the other. Second, it also captures the relative easy of mobility between the two working classes (V+VI and VIIa) that form a “blue collar bloc”. Third, it represents the proximity between the class of professionals and administrators (I+II) and the class of small proprietors (IVab), since professionals can help their sons to open small business or can be owners of professional practices or large business. Finally, positive affinity B also captures the relative easy of mobility from the rural class (IVc+VIIb) to the unskilled working class (VIIa). In order to fit the core model to the Brazilian data two modifications are necessary and both match somehow the sociological and historical literature on Brazilian social structure (Fernandes, 1974; Silva and Hasenbalg, 1988). The first modification is into positive affinity B. The propensity for mobility from the rural classes (IVc+VIIb) to the unskilled working class (VIIa) is extended to the skilled working class (V+VI). Since the skilled working class was formed very recently because the transformation from a predominantly agricultural to an industrial society happened during the 1960’s and 1970’s there is a high probability that people with origins in the rural classes would find working positions within the working class considered as a bloc. It also means that the distance between the two working classes is not very accentuated in Brazil. This view is shared by many interpreters of the Brazilian society who emphasize that within the skilled and unskilled working class there is a division between modern (industry using high technology) and traditional (industry using low technology) sectors (Serra and Tavares, 1971; and Silva e Hasenbalg, 1988). In fact a study using more class differentiation than in this chapter indicates the relevance of this division for mobility analysis (Ribeiro, 2007). The second modification to the core model is an extra negative affinity effect designed to capture the distance between the rural classes (IVc+VIIb) and the white-collar classes (I+II and III). As I have emphasized in the description of the Brazilian context above, the rural classes in Brazil are specially deprived. The rural structure of the country has been characterized by high levels of inequality, and in

－ 147 －

addition the access to education in rural areas has been extremely difficult during most of the 20th century. Therefore, the chances of sons and daughters of rural workers to reach the white-collar classes are much lower than the chances of those persons with other social origins. The negative affinity effect is designed to capture this extra barrier that people with rural origins face in Brazil. Table 12 presents the fit of nine models to the mobility table crossing six classes of origin by six of destination by four survey years for Brazilian men who had 30 to 64 years old in each year. Since the surveys have very high number of cases, I am obliged to evaluate the fit of the models based on the Bic statistic. The first three models (M1, M2, and M3) specify a full interaction pattern of association. The comparison between model 1, the Constant Social Fluidity model with full interaction, and model 2, the UNIDIFF model with full interaction, favors the later since its Bic is more negative (for CSF the Bic=-626.8, and for Unidiff the Bic=-648.5). Therefore, I conclude that there is a trend of increasing fluidity from 1973 to 1996. The expansion of fluidity is steeper between 1973 and 1982, but continues in a lower pace until 1996. Model 3 constrains the increase in fluidity to two periods, that is, it specifies that the fluidity increases from 1973 to 1982, but remains unchanged from 1982 onward. Since the difference in the Bic statistics between model 2 and model 3 is less then -5 points I decide to select model 2 as the preferred one, and must conclude that fluidity increased most between 1973 and 1982 but continued to increase in a slower pace until 1996. It is easy to interpret the expansion in fluidity between 1973 and 1982, since this period was marked by the fastest economic growth in the Brazilian history. The period after 1982 was marked by economic stagnation, nevertheless the increase in fluidity, although in a lower pace, is still present. It is hard to interpret this decrease, but it is probably related to the fact that from 1982 to 1988 and 1996 there is a significant increase in the proportion of men with secondary education. Models 4, 5 and 6 use the core social fluidity pattern defined by Ishida, Muller and Ridge (1995) for the six by six mobility table, and use respectively a constant social fluidity or no trend design, a uniform difference (Unidiff) trend, and a linear unidiff trend. No one of these three models fit the data well. The next three models (M6, M7 and M8) specify the same type of hypothesis for the change across survey years, but use the two modifications to the core model discussed above, that is, expand the positive affinity (AF2B) to the outflow from the rural class (IVc+VIIb) to the skilled manual class (V+VI), and includes an extra negative affinity to account for the difficulty of mobility from the rural classes (IVc+VIIb) to the white collar classes (I+II and III). With these two changes the core fluidity model fits the data. As it happened with the full interaction models (M1, M2 and M3) the Unidiff model with the Brazilian core fluidity pattern (M9) fits better than the CSF with the same pattern (M8) – M9 has a Bic=-582.2 and M8 has a Bic=-558.1. Model 9 specify a curvilinear trend for the unidiff parameter such that there is a big decline from 1973 to 1982 and a minor decline between 1982 and 1988, then constant fluidity until 1996. Although this model (M9) with a curvilinear trend fits better than model 8, I must be careful interpreting it because it is tailored to fit the data. In any case, model 9 still suggests that there is a minor increase in fluidity from 1982 to 1988 a period of economic turmoil in Brazil. Therefore our preferred model is M8 (unidiff core fluidity Brazilian version) and model 2 (unidiff with a full interaction pattern). Model 2 fits better then model 8, but it is not parsimonious and not theoretically informative. The parameters of model 8 are interesting because they allow the comparison with other late industrializing countries in

－ 148 －

this volume and industrialized countries in the paper by Ishida, Muller and Ridge (1995). All the parameters in model 8 follow the hypothesized pattern, that is to say they have negative and positive effects that match the ideas claimed by the formulators of the core model. In the Brazilian case there are, however, some particularities that should be highlighted. Two points seems to be different between Brazil and the industrialized world. First, immobility in the professional and managerial class (I+II) is significantly lower in Brazil than in the industrial nations (Ishida, Muller, and Ridge, 1995), and second long distance mobility seems to be more difficult in Brazil since I had to fit an extra negative affinity between the rural classes (IVc+VIIb) and the “white collar bloc” (I+II and III). 11 This negative affinity works as an extra hierarchy separating the rural classes form the white-collar classes, while the low value for inheritance in class I+II indicates low hierarchy at the top. Apparently this pattern is also present in Chile and Mexico, the two other Latin American nations in this volume. In table 13 I present the analyses for trends in OD association in the origin by destination by education by period table. The main goal is to measure the impact of controlling for education on the OD association represented by the Brazilian version of the core fluidity model. The analyses in table 13 confirm those in table 12, since the models for the uniform difference for OD association across period also fit better to the four-way table. Models 2, 4 and 6 (all hypothesizing change across period in the OD association) fit better than the respective constant fluidity models (1, 3, and 5). The last column in table 13 present the percentage reduction in the matrix effect parameters for the Brazilian version of the core fluidity model fit to the table without and with education. In other words it presents the comparison between the effect matrices in model 8 of table 12 and model 6 in table 13. The inclusion of education reduces significantly the effect of some parameter estimates and has minor effect on others. The clearest effect is that the propensity for inheritance in the service class (I+II) is eliminated once education is taken into account. This result reinforces the conclusion based on the analysis of ED association controlling for O (tables 8 and 9). That is, since the ED association does not change significantly when I accounted for origin and that the effect of origin on destination is eliminated after I control for education, I must conclude that after higher education is reached the entrance in the service class is not determined by class of origin. Inequality for entering the service class is therefore defined by the link between class of origin and educational attainment. As I have indicated (table 6) men with service class origin have considerable advantage in avoiding lower education and being included in higher education. In any one of the industrialized country studied by Ishida, Muller and Ridge (1995) the impact of education on reducing inheritance in the service class is as high as in Brazil. (WHAT ABOUT THIS VOLUME?).

The reduction in the hierarchy effects (HI1 and HI2) and on the negative affinity between rural classes and white-collar classes is also very significant in Brazil. The patterns of diminishment in these matrix effect parameters indicate that education is a powerful mean for eliminating class advantages and disadvantages. Education also diminishes significantly the effect of affinity A, that is, the affinity for exchange from class I+II and class IIIab reinforcing once again that education is relevant for diminishing

11 Changing the hierarchical position of rural class in the HI2 matrix does not have an effect on model fit. Therefore, I did not make this change and prefer to use the extra negative affinity.

－ 149 －

class inequality. In contrast, the inclusion of education does not have a very significant impact on

the inheritance in the classes below the service class (represented by diagonal effects 2, 3, 4, 5, and 6), on the one hand, and have a moderate impact (29.9% reduction) on affinity B involving the petty bourgeoisie, the farming classes, and the two working classes, on the other hand. Since this affinity captures very different social processes it is hard do say which linkages are reduced by controlling for education. Table 14 presents the fit of the same nine types of models discussed above to the mobility table for women crossing six origin classes by six destination classes by four survey years. The patterns of association are the same as those used to analyze the table for men, that is, models 1, 2 and 3 use a full interaction pattern; models 4, 5, and 6 use the core social fluidity pattern; and models 7, 8 and 9 use the Brazilian variant to the core fluidity model described above. Only the models with full interaction and the Brazilian core fluidity fit the data (models 1, 2, 3, 7, 8 and 9). The comparisons among these models indicate that the Unidiff specification is preferred (models 2, 3, 8 and 9) over the constant fluidity one (models 1 and 7). In other words, there is a trend of increasing fluidity for women. This trend, however, is different than the one observed for men. Models 2 and 8 (fitting better then 1 and 7 respectively) indicate that there is a clear increase in fluidity from 1973 to 1982, then constant fluidity between 1982 and 1988, and finally another increase in fluidity from 1988 to 1996. The linear specifications of the unidiff model either for the full interaction pattern (model 3) or the Brazilian core fluidity pattern (model 9) fit better, according to Bic, than the respective simple unidiff models with the same patterns of association (models 2 and 8). Therefore, the preferred models are model 3 and 9. Since model 9 is more parsimonious than model 3, and also more interesting for the comparative purposes of this volume, I will interpret the results of model 9 as the preferred one. The trend represented by model 9 (and also by model 3) indicates two moments of increasing fluidity. The first, from 1973 to 1982, is exactly the period of fastest economic growth in Brazil. During these years the country really changed and the increase in fluidity is almost certainly related to this economic expansion. The second moment of increasing fluidity, from 1988 to 1996, is probably related to educational expansion, and the increasing entrance of women in higher class positions (I+II and IIIab). The effect matrix parameters for model 9 follow the predicted expectations in terms of the signs indicating positive or negative effects. Since I do not have the results for women in other countries, I am not able to make comparisons across countries. I am, however, able to make some comparisons between men and women in Brazil, since there are some interesting differences. Immobility in the clerical class (IIIab) is higher for men than for women, this is probably a consequence of the fact that many women in this class destination have origins in other classes. The clerical class (IIIab) is a very common destination for women from different class origins. Also the immobility for women in the proprietor class (IVa) is lower than for men, indicating that males have probably higher chances of inheriting this positions. Another important difference is related to immobility in the working classes. Women have a higher tendency than man to be immobile in the unskilled working class, what is probably related to the fact that women from the lower classes tend to work in domestic labor (maids, nannies and etc). In contrast, men with origins in the skilled working class (V+VI) tend to remain there and women have higher propensities for moving to the clerical class (IIIab). The other parameters for women are similar from those for men.

－ 150 －

Table 15 presents the analysis for women including education as control variable. In other words, it presents the results of the exercise of model fitting to the origin by destination by period by education table for women. My interest is on observing the impact of including education on the matrix effect parameters modeling the origin by destination association across period. The last column of table 15 presents the percentage reduction on these parameters estimated by model 8 from table 14 and model 6 from table 15. Some interesting conclusions are obtained from this comparison. After controlling for education the effect of inheritance in the service class diminishes dramatically (190%). In fact, the sign is reverted indicating that education is really more important than having an origin in the service class to enter in this class position. The hierarchy effects also diminish considerably, since the effect of HI1 does not change much, but the effect of HI2 diminishes by 71%. The special negative affinity for the outflow from the rural classes to the white-collar classes (AF2C) also diminishes its effect by half. These changes indicate that once education is attained it becomes an important way of reducing class disadvantage for women in Brazil. Inheritance in the working classes (V+VI and VIIa) and the clerical class (IIIab) also diminish considerably with their effects becoming almost zero (check for SE).

Education also diminishes significantly the effect of the affinity linking class I+II and class IIIab, and reduces by half the effect of affinity B involving the petty bourgeoisie, the farming classes, and the two working classes. As I have mentioned above this affinity captures very different social processes and it is hard to say for which linkages education has a more significant effect.

8. Explaining trends in social fluidity In the previous sections I have shown that from 1973 to 1996 there was a significant decline in the association between origin and education for men but not for women. I also have shown that for both men and women there was an increase in social fluidity, as well as a decline in the association between education and destination. In this section I will present some models fit to the PEOD tables for men and women in order to judge in what measure the increase in fluidity are due to a decline in the association of origin and destination, and/or the decline of the association between education and destination. This analyses help to explain in what measure the decline in fluidity is due to economic expansion until 1982, and/or to educational expansion and other factors. Table 16 presents these models. The first is the independence model, the second is the constant OD association across period, the third is the constant ED association across period, the fourth is the constant OD and ED association, the fifth estimates a uniform difference (Unidiff) trend for OD across P, and the sixth a unidiff trend for ED across P. The models are fit to both the table for men and women. Models 1, 2 and 3 do not fit the data for men neither for women. Model 4, the constant fluidity across period for both OD and ED, is the first one to fit the data for men and women. The two following models also fit the data, but the model estimating a trend for ED across period (model 8) fits better than the model estimating a trend for OD across period (model 5). Finally, for men I fit model 7 that allows the OD association to vary across period and educational level. A similar model did not fit the table for women, and is not presented. The interpretations for men and women are different. For women the explanation for the increase in social fluidity is probably related to the fact that there was a

－ 151 －

considerable expansion of women participation in the labor market (37.4% in 1973, 37.8% in 1982, 44.8% in 1988, and 55.9% in 1996), as well as an increase in social fluidity due to the economic expansion between 1973 and 1982. As I have shown in section 5 the association between class of origin and educational qualification remained constant from 1973 to 1996 for women, what means that an equalization in inequality of educational opportunity cannot be claimed as an explanation for the increase in OD fluidity, which was fastest for women between 1973 and 1982 and then between 1988 and 1996. From 1973 to 1982 the economic expansion is probably the cause for the increase in fluidity, while from 1988 to 1996 this increase is probably due to the expansion of women participation in the labor market. A plausible explanation is that women entering in the labor market during the 1980’s and 1990’s were more likely to enter class positions in the service classes. In fact, table 2 indicates that in 1988 and 1990 the percentage of women in the service and clerical class is significantly higher than in the previous years. Finally, model 6 in table 16 (the preferred one for women) indicates that there is a decline in the ED association across the years. Tables 10 and 11 in section 6 estimating the models for inclusion and exclusion mechanisms for the ED association, indicate that the propensity for highly educated women to enter in the service class is almost independent from class of origin but that the propensity for avoiding working class position depends in some measure from origin class. Since there is a significant increase in women with secondary and tertiary education entering in the labor market in 1988 and 1996 (table 5), and the link between education and destination declines in this same period, I am inclined to conclude that education is loosing its value for women across the years. Nevertheless, the decline in the origin by destination association related to the economic expansion between 1973 and 1982, and to the entry of women in higher class positions in 1988 and 1996 accounts for the increase in fluidity observed for women in the labor market. For men the model that best fits the PEOD mobility table is model 7 (in table 16). This model indicates that social fluidity increases across both period and educational level. In general there is more fluidity in higher levels of education (secondary and tertiary), since there is a significant increase in educational attainment of secondary and tertiary education after 1982 (see table 5) the increase in fluidity during this period must be related to a compositional effect. 12 In other words, the fact that there are higher proportions of men in an educational level in which the effect of origin on destination is weaker probably accounts for the increase in fluidity specially during the 1980’s and 1990’s. Model 7 in this section also indicates that there is a significant decline in fluidity from 1973 to 1982 for those men with primary or less education and secondary education and then there is no significant change. This increase in fluidity is probably related to the fast economic expansion that characterizes Brazilian society during the 1970’s and came to halt in 1982. Finally, the increase in fluidity for men is probably also related to the diminishment of the effect of origin on educational qualification across the years.

9. Conclusion and summary In this chapter I have used a six-fold class schema and a three-fold educational classification to study trends in social mobility for men and women in Brazil during 1973 12 For other countries see Hout (1988) and Vallet (1999, and 2004).

－ 152 －

and 1996. I also used a class structure perspective and estimated the core social fluidity model (Ishida, Muller, and Ridge, 1995) and its Brazilian variant to analyze the pattern of origin and destination class association. For the origin to education, and the education to destination relations I used models designed to capture inclusion and exclusion mechanisms. These analyses help to determine trends in absolute and relative mobility rates that can be read in a comparative way with other country studied in this volume. In Brazil there are two marked economic periods during the second half of the 20th century. First, in the 1960’s and 1970’s the country experienced high rates of economic growth largely determined by industrialization. During this period the country’s social structure was transformed from rural and agricultural, to predominantly urban and industrial. In the beginning of the 1980’s the country entered in a period of economic stagnation and inflationary crises that lasted until the mid 1990’s. Trends in absolute and relative mobility rates in Brazil were affected from these two periods, since considerable increase in absolute rates happen between 1973 and 1982, as well as an increase in social fluidity. Absolute mobility rates in Brazil are marked by the very large discrepancy between origin and destination class distributions. Among the countries in this book only South Korea has comparable dissimilarity, due also to a rapid and recent transition from a rural to an urban social structure. In Brazil the economic expansion was accentuated until 1982. In fact, total mobility increased from 1973 to 1982 and remained constant afterwards for both men and women. 13 Outflow rates from the rural classes also increased from 1973 to 1982 (for both men and women). As it happen with other countries in this volume (???) immobility in the service class is lower in Brazil than in industrialized nations, and also declined most rapidly from 1973 to 1982 (for both men and women). Immobility in the working classes (V+VI and VIIa) also declined significantly from 1973 to 1982 for men and to 1988 for women. From inflow rates analyses I would like to highlight two points. First, that in Brazil a large proportion (about 1/3) of men and women in the service class have origins in the rural classes. This combined with the fact that for men at least two out of five members in the service class come from working class positions, indicates that there is no social closure (in an absolute mobility sense) in Brazil. Second, the inflow data indicates that self recruitment in the working class is not too large in Brazil, and also in other late industrializing countries, when compared to industrialized nations. The analyses of relative mobility rates also indicate change, especially, in the period of fast economic expansion from 1973 to 1982, but also afterwards. There is a significant increase in fluidity for both men and women. For men this increase can be explained by the effect of economic expansion and the absolute increase in higher educational provisions including more men in a level for which the origin and destination association is weaker. Therefore, the increase in fluidity should be attributed to both industrialization and compositional effects. The increase in fluidity for women is also explained by these two social processes, and also may be by the fact that the expanding proportion of women entering the labor market were doing so in higher class positions (I+II and IIIa). The fit of the Brazilian version of the core fluidity model also reveals some important aspects of social stratification in the country. First, the propensity for immobility in the service class in Brazil is considerably lower than in industrialized 13 This increase in total fluidity in a period of growth is also observed in Japan.

－ 153 －

nations. Second, the inclusion of an extra negative affinity for capturing the relative difficulty for the outflow from the rural classes to the white-collar bloc (I+II and IIIab), indicates the presence of an extra hierarchical level in Brazil. Long distance mobility seems to be more difficult in Brazil. This model applies well to men and women. Finally, the analysis on the role played by education in Brazil indicated that higher education is especially important in the country. Men with origin in the service class have considerably higher chances of attaining tertiary education and of avoiding lower educational levels. This is also valid for women, although they seem to be a little disadvantaged than men in Brazil. Higher education is extremely important to enter in the service class and to be excluded from the rural classes. For those having higher education the effect of social origin is almost inexistent.

References

Abreu, Marcelo Paiva. 1990. A Ordem do Progresso: 100 anos de política econômica republicana (1889-1989). Rio de Janeiro: Campus.

Abreu, Marcelo Paiva and Dorte Verner. 1997. Long-term Brazilian economic growth: 1930-34. Paris: Development Center of the Organization for Economic Co-operation and Development.

Adelman, Jeromy. 2001. "Institutions, property and Economic Development in Latin America." in The Other Mirror: Grand Theory through the Lens of Latin America, edited by M. A. Centeno and F. López-Alves. Princeton: Princeton University Press.

Albuquerque, Roberto Cavalcanti de and Renato Villela. 1993. O Brasil Social. Rio de Janeiro: IPEA.

Bacha, Edmar L. and Herbert S. Klein. 1989. Social Change in Brazil, 1945-1985: the Incomplete Transition. Albuquerque: university of New Mexico Press.

Bacha, Edmar L. and Lance Taylor. 1976. "The Unequalizing Spiral: A First Growth Model for Belindia." Quarterly Journal of Economics 90:197-218.

Barros, Ricardo Paes de, Ricardo Henriques, and Rosane Mendonça.2000. "Desigualdade e Pobreza no Brazil: retrato de uma estabilidade inaceitável." Revista Brasileira de Ciências Sociais 15:42-57.

Bresser-Pereira, L. 2003. Desenvolvimento e Crise no Brasil: História, Economia e Política de Getúlio Vargas a Lula. (Development and Crisis in Brazil: History, Economy, and Politics from Getulio Vargas to Lula) São Paulo: Editora 34. Becker, Gary S. [1964] 1993. Human Capital: A Theoretical and Empirical Analysis, with Special Reference to Education. Chicago: The University of Chicago Press.

Bendix, Reinhard, and Lipset, Seymour Martin. 1959. Social Mobility in Industrial Society. Berkeley: University of California Press.

－ 154 －

Bills, David, A. Haller, J. Kelly, M. Olson, and J. Pastore. 1985. "Class, class origins, regional socioeconomic development and the status attainment of Brazilian men." Pp. 89-127 in Research in Social Stratification and Mobility 4, edited by R.V.Robinson. Greenwich, Connecticut: JAI Press.

Bills, David and Archibal O. Haller. 1985. "Socioeconomic Development and Social Stratification: reassenssing the Brazilian case." Journal of Development Areas 19:59-69.

Blau, Peter Michael and Otis Dudley Duncan. 1967. The American Occupational Structure. New York: Wiley. Bourdieu, Pierre and Jean Claude Passeron. 1977. Reproduction in education, society and culture. London; Beverly Hills: Sage Publications.

Breen, Richard (ed.). 2004. Social Mobility in Europe. Oxford: Oxford University Press. Cardoso, A. 2002. “Direito do Trabalho e Relações de Classe no Brasil.” (The Right to Work and Class Relations in Brazil) in A Democracia e os Tres Poderes no Brasil, edited by Vianna, L.W. Belo Horizonte: Editora UFMG. Carvalho, José Alberto Magno and Laura Rodriguez Wong. 1997. "The Fertility Transition in Brazil: Causes and Consequences." in The Fertility Transition in Latin America, edited by J. M. Guzman. Oxford: Clarendon Press.

Castro, Claudio de Moura. 1986. "O que está acontecendo com a educação no Brasil?" in A transição incompleta: Brasil desde 1945, edited by E. e. K. Bacha, Herbert. Rio de Janeiro: Paz e Terra.

Erickson, Robert and John H. Goldthorpe. 1993. The Constant Flux: A Study of Class Mobility in Industrial Societies. Oxford: Clarendon Press.

Erickson, Robert, and John H. Goldthorpe. 1992. "The CASMIN Project and the American Dream." European Sociological Review 8:283-306.

Faria, Vilmar E. 1989. "Changes in the Composition of Employment and the Structure of Occupations." in Social Change in Brazil 1945-1985: The Incomplete Transition, edited by E. Bacha and H. Klein: University of New Mexico Press.

Fernandes, Florestan. 1974. A Revolução Burguesa no Brasil: Ensaio de Interpretação Sociológica. Rio de Janeiro: Zahar Editores. Gerber, T. and M. Hout. 2004. "Tightening Up: Declining Class Mobility during Russia's Market Transition." American Sociological Review 69:677-703.

Haller, Archibal O. 1991. "Ascription and Status Transmission in Brazil." Pp. 63-93 in Status Influence in Third World Labor Markets: Caste, Gender, and Custom, edited by J. Scoville. Berlin: Gruyter Inc.

Haller, Archibal O., Bam Dev Sharda, and George Miller. 1997. New Light on National

－ 155 －

Development: a challenge for stratification theory: Mimeo.

Haller, Archibal O., R. Torrecella, Maria Christina Haller, and Manuel Tourinho. 1996. "The Socioeconomic Development Level of the People of Amazoniam Brazil - 1970 and 1980." Journal of Development Areas 30:243-316.

Haller, Archibal O., Manuel Tourinho, David Bills, and Jose Pastore 1981. "Migration and Socioeconomic Status in Brazil: interregional and rural-urban variation in education, occupational status, and income." Luso-Brazilian Review 18:117-138. Hasenbalg, Silva and Nelson Valle. 1988. Estrutura Social, Mobilidade e Raça. Rio de Janeiro: IUPERJ/Vértice.

Hout, Michael. 1988. "More Universalism, Less Structural Mobility: The American occupational Structure in the 1980´s." American Journal of Sociology 93:1358-1400.

Hout, Michael and Robert Hauser. 1992. "Symmetry and Hierarchy in Social Mobility: A Methodological Analysis of the CASMIN Model of Class Mobility." European Sociological Review 8:239-266.

Hutchinson, Bertram. 1957. "The Social Grading of occupations in Brazil." British Journal of Sociology 3:176-189.

— 1958. "Structural and Exchange Mobility in the Assimilation of Immigrants to Brazil." Population Studies 12:111-120.

—. 1960. Mobilidade e Trabalho. São Paulo: Centro de Pesquisas Educacionais.

Ishida, Hiroshi. 1993. Social Mobility in Contemporary Japan: Educational Credentials, Class and the Labor Market in a Cross-National Perspective. Stanford: Stanford University Press.

Ishida, Hiroshi, Walter Muller, and John M. Ridge. 1995. "Class Origin, Class Destination and Education: A Cross-National Study of Ten Industrial Nations." American Journal of Sociology 101:145-93.

Kerr, C., J. Dunlop, F.Harbison, and C.Myers. 1960. Industrialism and Industrial Man. Cambridge, Mass: Harvard University Press.

Leal, Victor Nunes. 1976. Coronelismo, Enxada e Voto. São Paulo: Alfa-Omega.

Lopes, Juarez Brandão. 1971. Sociedade Industrial no Brasil. São Paulo: Difusão.

Park, H. 2002. "Social Mobility in South Korea." Research in Social Stratification and Mobility.

Pastore, José. 1981. Inequality and Social Mobility in Brazil. Madison: University of Wisconsin Press.

Pastore, Jose and Nelson Valle Silva. 2000. Mobilidade Social no Brasil. São Paulo:

－ 156 －

Markon.

Powers, Daniel and Yu Xie. 2000. Statistical Methods for Categorical Data Analysis. NY: Academic Press. Raftery, Adrian. 1986. "Choosing Models for Cross-Classification." American Sociological Review 51(1) : 145-46.

—. 1995. "Bayesian model Selection in Social Research." in Sociological Methodology, edited by P. Marsden. Washington D.C: The American Sociological Association.

Raftery, Adrian and Michael Hout. 1993. "Maximally Maintained Inequality: Expansion, Reform and Opportunity in Irish Education 1921-1975." Sociology of Education 66-1: 41-62. Ribeiro, Sérgio Costa and Rubem Klein. 1991. "O Censo Educacional e o Modelo de Fluxo." Relatório de Pesquisa e Desenvolvimento 24, Laboratório Nacional de Computação Científica - LNCC.

Ribeiro, Carlos Antonio Costa and Maria Celi Scalon. 2001. "Mobilidade de Classes no Brasil em Perspectiva Comparada." Dados: Revista De Ciências Sociais 44:53-96.

Ribeiro, Carlos Antonio Costa. 2003 "Estrutura de Classes, Condições de Vida e Oportunidades de Mobilidade Social." in Origens e Destinos: Desigualdades Sociais ao Longo da vida, edited by N. d. V. S. e. C. Hasenbalg. Rio de Janeiro: Topbooks.

—. 2005. "Classe e Gênero no Brasil Contemporâneo: Mobilidade Social, Casamento, e Divisão do Trabalho Doméstico." edited by C. e. A. Scalon, Clara. Rio de Janeiro: FGV Ed.

—. 2006. "Raça, Classe e Mobilidade Social no Brasil." Dados, Revista de Ciências Sociais 49.

—. 2007. Estrutura de Classe e Mobilidade Social no Brasil. São Paulo: EDUSC.

—. 2007b. “Educational Expansion and Inequality of Educational Opportunity in Brazil.” Paper presented at the RC28 meeting in Brno.

Scalon, Maria Celi. 1999. Mobilidade Social no Brasil: Padrões e Tendências. Rio de Janeiro: Revan-Ipuerj-UCM.

Schawartzman, Simon. 1982. Bases do Autoritarismo Brasileiro. Rio de Janeiro:Campus. Shavit, Y. and H.P. Blossfeld. 1993. Persistent Inequality.ChangingEducational Attainmet in Thirteen Countries. Boulder CO: Westview.

Silva, Nelson do Valle. 1979. "As duas faces da mobilidade." Dados: Revista de Ciências Sociais 21.

Silva, Nelson Valle. 1988. Uma Classificação ocupacional para o estudo da mobilidade

－ 157 －

e da situação de trabalho no Brasil: Mimeo.

—. 1999. "Mobilidade Social." in Sociologia 2 - O que ler na ciência social brasileira (1970-1995). São Paulo: Editora Sumaré.

Silva, Nelson Valle and Debora Roditi. 1988. "Et plus ça change.Tendências Históricas de fluidez social no Brasil." Dados 29. Tavares, Maria da Conceição e Serra, José. 1971. "Mas Allá del Estancamiento: Una Discusión sobre el Estilo de Desarrollo Reciente." El Trimestre Económico 33.

Torche, Florencia. 2005. "Unequal but Fluid: Social Mobility in Chile in Comparative Perspective." Americans Sociological Review 70:451-471.

Treiman, Donald J. 1970. "Industrialization and Social Stratification." in Social Stratification: research and Theory for the 1970´s, edited by E. O. Laumannn. Indianapoles: Bobbs-Merril.

Vallet, L. - A. 1999. "Quarent années de mobilité de mobilité sociale en France." Revue Française de Sociologie 40:3-64. Xie, Yu 1992. "The Long-Multiplicative Layer Effect Model for Comparing Mobility Tables." American Sociological Review 16:159-183.

－ 158 －

Figure 1GNP Per Capta in US$ (2003 prices): Brazil, 1940-2003.

0

500

1000

1500

2000

2500

3000

3500

1940

1942

1944

1946

1948

1950

1952

1954

1956

1958

1960

1962

1964

1966

1968

1970

1972

1974

1976

1978

1980

1982

1984

1986

1988

1990

1992

1994

1996

1998

2000

2002

Years

US$

－ 159 －

－ 160 －

Figure 2, Gross Elementary, Middle School, Primary, Secondary and Tertiary Enrollment Rates, Brazil 1960 to 2000.

0.0

10.0

20.0

30.0

40.0

50.0

60.0

70.0

80.0

90.0

100.0

1960 1965 1971 1975 1980 1985 1990 1995 2000

Elementary (4 years)Middle school (7 years)Primary (8 years)Secondary (3 years)Tertiary

Table 1 - Distribution of Origin and Destination by Birth Cohort and by Period (Survey Year), Men between 30 and 64 years old.

1909-17 1918-23 1924-31 1932-37 1938-43 1944-52 1953-58 1959-66

Father's ClassI+II 3.7% 3.5% 4.0% 4.0% 4.4% 5.4% 7.1% 8.5%IIIab 1.3% 1.3% 1.8% 1.8% 2.1% 2.7% 3.6% 4.6%IVab 5.7% 5.7% 5.9% 6.2% 6.4% 7.1% 6.9% 6.6%IVc+VIIb 80.0% 79.8% 75.8% 73.5% 70.4% 65.2% 57.7% 50.4%V+VI 5.2% 4.6% 5.5% 6.4% 7.0% 7.6% 9.4% 10.6%VIIb 4.2% 5.3% 7.1% 8.1% 9.6% 11.9% 15.3% 19.3%

Respondent's ClassI+II 7.5% 8.5% 10.0% 11.1% 12.3% 15.0% 16.3% 16.0%IIIab 4.7% 4.8% 5.1% 5.9% 6.4% 7.3% 8.5% 10.4%IVab 7.9% 7.8% 9.0% 9.1% 8.9% 8.9% 9.3% 8.3%IVc+VIIb 51.2% 47.8% 40.0% 35.6% 32.5% 27.0% 23.3% 21.1%V+VI 10.4% 9.3% 11.2% 12.7% 13.8% 14.6% 15.6% 14.5%VIIb 18.2% 21.9% 24.7% 25.6% 26.1% 27.1% 26.9% 29.6%

Dissimilarity 28.8% 32.0% 35.8% 38.0% 37.9% 38.2% 34.4% 29.2%Imobility 55.6% 52.7% 46.5% 42.5% 40.8% 37.3% 36.7% 37.0%

N 3854 7922 20238 22537 27775 39836 15884 11485

1973 1982 1988 1996

Father's ClassI+II 4.3% 4.3% 5.4% 6.7%IIIab 2.0% 2.1% 2.5% 3.5%IVab 6.5% 6.7% 6.7% 6.1%IVc+VIIb 72.2% 71.8% 66.2% 60.6%V+VI 6.6% 6.3% 7.7% 8.7%VIIb 8.4% 8.9% 11.5% 14.3%

Respondent's ClassI+II 11.6% 11.8% 14.1% 14.8%IIIab 6.1% 6.6% 6.9% 7.7%IVab 7.4% 8.7% 9.9% 9.5%IVc+VIIb 36.7% 34.1% 29.6% 26.7%V+VI 13.3% 13.2% 13.8% 13.5%VIIb 24.8% 25.6% 25.8% 27.8%

Dissimilarity 35.5% 37.7% 36.6% 33.9%Imobility 44.7% 42.0% 39.0% 38.8%Total Mobility 55.3% 58.0% 61.0% 61.2%

N 30742 53506 29966 35320

Birth Cohort Analysis

Period Analysis

0.0%

10.0%

20.0%

30.0%

40.0%

50.0%

60.0%

－ 161 －

－ 162 －

Table 2 - Distribution of Origin and Destination by Birth Cohort and by Period (Survey Year), Women between 30 and 64 years old.

1909-17 1918-23 1924-31 1932-37 1938-43 1944-52 1953-58 1959-66

Father's ClassI+II 3.3% 3.3% 4.2% 3.8% 4.9% 6.6% 7.6% 8.9%IIIab 0.9% 1.5% 1.4% 2.3% 2.4% 2.9% 4.1% 4.1%IVab 3.2% 5.2% 6.2% 6.3% 6.7% 7.6% 7.9% 7.5%IVc+VIIb 84.3% 78.9% 75.6% 72.9% 68.9% 63.5% 56.8% 53.1%V+VI 4.1% 4.4% 5.6% 6.1% 6.7% 7.5% 9.0% 10.7%VIIb 4.3% 6.8% 7.0% 8.7% 10.5% 11.9% 14.6% 15.8%

Respondent's ClassI+II 5.5% 6.3% 8.5% 9.5% 12.8% 17.5% 19.7% 18.3%IIIab 4.2% 7.1% 8.3% 9.7% 10.6% 12.7% 16.7% 17.4%IVab 3.6% 4.7% 5.2% 5.4% 5.7% 5.7% 6.1% 6.6%IVc+VIIb 49.8% 37.5% 31.5% 31.4% 27.4% 19.9% 17.1% 17.4%V+VI 9.5% 13.3% 13.9% 13.6% 14.1% 14.5% 12.6% 11.6%VIIb 27.5% 31.1% 32.5% 30.4% 29.4% 29.8% 27.7% 28.7%

Birth Cohort Analysis

Dissimilarity 34.5% 41.9% 45.2% 42.5% 42.6% 45.4% 41.5% 36.5%Imobility 52.2% 43.2% 38.3% 38.5% 35.2% 29.7% 27.6% 30.1%

N 910 2043 6025 8556 11891 19385 9574 7783

1973 1982 1988 1996

Father's ClassI+II 4.3% 5.2% 6.1% 7.2%IIIab 2.1% 2.4% 2.8% 3.6%IVab 6.7% 7.1% 7.6% 6.9%IVc+VIIb 72.9% 68.5% 62.6% 61.0%V+VI 5.7% 6.7% 8.2% 8.4%VIIb 8.4% 10.1% 12.7% 13.0%

Respondent's ClassI+II 11.1% 13.6% 16.8% 16.0%IIIab 9.9% 11.4% 13.1% 13.8%IVab 2.8% 5.5% 6.5% 6.7%IVc+VIIb 35.4% 23.1% 18.4% 23.3%V+VI 12.8% 14.5% 14.2% 12.7%VIIb 28.0% 31.9% 31.1% 27.6%

Dissimilarity 41.4% 47.0% 45.2% 37.9%Imobility 42.5% 31.8% 28.8% 32.6%Total Mobility 57.5% 68.2% 71.2% 67.4%

N 9894 19980 13313 22975

Perido Analysis

－ 16

3 －

Tabl

e 3

- Sel

ecte

d O

utflo

w R

ates

, Bra

zilia

n M

en a

nd W

omen

30

to 6

4 ye

ars

old

in 1

973,

198

2, 1

988

and

1996

, and

oth

er

Late

Indu

stria

lizai

ng C

ount

ries.

Serv

ice

Cla

ssW

orki

ng C

lass

Serv

ice

Cla

ssW

orki

ng C

lass

Serv

ice

Cla

ssW

orki

ng C

lass

Men

(I+

II)(V

+VI a

nd V

IIa)

(I+II)

(V+V

I and

VIIa

)(I+

II)(V

+VI a

nd V

IIa)

Braz

il19

7355

.3%

19.3

%17

.7%

58.0

%4.

9%36

.6%

1982

50.7

%20

.5%

18.7

%55

.3%

5.3%

38.1

%19

8851

.8%

18.9

%20

.8%

52.5

%6.

3%39

.5%

1996

50.0

%22

.9%

19.0

%56

.5%

6.9%

39.6

%

Japa

n 19

7555

.2%

11.9

%23

.7%

49.7

%17

.8%

35.4

%C

hile

55.1

%18

.3%

12.7

%54

.5%

10.6

%48

.1%

Taiw

an45

.1%

19.3

%20

.8%

46.4

%11

.6%

33.9

%C

hina

26.5

%14

.6%

13.0

%29

.1%

3.3%

5.5%

Kore

a45

.3%

13.5

%12

.3%

42.7

%8.

5%35

.3%

Wom

enBr

azil

1973

46.2

%16

.4%

12.7

%58

.6%

5.2%

40.8

%19

8250

.9%

19.9

%16

.2%

55.0

%6.

8%48

.9%

1988

50.6

%17

.1%

19.0

%49

.4%

8.5%

50.5

%19

9648

.2%

19.3

%18

.8%

47.8

%8.

1%42

.4%

Out

flow

from

Cla

ss I+

II (S

ervi

ce C

lass

) to:

Out

flow

from

Cla

ss V

/VI+

VIIa

(Wor

king

Cla

ss) t

o:O

utflo

w fr

om C

lass

IVc/

VIIb

(Rur

al C

lass

) to:

－ 164 －

Table 4 - Selected Outflow Rates, Brazilian Men and Women 30 to 64 years old in 1973, 1982, 1988 and 1996, and other Late Industrializaing Countries.

Working Class Rural Classes Working Class Rural ClassesMen (V/VI+VIIb) (Ivc+VIIb) (V/VI+VIIb) (IVc+VIIb)Brazil

1973 22.9% 30.6% 22.8% 69.3%1982 24.2% 32.5% 21.7% 70.4%1988 28.6% 29.6% 25.6% 66.1%1996 29.6% 28.1% 31.6% 58.1%

Japan 1975 8.2% 31.0% 15.7% 54.7%Chile 22.4% 12.9% 44.8% 27.2%Taiwan 19.9% 22.8% 28.2% 42.4%China 30.9% 16.9% 49.1% 20.1%Korea 7.5% 36.4% 11.2% 65.9%

WomenBrazil

1973 16.1% 34.3% 20.2% 73.0%1982 19.9% 34.4% 19.8% 72.2%1988 23.7% 31.9% 22.8% 69.9%1996 25.1% 30.8% 25.4% 64.3%

Inflow into Class I+II(Service Class) from:

Inflow into Class V/VI+VIIa(Working Class) from:

T able 5 - Distribution of Men and Women (30 to 64 years by Survey Year,

197 198 198 199Less than 59.7 54.3 45.6 36.2Elementary or Primary 30.3 32.6 35.7 39.6Secondar 4.6 6.3 9.7 13.9Tertiar 5.4 6.8 9.0 10.3

Tota 100.0 100.0 100.0 100.0

197 198 198 199Less than 64.3 53.1 42.7 34.4Elementary or Primary 28.0 29.3 32.3 36.9Secondar 5.6 9.4 13.0 15.6Tertiar 2.1 8.3 12.0 13.1

Tota 100.0 100.0 100.0 100.0

Men

Wome

－ 165 －

Table 6

fect matrix parameterEstimate SE Estimate

SO1 0.565 0.039 0.620SO2 -0.319 0.048 -0.321MO1 -0.627 0.031 -0.676MO2P 0.070 0.040 0.103MO2F 2.537 0.039 2.803MO2S 0.685 0.043 0.768MO2U 1.122 0.042 1.246

Estimates of OE Effect matrices parameter and Unidiff parameters OE over Period (1973, 1982, 1988 and 1996) in the Origin by Education by Period Table, Brazilian men 30 to 64 years old

M1 M2 M3 M4 M5Indep. CSF UNIDIFF CSF UNIDIFF

FI FI Matrix eff. Matrix eff.Goodness of Fit

L2 29913 184.5 116.1 192.1 123.5df 40 30 27 33 30p 0.00 0.00 0.00 0.00 0.00BIC 29437 -172.9 -205.6 -201.1 -233.9ID 0.136 0.008 0.006 0.008 0.006

Unidiff parameter:1973 1 11982 0.93 0.931988 0.92 0.921996 0.85 0.85

Ef

Table 7Estimates of OE Effect matrices parameter and Unidiff parameters OE over Period (1973, 1982, 1988 and 1996) in the Origin by Education by Period Table, Brazilian women 30 to 64 years old

M1 M2 M3 M4 M5Indep. CSF UNIDIFF CSF UNIDIFF

FI FI Matrix eff. Matrix eff.Goodness of Fit

L2 15095 140.6 105.8 164.7 129.4df 40 30 27 33 30p 0.00 0.00 0.00 0.00 0.00BIC 14651 -192.3 -194.5 -201.5 -203.6ID 0.164 0.011 0.009 0.012 0.010

Unidiff parameter:1973 1 11982 0.93 0.931988 0.90 0.901996 0.85 0.85

Effect matrix parameterEstimate SE Estimate

SO1 0.511 0.050 0.562SO2 -0.330 0.067 -0.355MO1 -0.635 0.040 -0.711MO2P -0.065 0.057 -0.051MO2F 2.385 0.054 2.664MO2S 0.702 0.060 0.790MO2U 1.160 0.058 1.300

－ 166 －

Table 8Estimates of ED Effect matrices parameter and Unidiff parameters OE over Period (1973, 1982, 1988 and 1996) in the Education by Destination by Period Table, Brazilian men 30 to 64 years old

M1 M2 M3 M4CSF UNIDIFF CSF UNIDIFFFI FI Matrix eff. Matrix eff.

Goodness of FitL2 399.8 143.6 553.9 298.5df 30 27 33 30p 0.00 0.00 0.00 0.00BIC 42.3 -178.1 160.7 -58.9ID 0.012 0.006 0.014 0.009

Unidiff parameter:1973 1 11982 0.93 0.931988 0.88 0.871996 0.77 0.76

Effect matrix parameterEstimate SE Estimate

HQ1 1.202 0.030 1.379HQ2 -1.295 0.058 -1.543LQ1 -0.651 0.030 -0.703LQ2P 0.528 0.029 0.607LQ2F 2.131 0.070 2.465LQ2S 1.635 0.036 1.928LQ2U 2.385 0.035 2.821

－ 167 －

－ 168 －

2.090 2.406 14.7%

Table 9Estimates of OD Effect matrices parameter and Unidiff parameters OD over Period (1973, 1982, 1988 and 1996) in the Origin by Destination by Education by Period Table, Brazilian women 30 to 64 years old

M1 M2 M3 M4 % reduc.CSF UNIDIFF CSF UNIDIFF M4 PEODFI FI Effect mat Effect mat M4 EDP

Goodness of FitL2 877.0 712.9 1016.797 858.6319df 230 227 233 230p 0.00 0.00 0.00 0.00BIC -1863.3 -1991.7 -1759.323 -1881.744ID 0.015 0.013 0.0164 0.0147

Unidiff parameter:1973 1 11982 0.96 0.951988 0.91 0.901996 0.79 0.79

Effect matrix parameterEstimate Estimate

HQ1 1.161 1.293 6.2%HQ2 -1.181 -1.376 10.8%LQ1 -0.613 -0.651 7.4%LQ2P 0.470 0.525 13.4%LQ2F 1.472 1.642 33.4%LQ2S 1.514 1.735 10.0%LQ2U

Table 10Estimates of ED Effect matrices parameter and Unidiff parameters OE over Period (1973, 1982, 1988 and 1996) in the Education by Destination by Period Table, Brazilian women 30 to 64 years old

M1 M2 M3 M4CSF UNIDIFF CSF UNIDIFFFI FI Matrix eff. Matrix eff.

Goodness of FitL2 416.2 124.0 431.7 132.1df 30 27 33 30p 0.00 0.00 0.00 0.00BIC 83.3 -175.7 65.5 -200.8ID 0.016 0.009 0.016 0.009

Unidiff parameter:1973 1 11982 0.85 0.841988 0.73 0.721996 0.65 0.64

Effect matrix parameterEstimate SE

HQ1 1.304 0.039 1.778HQ2 -1.004 0.081 -1.574LQ1 -1.261 0.040 -1.577LQ2P 0.991 0.044 1.376LQ2F 4.295 0.084 6.072LQ2S 2.220 0.048 3.071LQ2U 3.076 0.048 4.288

－ 169 －

－ 170 －

Table 11Estimates of OD Effect matrices parameter and Unidiff parameters OD over Period (1973, 1982, 1988 and 1996) in the Origin by Destination by Education by Period Table, Brazilian women 30 to 64 years old

M1 M2 M3 M4 % reduc.CSF UNIDIFF CSF UNIDIFF M4 PEODFI FI Effect mat Effect mat M4 EDP

Goodness of FitL2 730.1614 507.0062 745.3672 519.4109df 230 227 233 230p 0.00 0.00 0.00 0.00BIC -1822.299 -2012.162 -1840.387 -2033.05ID 0.0194 0.0157 0.0196 0.0159

Unidiff parameter:1973 1 11982 0.86 0.861988 0.74 0.731996 0.66 0.66

Effect matrix parameter

Estimate Estimate

HQ1 1.324 1.754 1.3%HQ2 -0.876 -1.331 15.4%LQ1 -1.333 -1.662 -5.4%LQ2P 1.009 1.363 1.0%LQ2F 3.656 5.064 16.6%LQ2S 2.092 2.833 7.8%LQ2U 2.804 3.835 10.6%

－ 17

1 －

Tabl

e 12

Estim

ates

of O

D Ef

fect

mat

rices

par

amet

er a

nd U

nidi

ff pa

ram

etO

D ov

er P

erio

d (1

973,

198

2, 1

988

and

1996

) in

the

Orig

in b

y Des

tinat

ion

by P

erio

d Ta

ble,

Br

azilia

n m

en 3

0 to

64

year

s ol

dM

1M

2M

3M

4M

5M

6M

7M

8M

9CS

F FI

UNID

IFF

UNID

IFF

CSF

UNID

IFF

UNID

IFF

CSF

UNID

IFF

UNID

IFF

FIFI

linea

rCo

reCo

reCo

re L

in.

Core

Br.

Core

Br.

Core

Br.

Lin.

Goo

dnes

s of

Fit

L226

6.8

209.

422

8.3

2201

.221

43.3

2152

.650

2.3

441.

945

9.36

86df

7572

7490

8789

8986

88p

0.00

0.00

0.00

0.00

0.00

0.00

0.00

0.00

0.00

BIC

-626

.8-6

48.5

-652

.211

28.8

1106

.611

01.5

-558

.1-5

82.8

-589

.2ID

0.01

0.01

0.01

0.03

0.03

0.03

350.

0159

0.01

490.

0154

Unid

iff p

aram

eter

:19

731

11

11

119

820.

950.

950.

950.

9569

0.95

0.96

1988

0.91

0.95

0.90

0.95

040.

900.

9519

960.

890.

950.

890.

9504

0.89

0.95

Effe

ct m

atrix

par

amet

erEs

timat

eSE

Estim

ate

SEEs

timat

eSE

Estim

ate

SEEs

timat

eSE

Estim

ate

SE

Diag

(I+II)

0.35

50.

037

0.39

40.

3756

0.41

90.

466

0.44

4Di

ag(II

Iab)

0.94

60.

046

1.02

30.

9811

0.65

70.

484

0.46

3Di

ag(IV

ab)

1.07

00.

029

1.14

01.

1007

1.21

11.

309

1.26

1Di

ag(IV

c+VI

Ib)

2.74

50.

023

2.93

32.

8248

3.25

52.

960

2.84

6Di

ag(V

+VI)

0.54

70.

025

0.58

20.

5623

0.85

20.

947

0.91

2Di

ag(V

IIa)

-0.0

790.

023

-0.0

80-0

.078

0.05

50.

191

0.18

3HI

1-0

.346

0.01

1-0

.369

-0.3

55-0

.444

-0.3

45-0

.331

HI2

-0.9

860.

021

-1.0

56-1

.015

-0.7

84-0

.617

-0.5

92AF

2A0.

631

0.02

90.

691

0.66

070.

600

0.43

90.

418

AF2B

0.26

40.

012

0.28

30.

2725

0.59

70.

473

0.45

5AF

2C-0

.542

-0.5

22

ers

－ 172 －

able 13Estimates of OD Effect matrices parameter and Unidiff parameters OD over Period (1973, 1982, 1988 and 1996) in the Origin by Destination by Education by Period Table, Brazilian men 30 to 64 years old

M1 M2 M3 M4 M5 M5 % reduc.CSF UNIDIFF CSF UNIDIFF CSF UNIDIFF M6 PEODFI FI Core Core Core Brazil Core Brazil M7 POD

Goodness of FitL2 922.5 841.3 1533.4 1476.1 1097.4 1024.7df 275 272 290 287 289 286p 0.00 0.00 0.00 0.00 0.00 0.00BIC -2354.0 -2399.4 -1921.8 -1943.3 -2345.9 -2382.9ID 0.019 0.018 0.026 0.026 0.020 0.020

Unidiff parameter:1973 1 1 11982 0.91 0.93 0.921988 0.83 0.84 0.831996 0.85 0.88 0.86

Effect matrix parameterEstimate Estimate Estimate Estimate

Diag(I+II) 0.016 0.030 -0.017 -0.004Diag(IIIab) 0.649 0.724 0.385 0.434Diag(IVab) 1.030 1.125 1.086 1.199Diag(IVc+VIIb) 2.464 2.700 2.491 2.759Diag(V+VI) 0.600 0.651 0.827 0.915Diag(VIIa) 0.089 0.100 0.142 0.162HI1 -0.245 -0.270 -0.248 -0.275HI2 -0.377 -0.417 -0.207 -0.228AF2A 0.343 0.394 0.198 0.236AF2B 0.233 0.257 0.297 0.331AF2C -0.301 -0.342

T

100.8%10.5%

8.4%6.8%3.5%

15.2%20.2%63.1%46.3%29.9%36.8%

－ 17

3 －

able

14

stim

ates

of O

D E

ffect

mat

rices

par

amet

er a

nd U

nidi

ff pa

ram

eter

s O

D o

ver P

erio

d (1

973,

198

2, 1

988

and

1996

) in

the

Orig

in b

y D

estin

atio

n by

Per

iod

Tabl

e,

Braz

ilian

wom

en 3

0 to

64

year

s ol

dM

1M

2M

3M

4M

5M

6M

7M

8M

9C

SF F

IU

NID

IFF

UN

IDIF

F C

SFU

NID

IFF

UN

IDIF

FC

SF

UN

IDIF

F U

NID

IFF

FIFI

line

arC

ore

Cor

eC

ore

Lin.

Cor

e Br

.C

ore

Br.

Cor

e Br

. Lin

.G

oodn

ess

of F

itL2

275.

315

4.7

167.

316

49.3

1532

.315

35.0

520.

439

2.8

407.

1df

7572

7490

8789

8986

88p

0.00

0.00

0.00

0.00

0.00

0.00

0.00

0.00

0.00

BIC

-557

.2-6

44.5

-652

.665

0.3

566.

656

4.1

-467

.5-5

61.8

-569

.6ID

0.01

90.

013

0.01

40.

049

0.04

70.

048

0.02

20.

018

0.01

8

Uni

diff

para

met

er:

11

11

1973

11

0.93

0.91

0.92

0.93

1982

0.92

0.93

0.91

0.91

0.93

0.93

1988

0.93

0.93

0.77

0.82

0.77

0.82

1996

0.77

0.82

Effe

ct m

atrix

par

amet

erEs

timat

eSE

Estim

ate

SEEs

timat

T E

eSE

Estim

ate

SEEs

timat

eSE

Estim

ate

SE

Dia

g(I+

II)0.

172

0.05

40.

216

0.20

40.

508

0.59

80.

573

Dia

g(III

ab)

0.63

20.

060

0.77

50.

747

0.09

40.

150

0.14

5D

iag(

IVab

)0.

697

0.05

10.

790

0.77

30.

691

0.77

40.

752

Dia

g(IV

c+VI

Ib)

2.95

10.

040

3.40

13.

265

3.43

03.

950

3.78

0D

iag(

V+VI

)-0

.175

0.04

2-0

.195

-0.1

890.

400

0.46

30.

445

Dia

g(VI

Ia)

-0.1

420.

035

-0.1

49-0

.146

0.39

60.

460

0.44

2H

I1-0

.469

0.01

5-0

.535

-0.5

19-0

.339

-0.3

90-0

.376

HI2

-0.9

750.

030

-1.1

18-1

.085

-0.4

44-0

.501

-0.4

86AF

2A0.

662

0.03

80.

790

0.76

10.

427

0.51

80.

497

AF2B

0.25

00.

018

0.28

70.

278

0.56

90.

652

0.62

7AF

2C-0

.546

-0.6

21-0

.599

Table 15Estimates of OD Effect matrices parameter and Unidiff parameters OD over Period (1973, 1982, 1988 and 1996) in the Origin by Destination by Education by Period Table, Brazilian women 30 to 64 years old

M1 M2 M3 M4 M5 M6 % reduc.CSF UNIDIFF CSF UNIDIFF CSF UNIDIFF M6 PEODFI FI Core Core Core Brazil Core Brazil M7 POD

Goodness of FitL2 714.9 573.8 1070.7 957.6 919.2 788.0df 275 272 290 287 289 286pBIC -2336.9 -2444.7 -2147.6 -2227.4 -2288.1 -2385.9ID 0.025 0.021 0.033 0.031 0.029 0.025

Unidiff parameter:1973 1 1 11982 0.84 0.86 0.851988 0.83 0.83 0.831996 0.63 0.66 0.64

Effect matrix parameterEstimate Estimate Estimate Estimate

Diag(I+II) -0.435 -0.533 -0.437 -0.537 189.9%Diag(IIIab) 0.233 0.385 0.001 0.077 48.4%Diag(IVab) 0.738 0.912 0.838 1.054 -36.1%Diag(IVc+VIIb) 2.574 3.232 2.616 3.323 15.9%Diag(V+VI) -0.168 -0.216 0.054 0.080 82.7%Diag(VIIa) -0.058 -0.069 0.057 0.092 80.1%HI1 -0.365 -0.462 -0.353 -0.448 -14.8%HI2 -0.243 -0.317 -0.113 -0.142 71.7%AF2A 0.184 0.286 0.053 0.111 78.6%AF2B 0.127 0.156 0.249 0.323 50.4%AF2C -0.195 -0.267 57.1%

－ 174 －

－ 175 －

Table 16. Goodness of fit statistics and unidiff parameter estimates for models fit to Period byEducation by Origin by Destination Tables for men and women. Brazil 1973, 1982, 1988 and 1996.

L2 df Bic1 Independence {OEC} {DC} 73948.1 340 69897.12.Constant {OD} association {OET} {DT} {OD} 29243.7 315 25490.53. Constant {ED} association {OET} {DT} {ED} 25057.2 330 21125.34. {OD} & {ED} association constant 1261.5 305 -2372.4

5. Only {OD} changes (UNIDIFF) 1167.4 302 -2430.8Unidiff parameters (1973) (1982) (1988) (1996)

1 0.91 0.82 0.836. Only {ED} changes (UNIDIFF) 1046.8 302 -2551.5

Unidiff parameters (1973) (1982) (1988) (1996)1 0.95 0.88 0.76

7.OD changes (UNIDIFF) acros E and P 975.0 299 -2587.5Unidiff parameters (1973) (1982) (1988) (1996)

E1 1 0.92 0.90 0.90E2 0.74 0.65 0.65 0.65E3 0.51 0.56 0.56 0.56

L2 df Bic1 Independence {OEC} {DC} 43203.7 340 39430.32.Constant {OD} association {OET} {DT} {OD} 24520.5 315 21024.53. Constant {ED} association {OET} {DT} {ED} 8460.3 330 4797.84. {OD} & {ED} association constant 1075.3 305 -2309.7

5. Only {OD} changes (UNIDIFF) 923.0 302 -2428.6Unidiff parameters (1973) (1982) (1988) (1996)

1 0.84 0.82 0.626. Only {ED} changes (UNIDIFF) 828.361 302 -2523.309

Unidiff parameters (1973) (1982) (1988) (1996)1 0.86 0.74 0.66

Women

Men

Social Mobility and Education in Contemporary Chile ∗

Florencia Torche

(New York University)

Abstract Using the 2001 Chilean Mobility Survey, this paper examines the pattern of intergenerational mobility among Chilean men. The pattern is found to be relatively similar to that of advanced industrial countries, as expressed in the core model of social fluidity. Departures from the core can be explained by historically specific characteristics of Chilean society -- an agrarian reform and counter-reform that disrupted the patterns of land inheritance, little financial capital associated with the voluminous Chilean self-employed sector, and the strong barriers preventing long-range downward mobility from the Chilean elite. No change over the past five decades in the level of fluidity is detected, in spite of significant economic development and change in the national political economy. The role of education in the mobility process is analyzed. Findings indicate that education is an important but not the only mediating role in the transmission of advantage and across generations. Education accounts for the intergenerational reproduction of the professional class, but other factors contribute to the strong hierarchical barriers to mobility in Chile. Keywords: intergenerational mobility, social class, inequality, industrialization.

1. Introduction Researchers have accumulated substantial knowledge about intergenerational class mobility in the industrialized world. Research has shown that the mobility pattern -- the location of the main intergenerational barriers in the social structure -- is relatively similar across countries, but the strength of these barriers displays more international variation (Erikson and Goldthorpe 1992, Hout and Di Prete 2006). We also know that international differences in mobility patterns are largely driven by nationally-specific institutional arrangements. As to mobility trends over time, a relative consensus about temporal constancy has been recently challenged, as national studies using repeated cross-sectional surveys have show growing fluidity in several industrialized countries (Breen 2004, Breen et al. 2005, see also Ganzeboom et al. 1989).

Evidence about macro- societal factors affecting mobility is less conclusive. Researchers have explored the influence of economic inequality, economic growth, educational enrollment and the political-economy regime on the rate of intergenerational fluidity. The most likely candidate to affect fluidity rates is the level of cross-sectional inequality, under the assumption that a more unequal distribution of resources will determine differential access to opportunity across generations. This hypothesis found ∗ This paper was presented at the meetings of the Project on Social Mobility in Late-industrializing Nations. I have benefited from comments from the participants of the meetings and the members of the Project on Social Mobility in Late-industrializing Nations.

－ 177 －

some empirical support (Tyree, Semyonov and Hodge 1979, Treiman and Yip 1989, Erikson and Goldthorpe 1992, chapter 12) but it has been questioned recently (Breen and Luijkx 2004a, Torche 2005a) 1 . Some empirical evidence supports the influence of the national political-economy regime on intergenerational fluidity (Goldthorpe 2007, chapter 7). For instance, intergenerational mobility was found to increase after the communist takeover in Hungary, as a result of drastic state intervention directly aimed at destratification (Wong and Hauser 1992, Szelenyi 1998). Growing fluidity is also found in Sweden between the 1930s and the 1970s in a context of social-democratic policies that implemented a comprehensive welfare state oriented to reduce class inequalities (Jonsson 1991). However, in these two national experiences fluidity has ceased to increase, and even slightly declined after the strong and systematic inequality-reduction policies have lost force (Jonsson 1996, Jonsson and Mills 1993, Roberts and Bukodi 2004). These recent trends suggest that even small increases in fluidity are not self-sustaining and they require strong and sustained intervention in the form of equalization policies.

Research suggests also that an important factor mediating intergenerational association is educational attainment. Three dimensions of the mediating role of education have been claimed to drive changes in social fluidity over time. These are changes in the association between social background and educational attainment, changes in the influence of educational attainment on class position, and a compositional effect driven by increase in the proportion of individuals attaining higher education combined with a lower intergenerational association of status among those with higher education (Hout 1988, Vallet 2004, Breen and Jonsson 2007, Breen and Luijkx 2007).

While in most national experiences, intergenerational fluidity has tended to increase or to remain constant, there is indication of growing rigidity in Russia (Gerber and Hout 2004) and Mexico (Torche and Costa-Ribeiro 2007). In both cases, the strengthening of the association between class origins and class destinations took place in a context of growing economic inequality and deep market reform. These recent findings suggest that there social fluidity is not as constant as we once thought and that it may be responsive -- even if only temporarily -- to institutional transformations and drastic changes in economic inequality. Part of the reason why the “constancy” hypothesis has not faced more empirical challenge may be that most countries analyzed to date are relatively homogeneous in terms of socioeconomic attributes -- they are mostly economically advanced nations with relatively developed welfare states. In this context, examining late-industrializing nations characterized by different levels and styles of development, degrees of inequality, and institutional arrangements may provide useful information on the sources of commonality and variation in intergenerational mobility.

Chile provides an interesting case in many respects. Historically very poor, Chile has experienced substantial economic growth in the last two decades, while it still remains one of the most unequal nations in the world. Furthermore, Chile implemented a deep and fast market reform during the 1970s and 1980s. This “great market transformation” (Martinez and Diaz 1999) has been used as a blueprint for reform in numerous countries over the last two decades. The main objective of this paper is to

1 Researchers tend to measure cross-sectional inequality contemporaneously rather than in the parental generation. However, the question about the timing and mechanisms whereby inequality of condition would potentially affect inequality of opportunity is still very much open.

－ 178 －

analyze the Chilean mobility pattern, the role of education in the mobility process, and mobility trends in the last few decades. The paper is organized in six sections. After this introduction, the second section provides basic information about the Chilean context. Section 3 introduces the data and variables used in this analysis. Section 4 describes absolute mobility patterns. Section 5 studies social fluidity -- the association between origins and destinations net of changes in the occupational structure -- its changes over time, and the role of education in social fluidity. Section 6 concludes.

2. The Chilean Context

By the 1950s Chile was one of the poorest countries in Latin America (Hofman 2000), based on an extractive economy organized around agriculture and mining. The country has experienced substantial economic and institutional transformation over the second half of the 20th century. Employment in agriculture and mining dropped from 37 to 15% between 1950 and 2000, while service employment jumped from 39% to 68% in the same period (see table 1). The economic transformation was coupled by migration from the countryside to the city, with the urbanization rate rising from 61% in 1950 to 86% in 2000. Even if post-war Chile was not as overwhelmingly rural and agricultural as other nations included in this volume such as Brazil, Mexico, China, or Korea, the transformation is nonetheless substantial.

The political economy regime has also significantly changed in Chile. The post-war economic landscape was defined by import-substitution industrialization (ISI) (Meller 1996). The ISI model was based on two types of policies: The closing of the national economy to world markets, and state-led industrial development (Munoz 1990, Stallings 1978). The growing industrial sector was, however, capital intensive and unable to provide employment for the large numbers of rural migrants (Corbo and Meller 1977). For example, between 1940 and 1970 the share of manufacture in total output expanded from 13 to 25% but its share in total employment expanded only from 17 to 18% (Braun et al. 2000). As a result, industrialization did not result in an expansion of the industrial working class. Therefore, the transformation of the Chilean class structure over the second half of the 20th century can be described as a leap from predominantly agricultural economy to one predominantly based on services, without an intermediate stage of industrial development.

In the early 1970s an aggressive market reform was implemented in the country by a military regime in power between 1973 and 1990. This market transformation was deeper and took place earlier than elsewhere in the world, forecasting similar processes taking place in the 1980s and 1990s in industrialized and developing countries. The reform included trade liberalization, deregulation of foreign direct investment and financial markets, and aggressive privatization of state-owned enterprises and of the social insurance system (Stalling and Peres 2000: 40-42, Edwards and Cox-Edwards 1991, Ffrench Davis 1999). After a deep economic recession and costly adjustment to the new market model in the late 1970s and early 1980s (Meller 1991) the country has experienced substantial economic growth and improvement of the living conditions of the population since the late 1980s (column 1 in table 1).

Another important characteristic of the Chilean case is the very high income inequality. With a Gini of .57 Chile currently ranks as the 11th most unequal country in

－ 179 －

the world (United Nations 2005). Inequality widened in the 1980s during the market reform and did not decline through the economic expansion of the 1990s. The Chilean pattern of inequality is characterized by high concentration at the top of the income distribution -- with the top income decile receiving more than 42% of the total national income -- but less disparities among the lower and middle classes. Indeed, if the wealthiest decile were excluded from the calculation, inequality would drop dramatically, and Chile would be more equal than the U.S. (Inter American Development Bank 1999). This pattern of inequality closely corresponds to the Chilean pattern of intergenerational and intragenerational mobility, characterized by high barriers to and from the wealthiest decile, largely populated by college-educated professionals, but significant fluidity among middle and lower classes (Nuñez and Risco 2004, Contreras et al. 2005, Torche 2005a, Beller and Hout 2006). These studies suggest that the hierarchical barriers to mobility are closely related to the structure of economic inequality prevalent in the country.

A class approach to mobility captures not only hierarchical but also non-hierarchical barriers to mobility, such as the propensity to remain in the class of origins (class inheritance) and barriers between sectors of the economy, such as those between agriculture and urban occupations, or between self-employment and salaried ones. Several characteristics of the Chilean society may affect these non-hierarchical barriers, inducing departures from the mobility pattern found in industrialized countries. Two need to be highlighted. First, an agrarian reform was implemented in the country during the 1960s and 1970s. Substantial redistribution of land was followed by a counter-reform undertaken by the military government in the 1970s. Rather than re-establishing the traditional hacienda model, the counter-reform created an active land market, and refocused agricultural production on the external market (Correa et al. 2001, Kay 2002). As a result, the pattern of land inheritance was drastically altered, which may have resulted in substantial intergenerational fluidity among the agricultural classes. Thus, Chilean farmers should display a particularly weak pattern of class inheritance.

Secondly, the proportion of self-employed reaches about 20% of the labor force. This is much higher than in the industrialized world, where the self-employment share is usually within the 1-digit range and (Arum and Muller 2004). In Chile (as elsewhere in Latin America) self-employment has been described as “forced entrepreneurship” as it includes a large share of low-capital, low-skill, relatively precarious and often unregulated activities (Portes and Hoffman 2002, Infante and Klein 1995, Portes, Castells and Benton 1989). Furthermore, the opening of the Chilean economy to international trade and the shrinking of the state resulting from the market reform in the 1970s-80s induced a decline in manufacturing and public employment, and to a transitory expansion of self-employment, used as a survival strategy by those who had lost their jobs in the manufacturing and public sectors (Leon and Martinez 2000, Velasquez 1990). Self-employment rose from 15% of the labor force in 1970 to 28% in 1980, to stabilize at about 20% in the 1990s (Thomas 1996). Most of these “forced entrepreneurs” lack material and financial resources that can be passed across generations. Thus, as the agricultural classes, the Chilean self-employed may display weak inheritance. This would further contribute to the weakness of sector barriers to mobility in Chile. Education in Chile. The Chilean educational system is organized into 8 years of compulsory primary education, four years of secondary schooling, and 2 to 6 years of

－ 180 －

post-secondary education, which can be pursued in vocational colleges, professional colleges or universities. In contrast to many European systems, which assign pupils to tracks early in their occupational careers, the Chilean system resembles the one in the U.S., organized as a linear accumulation of years of schooling, with very low stratification. Tracking exists only at the secondary level, when students select vocational or academic schools, but they can switch at any point of their secondary educational career, and both tracks have similar formal access to tertiary education.

Chile experienced substantial educational expansion throughout the second half of the 20th century. As columns 8 and 9 in table 1 indicate, gross secondary enrollment rose from 19 to 83 percent between 1950 and 2000, and tertiary enrollment rose from 1 to 38 percent in the same period. Most of the secondary expansion took place during the 1970s, and it slowed down during the economic crisis of the 1980s. The expansion of the post-secondary sector is more recent, and it accelerated in the 1990s. Even if smaller than the most impressive national experiences of post-secondary expansion such as Korea, the Chilean growth is nevertheless substantial, with about one-third of the relevant-age Chileans currently attending some form of post-secondary school.

This enormous expansion may result in a decline in the earnings returns to post-secondary education, as skilled labor becomes less scarce. However, by 2000, the college premium remained extremely high in Chile -- higher than other Latin American countries and the U.S. This high college premium accounts for most of the “excess earnings inequality” in Chile, in particular, for the concentration of income in the top decile (UNESCO 1999, Beyer 2000).

The expansion of lower levels of education was fostered by an educational reform in 1965, which extended compulsory primary schooling from 6 to 8 years. In 1981, in the context of the market transformation, a major privatization reform of the educational system was launched. This reform introduced a universal voucher system, in which a per-student subsidy was paid to public and private institutions in exchange for not charge tuition. The reform led to the creation of so-called “private-subsidized” schools funded by the government, which captured about one-third of the market by the early 1990s, at the expense of public school enrollment. It is important to highlight that the Chilean voucher system differs from the U.S. voucher experiments in that the government does not give a tuition certificate to the family but, rather, pays the subsidy directly to the school that the student chooses. Thus, the Chilean voucher model is known as a “funds follow the student” system (Mizala and Romaguera 2000).

The privatization of education pursued two objectives: Fostering competition among schools and providing choice to parents. Analysis of the consequences of the reform indicate significant sorting based on socioeconomic resources (and also unobserved characteristics), and a very minor changes on overall efficiency of the system, although this issue is highly contested (Mizala and Romaguera 2000, Carnoy and McEwan 2001, Sapelli 2003, Torche 2005b, Hsieh and Urquiola 2006).

Another component of the privatization reform was the expansion of higher education. This expansion was accomplished by the government’s promotion of private, fully paid post-secondary institutions, of widely diverse quality and academic standards (Brunner 1994; Nuñez 1997). The number of post-secondary institutions grew dramatically from 8 in 1980 to almost 300 in the early 90s. In addition, fees were implemented in formerly virtually-free public institutions.

As indicated, education plays an important role in intergenerational mobility. Changes in its influence could emerge from transformations of the association between

－ 181 －

origins and educational attainment, or the association between educational attainment and class of destination. The origin-education association has been found to decline in many countries (Breen et al. 2005). This is not the case in Chile. If something, this association may have increased in the recent past (Torche 2005b). As to the education-destination association, expansion of post-secondary enrollment may result in a decline in the college premium, and a consequent reduction of the influence of schooling on economic wellbeing. Preliminary evidence suggests that this may be the case among youngest cohorts of college graduates, born in the late 1970s (Sapelli 2007), but it is still too early to tell. A still unanswered question is the extent to which this potentially weakening influence of education on earnings translates to class position. With this background information about the Chilean case we move now to the analysis of mobility trends over the last few decades, and the role of education in the mobility process.

3. Data and Variables

This analysis uses the 2001 Chilean Mobility Survey (CMS). The CMS is a nationally representative multistage stratified sample of male heads of household aged 24-69 2 . The sampling strategy includes three stages. First, 87 primary sampling units (municipalities) were randomly selected from the 334 municipalities in the country. Then blocks within municipalities were selected, and finally households within blocks were chosen. Municipalities are stratified by size (fewer than 20,000, 20,000 to 100,000, 100,000 to 200,000 and more than 200,000 inhabitants) and by zone of the country (north, center, south). To increase efficiency, all municipalities in the large-size stratum are included. The fieldwork, conducted between April and June 2001, consists of face-to-face interviews in the respondents’ household carried out by trained personnel. The sample size is 3,544 and the response rate, after excluding non-eligible households, is 63%. The sample was weighted to correct for sample design and be able to generalize to the Chilean population.

In order to increase the likelihood that the respondents are in the labor force at the moment of the interview and to avoid recall problems, the age range was reduced to 30 to 64 for this analysis. For those respondents who were unemployed or otherwise not in the labor force at the time of the survey (about 12% of the sample) information was obtained about the last job they had. The key variables in this study are class or origin, class of destination and education. Class position is coded using the 6-fold version of the CASMIN class schema (Erikson and Goldthorpe 1992 for the original schema, Ishida et al. 1995 for the 6-class version). The following classes are distinguished: Service Class (I+II), Routine Non-Manual (III), Petty bourgeoisie (IVab), Agricultural (IVc+VIIb), Skilled Manual (V+VI) and Unskilled Manual (VIIa). Given that three- and four-way tables will be analyzed using this moderate sample size, sampling zeroes are likely to emerge. To avoid sampling zeros, 0.01 was added to each cell of the analyzed tables.

Education was operationalized by an ordinal variable that distinguishes lower, middle and higher education. Lower education refers to primary education or less, middle education includes incomplete and complete secondary schooling, and higher 2 Findings for intergenerational educational mobility (Torche 2007) suggest that there is no bias associated to reduction of sample to heads of household.

－ 182 －

schooling comprises any type of post-secondary schooling, including vocational, professional and academic colleges. Given that compulsory primary education was expanded from 6 to 8 years in 1965, the coding of educational attainment follows the educational levels rather than the years of schooling completed, under the assumption that the meaning of primary education has remained constant over time in spite of the expansion from 6 to 8 years.

4. Class Structure and Absolute Mobility

In order to provide a basic description of the Chilean class and educational structure, table 2 presents the origin and destination class distribution of Chilean men 30-64 years old. Even though the origin distribution is affected by relative fertility of men and the time in which the respondent was born, and therefore it does not represent any particular generation (Blau and Duncan 1967), the comparison of origin and destination distributions provide a coarse assessment of change over time. The most impressive transformation is the growth of the service class and the decline of the agricultural class. Indeed, very little change is observed beyond these two classes. This is entirely consistent with the flat trend in the share of manufacturing in total employment from 1950 to 2000 reported in table 1. Interestingly, the routine non-manual sector, quite small in the origin distribution, does not increase with the “tertiarization” of the Chilean economy, which is likely due to the fact that this sector involves a large percentage of female workers, excluded from this analysis. For instance, among the European countries compared by Breen and Luijkx (2004b: 74-5) the proportion of men in class III ranges between 7% and 18% across countries, while the proportion of women in this class ranges between 20% and 46%. The petty bourgeoisie is much larger in Chile than in the industrialized world, both in the origin and destination distribution. As discussed in the previous section, this highlights a particularity of Latin American (and developing-countries’ in general) labor markets, characterized by a voluminous informal sector (Infante and Klein 1995, Portes, Castells and Benton 1989).

The last column of table 1 presents the class distribution of respondent’s first job after completing his education, and it provides additional information on individual intergenerational careers. Not surprisingly, the lower echelons of both the non-manual and manual sectors -- classes III and VIIa -- are much more voluminous than the respective upper echelons (classes I and V+VI) in the class distribution of first job. This suggests that individuals enter the labor market in lower-level positions, to then experience intra-generational mobility within sectors. The very small percent of those who start working as self-employed (class IVab) suggests that independent work is a later step in the individual career, after individuals have participated in the labor market as salaried employees. Given that self-employment is not likely to be capital intensive in Chile, and that the large majority (about three-quarters) of self-employed do not hire the work of others, we assume that big sums of capital is not the main constraints preventing young Chileans to start a business, but factors such as social capital, access to a market niche, and experience might also play a substantive role.

The index of dissimilarity for the origin-current job table, measuring the proportion of cases in the table that would need to be reallocated to make origin and destination distributions identical, is 13.4. In comparative perspective this index is fairly

－ 183 －

low -- in most European countries, for instance, it reaches around 20 (Breen and Luijkx 2004b: 47).

In order to examine trends in intergenerational mobility, I divide the sample into three cohorts, born between 1937-1950 (C1), 1951-1960 (C2), and 1961-1971 (C3), which are aged 51-64, 41-50, and 30-40, respectively, at the time of the survey. The oldest cohort entered the labor market in the 1960s and 70s during the “import substitution industrialization” political economy regime. The second cohort entered the labor market during the 1970s and 80s, decades defined by deep market transformation and economic crisis. The youngest cohort entered the labor market during the period of economic stabilization and growth in the 1980s and 90s. It is well known that cohort analysis of a single cross-section confounds age, period and cohort effects (Ryder 1965). Age effects will be controlled by using two approaches. First, I restrict the sample to respondents aged 30 years old and older, under the assumption that they have reached occupational maturity, a stage in the life-cycle at which little further intragenerational mobility is expected. Note that while in the industrialized world occupational maturity is claimed to be reached around the age of 35 (Goldthorpe 1980, Heath and Payne 1999), I selected 30 years old under the assumption that this stage is reached at a younger age in Latin America due to lower overall educational attainment. In fact, the Chilean sample indicates that the mean age of labor market entry after finishing school is 18 years old, and it ranges between 17.3 in C1, 18.7 in C2, and 19.0 in C3. As a second strategy to control for age effects, I identify class of labor market entry as class of destination. Although there is variation across respondents in the chronological age at which they entered the labor market, this definition identifies a similar stage in the life-cycle for all respondents. Period and cohort effects cannot be disentangled in this analysis of a single cross-sectional survey, therefore both interpretations of change will be offered if a temporal trend is discovered.

Panel B in table 2 presents the change across cohorts in the class of origin, class of destination, and class of labor market entry distributions. Superficial examination of changes in the origins and destination distributions across cohorts confirms the transformation of the Chilean class structure driven by growth of the class of professionals and managers (classes I-II), and the decline of the agricultural class. The figures suggest also a stagnation of this structural change from the second (1951-60) to the youngest (1961-71) cohort, signaled by a very similar proportion of workers involved in the professional class in c2 and c3 (22.4% in C2 and 21.7% in C3) and in the agricultural class (10.6% in C2 and 11.6% in C3). However, this apparent stabilization can be partly a result of age effects if individuals move into the service class and out of agriculture rather late in their occupational career. To check for this possibility, cohort trends in class of first job are useful. Comparison of the proportion of workers who entered the labor market in the professional and agricultural classes from C2 (1951-61) to the youngest cohort (1962-76), provide further indication of stabilization: The proportion of those with first jobs in the professional class slightly declines from the second to the youngest cohort, and the proportion with first jobs in agriculture remains stable. These trends in the class distribution of first job are inconsistent with an age interpretation of change, and they support the interpretation of decline in structural mobility and stabilization of the Chilean class structure 3 .

3 Note also that the proportion of the origin distribution in the professional class substantially drops from 17 in C2 to 14 percent for the youngest cohort, accompanied by a comparable increase in

－ 184 －

Table 2B reports also the Index of Dissimilarity for two pairs of distributions: class of origin-current class position, and class of origin-class of first job. Both show a decline in intergenerational change in the class structure over time, further indication of reduced structural mobility.

In contrast to other national experiences, the amount of total mobility (i.e. the percent of individuals with destination different from their origins) is relatively high in Chile, reaching more than two-thirds of the population and remains roughly constant over time. Detailed analysis of the class-specific inheritance propensities suggests that the stability is driven by a compositional effect. The numerical decline of a class with a high level of inheritance (the agricultural class) is offset by the increase of the service class, which also has a very high inheritance level. Therefore, even though the occupational structure experiences significant upgrading, the level of inheritance remains constant over the second half of the 20th century, but is driven by the immobility propensity of a different class. Inflow and Outflow Distributions. Table 3 presents the outflow distribution across class of origin, and the inflow distribution across class of destination, and table 4 breaks these flows by cohort. The following findings from these distributions can be highlighted. First, the professional class has, by far, the highest inheritance rate (table 3A), with more than half of the sons of this class remaining in the top of the hierarchy, which is driven by the significant expansion of this sector across generations. However, cohort trends shown in table 4 reveal that class inheritance among professionals grows, even as the proportion of professionals in the total class structure flattens (note the stabilization in the size of the professional class from 22.7% in C2 to 21.7% in C3). This suggests persistent inheritance of class I-II, net of marginal changes in the class structure, an issue that will be further explored. Inflow patterns (table 3B) indicate that this expanding class recruits widely, with about one-third of its members having origins in the working and agricultural classes, and 20% coming from the self-employed sector. Table 4b shows a substantial increase in self-recruitment of class I-II among professionals from cohort 1 to cohort 2, apparently at the expense of recruit from the routine non-manual class.

Second, the substantial intergenerational shrinkage of the agricultural class results in low levels of inheritance (table 3A). Less than one-third of men with origins in this class remain farmers, 40% go to the working class, 20% become self-employed and only 13% move to non-manual classes. More surprising is the very low level of self-recruitment of the agricultural classes, which is significantly lower than other national experiences (for instance China) and which remains roughly constant across cohorts. This can be accounted for by the massive transformation of land ownership in Chile, resulting from the agrarian reform and counter-reform, land privatization and emergence of export-oriented agricultural businesses.

Third, almost half of those with origins in the working class (classes VI-VIIa) remain manual workers, with very little difference in outflow patterns between skilled and unskilled workers. Low differentiation within the working class probably results from very low specialization of these occupations. A non-negligible 10% of those with origins in the working class are engaged in agriculture (most likely in the new type of

self-employment. This may be explained by the fact that social origins for the youngest cohort are measured when cohort incumbents were around 14 years old, approximately the period between 1975 and 1985, characterized by a deep economic crisis and market transformation.

－ 185 －

export-oriented agriculture) and more than 20% become self-employed. Inheritance and self-recruitment are also relatively low among the self-employed,

reaching about 30%. With 17% of the members of class IVab coming from the non-manual sector, 25% from the manual one, and 23% from the agricultural classes, the class of self-employed recruit widely. This is further suggestion of the internal heterogeneity of the self-employed in terms of their living conditions.

5. Social Fluidity in Contemporary Chilean Society Trends in absolute mobility are driven by two different factors: Change in the relative size of different classes, known as structural mobility, and change in the association between origins and destinations, net of structural transformations. Now I analyze this latter component, claimed to measure equality of opportunity in a society. I estimate a series of models for the overall association between class of origin and class of destination (table 5).

The independence model, postulating no OD association after control for the margins, fits, unsurprisingly, poorly. Model 2 in table 5 is the 6-class version of the core model of social fluidity, as formulated by Ishida et al. (1995). This is a multi-matrix formulation that includes one inheritance effect for each class, two hierarchical effects, and two affinity effects. The model explains a large 90% of the association under independence, but it does not fit the data under standard statistical criteria. Information about specific characteristics of Chilean society supplemented by examination of the residuals suggests that some modifications might more accurately account for particularities of the Chilean mobility pattern. First, the core model distinguishes the following hierarchical strata: Upper stratum: Classes I+II, Middle stratum: Classes III, IVab and V-VI, Lower stratum: Classes VIIa, IVc+VIIb (where IVc+VIIb is in the lower stratum only as class of destination). This hierarchical classification is arbitrary, at least for countries where no previous information on resources and rewards associated with each social class is available (Torche 2005a). In fact, examination of the mean level of education and earnings by class in Chile (Appendix A) indicates that the Chilean skilled manual class does not rank significantly higher than the unskilled manual workers in terms of education and earnings. As shown, absolute mobility flows are nearly identical for these two classes.

Instead of using an arbitrary hierarchical classification, I apply an empirical strategy based on cluster analysis of classes’ mean years of schooling and logged monthly income, and obtain the following hierarchical ranking of classes for the Chilean case: Upper stratum: I+II, Middle stratum: III, IVab, Lower stratum: V-VI, VIIa, IVc+VIIb. Using these redefined hierarchical effects in model 2, the fit improves significantly -- the BIC coefficient drops from 7.02 to -12.02 -- without using any degree of freedom.

Even if model 2 accounts for 93% of the association under independence, it still does not fit the data according to standard fit statistics (p=.000). When the core model does not fit well to a particular country, Erikson and Goldthorpe (1992) suggest introducing country-specific affinity effects. To prevent arbitrary fitting, these effects should be grounded on knowledge about specific features of the country. Two such modifications will be introduced. First, the use of the empirically-obtained class ranking to model hierarchical effects renders the positive affinity effect AF2b (affinity between

－ 186 －

the classes [V-VI and VIIa], professionals and employers [I and IVab], self-employed and farmers [IVab and IVc-VIIb], farmers and unskilled manual workers [IVc-VIIb and VIIa], and skilled and unskilled working class [V-VI an VIIa]) unnecessary, so I eliminate it. Secondly, examination of residuals indicates that there are two cells with a very large negative residual: The combination of origins in the non-manual classes (I+II and III) and destination in the agricultural class (IVc+VIIb). Therefore, a Chilean-specific parameter captures this negative asymmetric affinity (AF-Ch). This parameter adds to the hierarchical distance between these classes, and suggests significant elite closure that prevents members of the non-manual classes from experiencing long-range downward mobility to class IVc-VIIb. The model that introduces these changes (model 3) fits significantly better (p=.06), and it explains a large 96% of the association under independence. Therefore Model 3 will be selected as the preferred model.

An alternative approach to modeling the association in the mobility table is to understand classes of origin and destination as hierarchically ordered categories, an approach that adequately models hierarchical mobility effects (Hout and Hauser 1992). A flexible version of this approach is the RC2 model (Goodman 1979), which empirically ranks classes of origin and destination each in a single unidimensional scale representing distances in terms of intergenerational mobility, and estimates a single association parameter. Because of the well known higher density in the main diagonal in mobility tables, this model can be accompanied by inheritance parameters (Gerber and Hout 2004). Model 4 presents the fit statistic of this quasi-RC2 model. Even though this model explains 97% of the association under independence, it is not very parsimonious and it fits the data substantially worse than the preferred version of the core, according to the BIC coefficient.

An additional modification of the RC2 model toward a more parsimonious formulation consists of assuming that the hierarchical ranking of classes has not changed between origins and destinations, and therefore to impose six equality constraints. This is done in model 5, and it substantially improves the fit. As a final adjustment, the Chilean negative affinity between classes of origin I+II and III and classes of destination IVc+VIIb (AF-Ch), capturing the protective barrier to long-range downward mobility from the Chilean elite, is added in model 6. Now the fit of the model is the best of all formulations. The BIC coefficient is more negative than the preferred specification of the core model, and the association under independence accounted reaches 98%. Even though I will not use this model in the analysis to ensure comparability of the Chilean case, it provides useful information about mobility in Chile. It highlights that the hierarchical distance between classes is the most important dimension of the Chilean mobility regime, and that the Chilean nonmanual classes (I-II and III) are able to prevent long-range downward mobility of their members. In what follows I focus on the preferred version of the core model (model 3) to analyze the Chilean mobility pattern in comparative perspective.

Table 6 presents the parameter estimates of the preferred version of the core model. As a reference I present the coefficients obtained from the unmodified version of the core model as formulated in Ishida et al. (1995) [model 2 in table 5], but the parameters to be interpreted are those from the preferred version. Parameter estimates for England, France (the two ‘central countries’ from which the core model was obtained) and Japan (Ishida et al 1995) are also presented as reference for comparison.

As in the core countries, the propensity for inheritance of the agricultural

－ 187 －

(IVc+VIIb) and the professional (I+II) classes are the highest in Chile. However, the agricultural class’ inheritance parameter is significantly lower in Chile than in the core countries (about one-third of the value), which is consistent with the aforementioned disruption in the process of land inheritance resulting from the agrarian reform and counter-reform that took place in the country between the 1960s and the 1990s.

Another particularity of the Chilean case is the much weaker inheritance among the self-employed and the skilled working class, indicating the limited class reproduction based on the transfer of specialized capital, skills or socialization for these classes in Chile. This is consistent with the transient and precarious nature of self-employment, which became a “refuge” against unemployment during the crisis and market transformation during the 1970s and 1980s. It is also consistent with the lack of functional specialization among the skilled working class, expressed in a similar hierarchical position to the unskilled manual sector.

The redefined Chilean hierarchical parameters are roughly comparable to those in the core. The positive affinity among non-manual classes I-II and III (AF2) is much higher in Chile than in the core countries and Japan, and it adds to the Chilean-specific negative affinity effect – which captures the difficulty of long-range downward mobility from the non-manual classes to the agricultural occupations at the lower end of the hierarchy – in signaling the substantial closure of the nonmanual elite.

In sum, this analysis confirms findings by Torche (2005a) and Beller and Hout (2006). In spite of its high economic inequality, there is considerable class fluidity in Chile. This “excess fluidity’ appears to be largely driven by weak sectoral barriers to mobility. The most striking Chilean departures from the pattern depicted by the core model are the stronger barriers protecting the non-manual elite from downward mobility and the weaker inheritance of farmers, self-employed and skilled manual workers. Note that the class schema used in this analysis is highly aggregated, and that a more detailed classification – distinguishing, for instance, classes I and II -- will capture even more pronounced hierarchical barriers. Change in the Origin-Destination Association Over Time. This section uses cohort analysis to explore whether the level of social fluidity has changed over time in the context of substantial economic and institutional transformation over the last four decades. I utilize two formulations of class of destination -- the current class position of the respondent, and the class position of his first job after completing school. If findings are consistent across these two specifications, this would provide some assurance that observed trends are not an artifact of life-cycle effects. Panel 1 in table 7 presents models for the association between class of origins and current class position of the respondent across cohorts. Model 1 is the independence model, assuming null intergenerational association. It fits the data poorly and is only used as a baseline for comparison. Model 2 captures the O-D association using a full-interaction pattern and assumes no change in the association across cohorts, i.e. constant social fluidity (CSF). Models 3 and 4 also assume CSF, but they model the association using the unmodified and the Chilean version of the core model respectively (models 1 and 2 in table 6). Not surprisingly, the best fitting model is model 4, which uses the Chilean version of the core model to capture the O-D association.

Models 5, 6 and 7 assess changes across cohorts using the uniform difference (UNIDIFF) or multiplicative-layer effect model (Erikson and Goldthorpe 1992, Xie 1992) which assumes the pattern of association to be invariant, but allows its strength to

－ 188 －

change across cohorts. While model 5 describes the O-D association as full interaction, model 6 uses the unmodified version of the core model, and model 7 uses the Chilean version of the this model. The comparison between model 7 and the constant fluidity model (model 4) indicates that allowing the origin-destination association to vary across cohorts does not significantly improve the fit 4 . This indicates that social fluidity has not changed over the last few decades in Chile. Examination of the cohort-specific UNIDIFF parameters is nevertheless interesting. They suggest that the intergenerational association is similar for cohorts 1 and 2 and stronger for the youngest cohort (c3). Model 8 therefore constraints the O-D association to be the same for c1 and c2, saving one d.f. The fit is still worse than the CSF. Thus, we conclude that even if there is some indication of declining fluidity, change in intergenerational mobility over time is not significant in spite of significant economic and institutional transformation.

To provide a more robust assessment of trends, a replicate analysis replacing respondent’s current class position with first job class position as class of destination is presented in panel B in table 7. The O-D association is again captured by the preferred version of the core model. Findings are similar to the previous specification: Although the UNIDIFF parameters (models 5-7) suggest growing rigidity from c1 to the younger cohorts, changes are not statistically significant, and we cannot reject the hypothesis of temporal constancy. The Role of Education in the Origin-Destination Association. As established by the seminal work of Blau and Duncan (1967), education plays a double role in the stratification process. On the one hand, educational attainment is the main vehicle for reproduction of socioeconomic status across generations; on the other hand, it creates opportunities for mobility independently from social origins. Three components of this dual role can be distinguished: The influence of social origins on education, the impact of education on destination, and the origin-destination association net of education. In order to analyze the mediating role of education in the mobility process and its potential influence on trends over time, table 8 presents models for the 4-way origin-destination-education-cohort table.

Table 8 presents models for the change over time in the association between O-D net of education, i.e. the portion of the intergenerational association that is not mediated by educational attainment but by factors such as the direct transfer of assets, and use of social connections. Note that in this case, the association between origin, education and cohort, and the association between education, destination and cohort are modeled as full-interaction instead of the more parsimonious multi-matrix formulation 5 . Model 1 captures the OD association as full interaction and it assumes no change in OD net of education over time, while models 2 and 3 are also based on the CSF assumption but it models the OD association using the effect matrix formulation, in its unmodified version (model 2) and Chilean variant (model 3). Models 4, 5 and 6 let the O-D association to

4 This conclusion is based on the BIC coefficient. Comparison based on L2 will slightly prefer the unidiff model (L2 CSF- L2 unidiff=8.28, 2 d.f. p=.02. Given the moderate sample size, the L2 statistic will detect rather minor differences in model fit. 5 The three-way full interaction terms OEC and EDC are included even if there is no significant change in OE across cohorts and no significant ED change across cohorts (results not shown, available from the author upon request). Results do not change if OEC EDC terms are replaced by the more parsimonious OC EC DC OE ED formulation.

－ 189 －

vary across cohorts in UNIDIFF manner using full interaction, unmodified version of the core, and Chilean-variant of the core formulation to capture association, respectively. Based on the BIC statistic, the preferred model is 3, assuming no change in the net OD association over time. We conclude therefore that the factors affecting intergenerational fluidity other than education have not change their strength over time. Table 9 presents the parameter estimates for the OD association net of education from the preferred model 3 Table 8, and it compares these net parameter estimates with the gross coefficients presented in table 6. Comparison between gross and net parameters gauges the extent to which educational attainment works as a mechanism of class reproduction. The reduction in the magnitude and significance of the parameters is only moderate. Of the six significant gross O-D parameters, two become insignificant at the p<.05 level, and the others are reduced in varying proportions. The parameter capturing inheritance of the class of professionals and managers becomes insignificant, indicating that the reproduction of the Chilean nonmanual elite is largely mediated by educational attainment. In contrast, education mediated the inheritance of the self-employed class only marginally (reduction of the inheritance parameter of 26%). This finding is in line with other nations, and it highlights that resources that ensure intergenerational reproduction of the self-employed are to large extent orthogonal to educational attainment. In the industrialized world, these resources refer mostly to real and financial capital, as well as occupational socialization. Given that a large proportion of Chilean self-employed lack capital, these resources may include social networks, and specific market knowledge, in addition to socialization.

Interestingly, education plays a negligible role at mediating short-range hierarchical barriers, but a more substantial one in explaining long-range hierarchical barriers, reducing this effect by about 40%. I expected a larger role of education in mediating hierarchical barriers given that Chilean socioeconomic inequality is largely driven by returns to education. Furthermore, the Chilean disaffinity between the non-manual classes of origin and destination in the agricultural class (AF-Ch) is only marginally mediated by education, confirming the relevance of extra-educational resources in preventing downward mobility from the nonmanual elite. This places schooling as a relevant but by no means only mechanism of status reproduction, and it highlights other factors such as family connections, cultural capital, etc. In sum, the analysis of the role of education in the mobility process indicates educational investments are a relevant, but by no means the only, mechanism for the intergenerational persistence of economic advantage. While reproduction of the nonmanual Chilean elite is largely mediated by educational attainment, hierarchical barriers to mobility are determined by factors other than the ability of parents to afford human capital accumulation by offspring.

6. Conclusions This paper has analyzed the Chilean intergenerational mobility regime and its change over the last five decades. The Chilean case is interesting because the country has experienced substantial economic growth since the late 1980s, and a deep market reform during the 1970s and 1980s. The main change in the Chilean class structure consists of the reduction of the agricultural classes, and the increase in the class of professionals and managers. Interestingly, Chile did not experience an expansion of the working class as an

－ 190 －

intermediate stage in the transformation of its class structure, which appears to be a path characteristic of some late industrializing countries. Other particularities of the Chilean class structure is the persistently voluminous self-employed class, a feature that it shares with other Latin American nations, and that has been highlighted by the “informal sector” literature. The Chilean occupational structure appears to have stabilized in the last decade, which suggests convergence with the industrialized countries.

Turning to the pattern of intergenerational fluidity, the analysis highlight several particularities of the Chilean case. Firstly, Chile displays very weak sectoral barriers to mobility, between self-employed and dependent workers, and between agricultural and urban classes, and very weak agricultural inheritance. These traits appear are interpreted as the result of the deep market transformation experienced by the country. Secondly, hierarchical barriers to mobility are substantial and, in the case of the nonmanual classes in the top of the hierarchical strategy, these barriers are supplemented by a protective barrier to long-range downward mobility of their members.

In spite of substantial transformation of the class structure no change in social fluidity is detected in Chile over the last five decades. Even though the models allowing for temporal change suggest growing origin-destination association over time, this change is not statistically significant. Finally, the analysis suggests that educational attainment plays only a partial role in mediating the intergenerational transmission of advantage. While the reproduction of the professional class is largely mediated by education, the inheritance of other classes – in particular, the class of self-employed – is not. Furthermore, hierarchical barriers to mobility appear to largely depend on factors other than education. This finding highlights the need to study factors such as the use of social networks, the direct transmission of assets, and parental investment in quality and not only quantity of education to further understand the mechanisms of intergenerational reproduction in Chile.

－ 191 －

References. Arum, R. and W. Muller. 2004 (eds). The Reemergence of Self-Employment: A Comparative Study of Self-Employment Dynamics and Social Inequality. Princeton: Princeton University Press. Beller, E. and M. Hout. 2006. “Welfare States and Social Mobility: How Educational and Social Policy may Affect Cross-National Differences in the Association between Occupational Origins and Destinations.” Research in Social Stratification and Mobility 24: 353-365. Beyer, H. 2000. “Educacion y Desigualdad de Ingresos: Hacia una Nueva Mirada.” (Education and Earnings Inequality: Towards a New Approach). Estudios Publicos 77. Blau, P. and O.D. Duncan. 1967. The American Class Structure. New York: Free Press. Braun, J. M. Braun, I. Briones and J. Diaz .2000. “Economia Chilena 1810-1995. Estadisticas Historicas.” (Chilean Economy 1810-1995. Historical Statistics) in Documento de Trabajo 187. Santiago: Instituto de Economia P. Universidad Catolica de Chile. Breen, Richard (ed.). 2004. Social Mobility in Europe. Oxford: Oxford University Press. Breen, R. and R. Luijkx 2004a. “Conclusions,” Chapter 15 in Social Mobility in Europe, edited by Richard Breen. Oxford: Oxford University Press. ______ 2004b. “Social Mobility in Europe between 1970 and 2000.” Pp. 37-76 in Social Mobility in Europe, edited by Richard Breen. Oxford: Oxford University Press. Breen, R., R. Luijkx, W. Muller and R. Pollak. 2005. “Non-Persistent Inequality in Educational Attainment: Evidence from Eight European Countries.” Pp. 18-21 in Paper prepared for the meeting of Research Committee 28 (ISA) Inequality and Mobility in Family, School, and Work, Los Angeles. Breen, R. and J. Jonsson. 2007. “Explaining Change in Social Fluidity: Educational Equalization and Educational Expansion in Twentieth-Century Sweden.” American Journal of Sociology 112 (6): 1775–1810. Breen, R. and R. Luijkx. 2007. “Social mobility and education: A comparative analysis of period and cohort trends in Britain and Germany.” in From Origin to Destination. Trends and Mechanisms in Social Stratification Research, edited by S. Scherer, R. Pollak, G. Otte, and M. Gangl . New York: Campus. Brunner, J.J. 1994. “Educacion Superior. Chile en el Contexto Internacional Comparado.” (Higher Education. Chile in the International Comparative Context). Series Documentos de Trabajo. Santiago: CPU. Carnoy, Martin, and Patrick McEwan. 2001. “Privatization Through Vouchers in Developing Countries: The Cases of Chile and Colombia.” Pp. 151–77 in Privatizing Education, edited by

－ 192 －

Henry M. Levin. Boulder CO: Westview Press. Contreras, D., R. Cooper, J. Hermann and C. Nielson. 2005. “Poverty Dynamics and Relative Income Mobility: Chile 1996 and 2001.” Unpublished manuscript, Department of Economics Universidad de Chile. Correa, S. C. Figueroa, A. Jocelyn-Holt, C. Rolle, and M. Vicuna. 2001. Historia del Siglo XX Chileno (History of the Chilean 20th Century). Santiago: Sudamericana. Edwards, S. and A. Cox-Edwards.1991. Monetarism and Liberalization: The Chilean Experiment. Cambridge MA: Ballinger. Erikson, R. and J. Goldthorpe. 1992. The Constant Flux. Oxford: Oxford University Press. Ffrench-Davis, R. 1999. Entre el Neoliberalismo y el Crecimiento con Equidad. Tres Decadas de Politica Economica en Chile. (Between Neoliberalism and Growth with Equity: Three Decades of Economic Policy in Chile.) Santiago: Dolmen. Ganzeboom, H., R. Luijkx, and D. Treiman. 1989. “Intergenerational Class Mobility in Comparative Perspective.” Research in Social Stratification and Mobility 8: 3-55. Gerber, T. and M. Hout. 2004. “Tightening Up: Declining Class Mobility during Russia’s Market Transition.” American Sociological Review 69: 677-703. Goldthorpe, J. (with C. Llewellyn and C. Payne). 1980. Social Mobility and Class Structure in Modern Britain. Second Edition. Oxford: Clarendon Press. Goldthorpe, J. 2007. On Sociology. Volume Two. Second Edition. Stanford: Stanford U. Press. Goodman, Leo. 1979. “Simple Models for the Analysis of Association in Cross-Classifications having Ordered Categories.” Journal of the American Statistical Association 74:537–52. Heath, A. and C. Payne. 1999. “Twentieth Century Trends in Social Mobility in Britain.” Working Paper 70 CREST. National Centre for Social Research and Department of Sociology University of Oxford, Oxford, UK. Hofman, A. 2000. The Economic Development of Latin America in the Twentieth Century. Cheltenham, UK: Edward Elgar. Hout, M. 1988. “More Universalism, Less Structural Mobility: The American Occupational Structure in the 1980s.” American Journal of Sociology 93(6): 1358-1400. Hout, M. and R. Hauser 1992. “Symmetry and Hierarchy in Social Mobility: A Methodological Analysis of the CASMIN Model of Class Mobility.” European Sociological Review 8 (3): 239-265. Hout, M. and T. Di Prete. 2006. “What We Have Learned: RC28’s Contributions to Knowledge About Social Stratification.” Research in Social Stratification and Mobility 24: 1-20.

－ 193 －

Hsieh, C.T. and M. Urquiola. 2006. “The Effects of Generalized School Choice on Achievement and Stratification: Evidence from Chile’s School Voucher Program.” Journal of Public Economics 90: 1477-1503. Infante, R. and E. Klein. 1995. “The Latin American Labor Market 1950-1990.” in Latin America’s Economic Development: Confronting Crisis, edited by Dietz, J. Boulder. CO: Lynne Rienner. Inter American Development Bank 1999. Facing Up to Inequality in Latin America: 1998-99 Report. Washington D.C.: IADB and The Johns Hopkins University Press. Ishida, I., W. Muller and J. Ridge. 1995. “Class Origin, Class Destination and Education: A Cross-National Study of Ten Industrial Nations.” American Journal of Sociology 101: 145-193. Jonsson, J. 1991. “Education, Social Mobility an Social Reproduction in Sweden: Patterns and Changes” in Scandinavian Trends in Welfare and Living Conditions, edited by. E. Hansen et al. Armonck: Sharpe. ______1996. “Stratification in Post-Industrial Society: Are Educational Qualifications Growing in Importance?” in Can Education be Equalized? The Swedish Case in Comparative Perspective, edited by Erikson, R. and J. Jonsson. Boulder. CO: Westview Press. Jonsson, J. and C. Mills. 1993. “Social Mobility in the 1970s and 1980s: A Study of Men and Women in England and Sweden.” European Sociological Review 9: 229-247. Kay, C. 2002. “Chile’s Neo-liberal Agrarian Transformation and the Peasantry.” Journal of Agrarian Change 2 (4): 464-501. Leon, A. and J. Martinez. 2000. “Social Stratification in Chile at the Close of the 20th Century.” in Chile in the Nineties, edited by Toloza, C. and E. Lahera. Santiago: Dolmen. Lipset S. and H. Zetterberg. 1959. “Social Mobility in Industrial Societies.” Pp. 11-75 in Social Mobility in Industrial Society, edited by S. Lipset and R. Bendix. Berkeley: University of California Press. Mare, R. and H.C. Chang. 2006. “Family Attainment Norms and Educational Stratification in the United States and Taiwan: The Effects of Parents’ School Transitions.” in Mobility and Inequality, edited by Morgan, S. D. Grusky and G. Fields. Stanford: Stanford U. Press. Martinez, J. and A. Diaz. 1999. Chile: The Great Transformation. Washington DC: The Brookings Institution and Geneva: The United Nations Research Institute for Social Development. Meller, P. 1991. “Adjustment and Social Costs in Chile During the 1980s.” World Development 19:1545-1551. ______ 1996. Un Siglo de Economia Politica Chilena. (One Century of Political Economy in Chile). Santiago: Andres Bello.

－ 194 －

Mizala, A. and P. Romaguera. 2000. “School Performance and Choice: The Chilean Experience.” Journal of Human Resources 25:392–409. Munoz, O.1990. “El Proceso de Industrializacion: Teorias, Experiencias y Politicas.” Serie Docente 1. Santiago: CIEPLAN. Nunez, J. and C. Risco. 2004. “Movilidad Intergeneracional del Ingreso en un Pais en Desarrollo: El Caso de Chile.” (Intergenerational Income Mobility in a Developing Country: The Chilean Case). Working Document, Department of Economics Universidad de Chile. Nuñez, I. 1997. Historia Reciente de la Educacion Chilena ( Recent History of Chilean Education.) online at www.udec.cl/educacion/biblioteca/principios1/documento2.htm Portes, A. M. Castells and L. Benton.1989. “ The Policy Implications of Informality.” Pp. 298-311 in The Informal Economy: Studies in Advanced and Less Developed Countries, edited by Portes , Castells and Benton. Baltimore: The Johns Hopkins University Press. Portes, A. and K. Hoffman 2002. “Latin American Class Structures: Their Composition and Change During the Neoliberal Era.” Latin-American Research Review 38 (1): 41-82. Robert, P. and E. Bukodi. 2004. “Changes in Intergenerational Class Mobility in Hungary 1973-2000.” in Social Mobility in Europe, edited by Breen, R. Oxford: Oxford University Press. Ryder, N. 1965. “ The Cohort as a Concept on the Study of Social Change.” American Sociological Review . 30 : 843-61. Sapelli, C. 2003. “The Chilean Voucher System: Some New Results and Research Challenges.” Cuadernos de Economia 40:530–38. Sapelli, C. 2007. “A Cohort Analysis of the Income Distribution in Chile.” Working Document #209, Instituto de Economia Universidad Catolica, Mayo. Stallings, B. 1978. Class Conflict and Economic Development in Chile 1958-1973. Stanford, CA: Stanford University Press. Stalling, B. and W. Peres 2000. Growth, Employment and Equity. The Impact of the Economic Reforms in Latin America and the Caribbean. Washington: Brookings. Szelenyi, S. 1998. Equality by Design: The Grand Experiment in Destratification in Socialist Hungary. Stanford: Stanford University Press. Thomas, J. 1996. “The New Economic Model and Labor Markets in Latin America.” in The New Economic Model in Latin America and its Impact on Income Distribution and Poverty, edited by V. ,Bulmer-Thomas. New York: St. Martin’s Press. Torche, F. 2005a. “ Unequal but Fluid. Social Mobility in Chile in Comparative Perspective.” American Sociological Review 70: 422-450.

－ 195 －

－ 196 －

______. 2005b. “Privatization Reform and Inequality of Educational Opportunity: The Case of Chile.” Sociology of Education 78: 316-343. Torche, F. and C. Costa-Ribeiro. 2007. “Educational Expansion and Decline in the 'Mobility Returns' of Higher Education: The Case of Brazil and Mexico.” Pp. 24-27 in paper presented at the RC28 Meeting Social Inequality and Mobility in the Process of Social Transformation, Brno Czech Republic. Torche, F. 2007. Movilidad Intergeneracional Educacional en Chile (Intergenerational Educational Mobility in Chile.) Mimeo. Treiman, D. and K.B. Yip. 1989. “Educational and Occupational Attainment in 21 Countries.” in Cross National Research in Sociology, edited by M. Kohn. Newbury Park: Sage. Tyree, A. M. Semyonov, R. Hodge. 1979. “Gaps and Glissandos: Inequality, Economic Development and Social Mobility in 24 Countries.” American Sociological Review 44: 410-424. UNESCO. 1999. OECD/UIS World Education Indicators. Geneva: UNESCO. United Nations. 2005. World Development Report. New York: United Nations. Vallet, L.A. 2004. “Change in Intergenerational Class Mobility in France from the 1970s to the 1990s and its Explanation: An Analysis Following the CASMIN Approach.” in Social Mobility in Europe, edited by Breen, R. Oxford: Oxford U. Press. Velasquez, M. 1990 “Evolucion del Empleo Publico en Chile: 1974-1985.” (Evolution if Public Employment in Chile: 1974-1985) in Empleo Publico Frente a la Crisis: Estudios sobre America Latina, edited by Marshall, A. Geneva: Instituto Internacional de Estudios Laborales. Wong, R. and R. Hauser. 1992. “Trends in Occupational Mobility in Hungary under Socialism.” Social Science Research 21: 419-444. Xie, Yu. 1992. “The Log-Multiplicative Layer Effect Model for Comparing Mobility Tables.” American Sociological Review 57 (3): 380-395.

－ 19

7 －

Tabl

e 1.

Bas

ic S

ocio

econ

omic

Indi

cato

rs C

hile

195

0-20

00.

Year

1.

GD

P Em

ploy

men

t Dis

tribu

tion

5. U

rban

izat

ion

6. R

eal

7. G

ini

Educ

atio

nal E

nrol

lmen

t

capi

ta

2. A

gric

.& M

inin

g3.

Man

ufac

ture

4.

Ser

vice

sR

ate

Wag

e C

oeffi

cien

t 8.

Sec

onda

ry

9. T

ertia

ry

1950

$1

,857

39

%

20%

41

%

61%

42

.3

19

1

1955

$1

,998

37

%

20%

42

%

36

.0

23

2

1960

$2

,150

35

%

20%

46

%

68%

52

.6

0.38

9 24

3

1965

$2

,294

31

%

19%

50

%

72%

58

.7

28

4

1970

$2

,606

26

%

18%

55

%

75%

87

.6

0.43

4 35

7

1975

$2

,139

22

%

16%

62

%

78%

54

.3

50

12

1980

$2

,857

20

%

15%

65

%

81%

78

.0

0.46

9 64

8

1985

$2

,530

22

%

14%

64

%

83%

71

.8

0,48

5 72

13

1990

$3

,224

23

%

17%

60

%

83%

80

.9

0,53

4 75

15

1995

$4

,556

19

%

18%

64

%

84%

10

0.0

0,46

8 76

20

2000

$4

,982

15

%

16%

68

%

85%

12

0.0

83

37

Table 2A. Distribution of Class Origins, Class Destination and Class at Labor Market Entry. Chilean Men ages 30-64. 2001. Origin Distribution Destination Distribution Labor market entry I+II Service Class 13.9 20.4 13.8 III Routine Non-Manual 5.5 6.5 13.8 IVab Petty Bourgeoisie 19.1 22.8 7.8 IVc+VIIb Agricultural 25.8 12.7 17.7 V-VI Skilled Manual 19.3 19.0 18.4 VIIa Unskilled Manual 16.3 18.5 28.6 Total 100% 100% 100% ID 13.4 Total Mobility Rate 70.1%

Table 2B. Distribution of class of origin, class of destination, and class of labor market entry by birth cohort. Chilean men ages 30-64. 2001. Origin Distribution Cohort 1937-50 Cohort 1951-60 Cohort 1961-71 I+II 11.3 17.2 13.7 III 5 5 6.7 IVab 19.8 15.8 21.5 IVc-VIIb 31.8 23.9 20.9 V-VI 17.8 21.1 19.3 VIIa 14.3 17 17.9 Destination Distribution Cohort 1937-50 Cohort 1951-60 Cohort 1961-71 I+II 17.7 22.4 21.7 III 5 7.3 7.5 IVab 24.8 23.7 19.6 IVc-VIIb 15.4 10.6 11.6 V-VI 18.3 19.8 19.1 VIIa 18.7 16.2 20.5 ID 16.4 15.4 11.5 Total Mob. 71.7 71.1 67 Labor market entry Distribution Cohort 1937-50 Cohort 1951-60 Cohort 1961-71 I+II 13.3 16.1 12.3 III 10.9 15.9 15 IVab 7.5 7.3 8.4 IVc-VIIb 22 14.6 15.8 V-VI 17.4 18.2 19.7 VIIa 29 27.8 28.8 ID 20.5 22.6 21.7 Total Mob. 64.4 63.1 65.3 Weighted Sample size 1103 952 940

－ 198 －

Table 3.A. Outflow Distribution Class of Origin- Current Class. Chilean Men ages 30-64. 2001. I-II IIIab IVab IVc-VIIb V-VI VIIa Total I-II 0.53 0.11 0.20 0.01 0.09 0.07 1.00 IIIab 0.38 0.09 0.21 0.01 0.16 0.16 1.00 IVab 0.21 0.07 0.31 0.08 0.18 0.15 1.00 IVc-VIIb 0.09 0.04 0.20 0.28 0.19 0.20 1.00 V-VI 0.15 0.06 0.21 0.09 0.26 0.24 1.00 VIIa 0.10 0.08 0.23 0.13 0.22 0.25 1.00 Total 0.21 0.07 0.22 0.12 0.19 0.19 1.00

Table 3.B. Inflow Distribution Class of Origin- Current Class. Chilean Men ages 30-64. 2001. I-II IIIab IVab IVc-VIIb V-VI VIIa Total I-II 0.36 0.23 0.12 0.01 0.07 0.05 0.14 IIIab 0.10 0.08 0.05 0.00 0.05 0.05 0.06 IVab 0.20 0.20 0.26 0.12 0.18 0.15 0.19 IVc-VIIb 0.12 0.14 0.23 0.56 0.26 0.28 0.26 V-VI 0.15 0.16 0.18 0.13 0.26 0.25 0.19 VIIa 0.08 0.19 0.17 0.16 0.19 0.22 0.16 Total 1.00 1.00 1.00 1.00 1.00 1.00 1.00

－ 199 －

Table 4.A. Outflow Rates by Cohort. Chilean Men ages 30-64. 2001. I-II IIIab IVab IVc-VIIb V-VI VIIa I-II 1937-1950 .40 .08 .28 .03 .13 .09 1951-1960 .57 .08 .23 .00 .07 .06 1961-1971 .61 .17 .10 .01 .08 .04 III 1937-1950 .52 .08 .19 .00 .12 .10 1951-1960 .27 .11 .27 .00 .16 .18 1961-1971 .32 .08 .19 .02 .20 .19 IVab 1937-1950 .24 .04 .31 .08 .18 .15 1951-1960 .18 .10 .29 .04 .24 .15 1961-1971 .21 .07 .32 .11 .14 .15 IVc+VIIb 1937-1950 .08 .04 .24 .29 .19 .17 1951-1960 .13 .04 .21 .23 .21 .17 1961-1971 .07 .03 .12 .32 .18 .28 V+VI 1937-1950 .15 .07 .22 .06 .23 .26 1951-1960 .17 .05 .20 .13 .24 .21 1961-1971 .14 .04 .21 .08 .29 .24 VIIa 1937-1950 .05 .04 .21 .23 .19 .28 1951-1960 .09 .09 .28 .11 .23 .20 1961-1971 .15 .10 .20 .12 .19 .20 Table 4.B. Inflow Rates by Cohort. Chilean Men ages 30-64. 2001. I-II IIIab IVab IVc-VIIb V-VI VIIa I-II 1937-1950 .25 .17 .13 .02 .08 .06 1951-1960 .44 .19 .16 .00 .06 .06 1961-1971 .38 .30 .07 .01 .05 .03 III 1937-1950 .15 .08 .04 .00 .03 .03 1951-1960 .06 .08 .06 .00 .04 .06 1961-1971 .10 .08 .06 .01 .07 .06 IVab 1937-1950 .27 .15 .25 .11 .19 .16 1951-1960 .13 .22 .19 .05 .19 .15 1961-1971 .20 .21 .35 .21 .16 .16 IVc+VIIb 1937-1950 .14 .23 .31 .59 .32 .30 1951-1960 .14 .14 .21 .52 .25 .26 1961-1971 .07 .08 .13 .58 .20 .29 V+VI 1937-1950 .15 .25 .16 .07 .23 .25 1951-1960 .16 .16 .18 .25 .26 .27 1961-1971 .13 .11 .20 .13 .30 .23 VIIa 1937-1950 .04 .12 .12 .21 .15 .21 1951-1960 .07 .22 .20 .17 .20 .21 1961-1971 .12 .23 .19 .07 .22 .24

－ 200 －

Table 5. Models of association class of origin*current class. Chilean men ages 30-64. 20011. L2 df L2 adj.2 BIC Assoc. Expl. p

1. Independence model 598.83 25 437.39 509.34 0.00% .000

2. Unmodified core model (Ishida et al 1995) 60.73 15 47.86 7.02 89.96% .000

2. Core, modified Hi effects 41.70 15 34.17 -12.02 93.04% .000

3. Core, modified Hi, and Af-Ch 24.54 15 21.86 -29.20 96.00% .060

4. Quasi-RC2 20.22 10 17.34 -15.58 96.62% .027

5. Quasi- Equal RC2 27.11 14 23.43 -23.00 95.47% .019

6. Quasi-Equal RC2 and Af-Ch 13.42 13 13.30 -33.12 97.76% .461 1 Gray shading indicates preferred model. 2 Adjusted to n=2,000.

Table 6. Parameter Estimates, Preferred Core Model for 6*6 Chilean Mobility Table. Men ages 30-64. 2001.

1 Core Unmodif. 2 Core Chile 3 England 4 France 5 Japan

Coeff. SE Coeff. SE Coeff. SE Coeff. SE Coeff. SE

IN I-II 1.062*** (.189) .691*** (.190) 1.229** (.139) .954** (.178) -.378 (.354)

IN III .658** (.305) .498† (.308) .118 (.162) .124 (.191) .381 (.332)

IN IVab .394*** (.116) .290* (.159) 1.115** (.104) 1.409** (.125) .839** (.203)

IN IVc-VIIb 1.452*** (.130) 1.106*** (.124) 3.528** (.161) 3.298** (.151) 2.138** (.228)

IN V-VI .295* (.119) .024 (.123) .433** (.064) .531** (.103) .878** (.256)

IN VIIa .151 (.156) .114 (.124) .398** (.125) -.103 (.148) .184 (.325)

Hi1 (Ch) -.043 (.065) -.177** (.075) -.194** (.055) -.343** (.056) -.363** (.109)

Hi2 (Ch) -.539*** (.129) -.756*** (.119) -.437** (.086) -.879** (.119) -.827** (.217)

AF2A 1.076*** (.143) .675*** (.157) .470** (.091) .350** (126) -.058 (.206)

AF2B .196*** (.058) NA .388** (.044) .351** (.062) -.004 (.134)

AF-Ch/ AF-JP NA -1.611*** (.464) NA NA -1.003** (.267)†p<.10 * p<.05, ** p<.01 *** p<.001. Standard Errors in parentheses. Parameter estimates for England, France, and Japan obtained from Ishida et al. (1995) and presented as referent for comparison.

－ 201 －

－ 202 －

Table 7. Goodness of Fit Statistics for Chilean Mobility Models across Cohorts. Men 30-64. 2001.1

Panel 1. Father’s Class Position * Current Class Position Model L2 df BIC L2 adj. 2 Assoc. Expl.

1. Conditional Independence (OC DC) 708.63 75 113.83 530.79 0%

2. CSF-Full interaction (OC DC OD-f) 111.67 50 -284.86 94.35 84%

3. CSF-Core model unmodified (OC DC OD-m) 176.13 65 -339.37 144.92 75%

4. CSF-Core model Chile (OC DC OD-mCh) 133.58 65 -382.15 114.32 81%

5. Unidiff-Full interaction (OC DC OD-f[uni]) 103.9 48 -276.8 88.2 85%

C1 1937-50 C2 1951-60 C3 1961-71

UNIDIFF parameters 1.00 1.069 1.373

6. Unidiff-Core unmodified (OC DC OD-m[uni]) 164.0 63 -225.67

C1 1737-50 C2 1951-60 C3 1961-71

UNIDIFF parameters 1 1.023 1.404

7. Unidiff-Core Chile (OC DC OD-mCh[uni]) 125.78 63 -373.86 108.15 82%

C1 1737-50 C2 1951-60 C3 1961-71

UNIDIFF parameters 1 1.085 1.365

8. Unidiff-Core Chile Constr. (OC DC OD-mCh[uni-CR]) 126.18 64 -381.79 108.72 82% (c1=c2) C1 1737-50 C2 1951-60 C3 1961-71

UNIDIFF parameters 1 1 1.315 1 Gray shading indicates preferred model.

2 L2 adjusted to n=2000

Panel 2. Father’s Class Position * First Job Class Position Model L2 df BIC L2 adj. Assoc. Expl.

1. Conditional Independence (OC DC) 1083.9 75 489.27 800.57 0% 2. CSF-Full interaction (OC DC OD-f) 83.52 50 -312.90 74.11 92% 3. CSF-Core model unmodified (OC DC OD-m) 132.03 65 -383.32 113.21 88% 4. CSF-Core model Chile (OC DC OD-mCh) 112.11 65 -403.24 98.88 90% 5. Unidiff-Full interaction (OC DC OD-f[uni]) 78.8 48 -301.76 70.15 93% C1 1737-50 C2 1951-60 C3 1961-71 UNIDIFF parameters 1.00 1.195 1.145 6. Unidiff-Core Chile (OC DC OD-mCH[uni]) 126.59 63 -372.74 C1 1737-50 C2 1951-60 C3 1961-71 UNIDIFF parameters 1.00 1.204 1.159 7. UNIDIFF-Core Chile (OC DC OD-mCh[uni]) 107.91 63 -391.58 95.30 90% C1 1737-50 C2 1951-60 C3 1961-71 UNIDIFF parameters 1 1.187 1.118 8. Unidiff-Core Chile Constr. (OC DC OD-mCh[uni-CR]) 108.4 64 -399.02 95.94 90% (c2=c3) C1 1737-50 C2 1951-60 C3 1961-71 UNIDIFF parameters 1 1.152 1.152

－ 20

3 －

Tabl

e 8.

Res

ults

of m

odel

s fitt

ed to

4-w

ay ta

ble

Orig

in b

y D

estin

atio

n by

Edu

catio

n by

Coh

ort,

Chi

lean

men

30-

64 y

ears

old

, 200

1.

Mod

el

L2D

f B

IC

Tren

ds in

OD

1

1. C

SF fu

ll in

tera

ctio

n (O

EC E

DC

OD

) 22

8.84

20

0 -1

359.

4 2.

CSF

mat

rix fo

rmul

atio

n, o

rigin

al (O

EC E

DC

OD

-mO

) 27

5.76

21

5 -1

431.

7 3.

CSF

mat

rix fo

rmul

atio

n, C

hile

an v

ersi

on (O

EC E

DC

OD

-mC

h)

246.

20

215

-146

1.2

4. U

nidi

ff O

D-C

full

inte

ract

ion

(OEC

ED

C u

niO

DC

) 22

2.64

19

8 -1

349.

8

C1

1737

-50

C2

1951

-60

C3

1961

-71

U

nidi

ff pa

ram

eter

s

1.00

1.

17

1.66

5. U

nidi

ff O

D-C

mat

rix fo

rmul

atio

n, o

rigin

al (O

EC E

DC

uni

OD

C-m

O)

264.

69

213

-142

6.8

C

1 17

37-5

0 C

2 19

51-6

0 C

3 19

61-7

1

Uni

diff

para

met

ers

1.00

1.

02

1.75

6. U

nidi

ff O

D-C

mat

rix fo

rmul

atio

n, C

hile

an v

ersi

on (O

EC E

DC

uni

OD

C-m

Ch)

23

8.82

21

3 -1

452.

7

C1

1737

-50

C2

1951

-60

C3

1961

-71

U

nidi

ff pa

ram

eter

s

1.

00

1.14

1.

64

7. U

nidi

ff-co

nstra

ined

OD

-C m

atrix

form

ulat

ion

(OEC

ED

C u

ni c

onst

r.OD

C-m

Ch)

23

9.25

21

4 -1

460.

22

(c1=

c2)

C1

1737

-50

C2

1951

-60

C3

1961

-71

U

nidi

ff pa

ram

eter

s

1.

00

1.00

1.

52

Not

e: O

= C

lass

Orig

ins,

D=C

lass

Des

tinat

ions

, E=E

duca

tiona

l Lev

el, C

=Coh

ort.

1 Thr

ee-w

ay f

ull i

nter

actio

n O

EC a

nd E

DC

incl

uded

eve

n if

ther

e is

no

sign

ifica

nt c

hang

e in

OE

acro

ss c

ohor

ts a

nd n

o si

gnifi

cant

ED

cha

nge

acro

ss c

ohor

ts

(res

ults

not

sho

wn,

ava

ilabl

e fr

om th

e au

thor

upo

n re

ques

t). R

esul

ts fr

om m

odel

s re

plac

ing

OEC

ED

C w

ith m

ore

pars

imon

ious

OC

EC

DC

OE

ED fo

rmul

atio

n ar

e su

bsta

ntiv

ely

iden

tical

.

－ 20

4 －

Ta

ble

9. P

erce

nt R

educ

tion

in th

e O

D E

ffect

Mat

rix P

aram

eter

s Afte

r Con

trolli

ng fo

r Edu

catio

n.

Mod

el 1

M

odel

2

OD

1O

D-E

C2

%

Red

. Coe

ffs.

b

s.e

. M

odel

1-2

IN

I-II

0.

691*

**

(.1

90)

0.07

4

(

.217

) 89

.2%

IN

III

0.49

8

(.308

)

0.

202

(.3

04)

---

IN IV

ab

0.29

0†

(.1

59)

0.27

5†

(.1

58)

5.1%

IN

IVc-

VII

b 1.

106*

**

(.1

24)

0.81

0***

(

.130

) 26

.7%

IN

V-V

I 0.

024

(.1

23)

0.11

7

(

.125

) --

- IN

VII

a 0.

114

(.1

24)

0.13

0

(

.126

) --

- H

i1 (C

h)

-0.1

77*

(.0

75)

-0.1

74*

(.075

) 1.

5%

Hi2

(Ch)

-0

.756

***

(

.119

) -0

.467

***

(.1

29)

38.2

%

AF2

A

0.6

75**

*

(.1

57)

0.32

3†

(.1

76)

52.2

%

AF-

Ch

-1.6

11**

*

(.4

64)

-1.2

11**

(.406

) 24

.8%

1 Orig

in-D

estin

atio

n as

soci

atio

n, n

o co

ntro

ls (O

D),

Mod

el 3

tabl

e 5.

2 O

rigin

-Des

tinat

ion

asso

ciat

ion

net o

f edu

catio

n an

d co

hort.

OEC

ED

C O

D-m

Ch,

Mod

el 3

tabl

e 8.

Appendix A: Mean Total Years of Schooling and Logged Monthly Earnings (Chilean Pesos) by Class, Chilean Males and Females. Class Education Earnings

I-II 14.9 6.3

III 11.4 5.4

Ivab 9.3 5.2

IVc-VIIb 6.6 4.6

V-VI 9 5.0

VIIa 8.3 4.9

Total 10.0 5.3

－ 205 －

Publication & Presentation List (既発表成果一覧) 1. Publications (論文) None 2. Presentations (学会等報告) Hiroshi Ishida and Satoshi Miwa, "Social Mobility and Education in Postwar Japan," paper presented at the meeting of the International Sociological Association, Research Committee on Social Stratification and Mobility (RC28) at UCLA, Los Angeles, CA, U.S.A, 19-21 August, 2005 Hiroshi Ishida and Satoshi Miwa, "Trends in Intergenerational Class Mobility in Japan in the late 20th Century," paper presented at the meeting of the International Sociological Association, Research Committee on Social Stratification and Mobility (RC28) at Masaryk University, Brno, Czech Republic, 25-27 May, 2007 Hiroshi Ishida and Satoshi Miwa, ”Social Mobility in Post-war Japan: Cross-temporal and Cross-national Comparisons,” paper presented at the International Conference on "East Asian Comparative Research," National Taiwan University, Taipei, Taiwan, 24-25 November, 2007 Park, Hyunjoon and Jongchun Cha. 2007. “Intergenerational Social Mobility in Korea: 1988-2005,” paper presented at the annual meeting of American Sociological Association, New York, 11-14 August, 2007. Xiaogang Wu, 2007. "Trends in Class Mobility and Education in China," paper presented at the International Conference on "East Asian Comparative Research," National Taiwan University, Taipei, Taiwan, 24-25 November, 2007

－ 206 －

２００５年 SSM 調査シリーズ １４ 後発産業社会の社会階層と社会移動

石田 浩 編

２００８年３月１０日発行

発行 ２００５年 SSM 調査研究会

〒113-0033 東京都文京区本郷 7-3-1

東京大学社会科学研究所石田研究室（発行担当）

〒980-8576 仙台市青葉区川内 27-1 東北大学大学院文学研究科行動科学研究室内

2005 年社会階層と社会移動調査研究会事務局（事務局）