shades of blue: confidence in the police in the world

10
Shades of blue: Condence in the police in the world Liqun Cao a, , Yung-Lien Lai b , Ruohui Zhao c a Faculty of Social Science and Humanities, University of Ontario Institute of Technology, 55 Bond Street East, Oshawa, ON, Canada L1G 0A5 b Graduate School of Criminology, National Taipei University, 151 University Road, Sanxia District, New Taipei City, 23741, Taiwan, Republic of China c Department of Sociology, University of Macau, China abstract article info Available online 13 December 2011 Purpose: The present study tests the hypothesis that regime nature as a structural characteristic explains variations in public condence in the police. Methods: Combining ve sources of data from 50 nations with 69,309 respondents, the current article extends the extant research by using hierarchical logistic regression analyses with ample sample sizes at both levels to test the hypothesis with a series of control variables. Findings: In addition to the largely consistent ndings from the individual-level predictors, the results show that that there is a U-shaper convex curvilinear relationship between the levels of democracy and condence in the police. Residents in long-term stable authoritarian regimes as well as in long-term stable democracies display elevated levels of condence in the police, whereas short-term or unstable authoritarian nations and nations in democratic transition have the lowest level of condence in the police. Besides, condence in the police is higher among citizens in nations with more government efciency and is lower among residents of countries with higher homicide rates. Conclusion: Regime nature is important in understanding condence in the police. In addition, governments should make more efforts to promote their efciency in order to win citizenssupport and they are expected to reduce homicide rates. © 2011 Elsevier Ltd. All rights reserved. Introduction A burgeoning literature on public condence in the police has emerged in the wake of the third waveof democratization (Huntington, 1991) and as the spread of community policing (Eck & Rosenbaum, 1994). Condence in the police is often regarded as an embodiment of low policing (Brodeur, 1983) or democratic policing (Goldsmith, 2005; Skolnick, 1994), as an indicator of police effectiveness (Sampson & Jeglum-Bartusch, 1998), and as a gauge of police perfor- mance (Ren, Cao, Lovrich, & Gaffney, 2005). In addition, a proliferation of hierarchical linear modelling techniques (HLM) has resulted in taking into consideration of the impacts of structural characteristics on con- dence in the police (Dai & Johnson, 2009; Huebner, Schafer, & Bynum, 2004; Reisig & Parks, 2000; Sampson & Jeglum-Bartusch, 1998; Wu, Sun, & Triplett, 2009), including the recent expansion of such studies in the comparative literature of condence in the police (Ivković, 2008; Jang, Joo, & Zhao, 2010). The limited comparative literature on this topic has neglected three issues. First, the existing comparative literature on condence in the police has not paid enough attention to the nature of political regime on condence in the police. A detailed comparison of condence in the police between China and Taiwan shows that the regime nature is important in shaping public condence in the police (Lai, Cao, & Zhao, 2010). Condence in the police is higher in China because the commu- nist regime has a tight grip over the media and over the public expres- sion of emotion. Jang et al.'s test (2010) of the impact of democracy on condence in the police, however, did not nd the relationship signi- cant. His thumbnail sample of fteen nations is inadequate to draw any rm conclusion. Whether this thesis can survive a test with a larger sample and more diversied pool of nations remains empirical. Second, while individual level factors, such as socio-demographic, experiential (personal and vicarious experiences with crime), and attitudinal variables (Brown & Benedict, 2002; Cao & Burton, 2006; Cao, Stack, & Sun, 1998; Cao & Zhao, 2005; Correia, Reisig, & Lovrich, 1996; Hurst & Frank, 2000; Lai et al., 2010; Skogan, 2009; Weitzer & Tuch, 2006) are extensively studied, national characteris- tics on condence in the police have been scant. Prior research within the U.S. has demonstrated the power of structural variables, such as concentrated disadvantage, on condence in the police. The thesis, however, has not been tested with international data. Neither of the two studies (Ivković, 2008; Jang et al., 2010) with multi-level data examines the effect of inequality on condence in the police. While most sources of public condence in the police have been extensively studied by often excellent research at the individual level, additional studies, using more rened measures at the national/structural level, are warranted. Journal of Criminal Justice 40 (2012) 4049 Corresponding author. Tel.: + 1 905 721 8668x3804; fax: + 1 905 721 3371. E-mail address: [email protected] (L. Cao). 0047-2352/$ see front matter © 2011 Elsevier Ltd. All rights reserved. doi:10.1016/j.jcrimjus.2011.11.006 Contents lists available at SciVerse ScienceDirect Journal of Criminal Justice

Upload: liqun-cao

Post on 05-Sep-2016

216 views

Category:

Documents


1 download

TRANSCRIPT

Page 1: Shades of blue: Confidence in the police in the world

Journal of Criminal Justice 40 (2012) 40–49

Contents lists available at SciVerse ScienceDirect

Journal of Criminal Justice

Shades of blue: Confidence in the police in the world

Liqun Cao a,⁎, Yung-Lien Lai b, Ruohui Zhao c

a Faculty of Social Science and Humanities, University of Ontario Institute of Technology, 55 Bond Street East, Oshawa, ON, Canada L1G 0A5b Graduate School of Criminology, National Taipei University, 151 University Road, Sanxia District, New Taipei City, 23741, Taiwan, Republic of Chinac Department of Sociology, University of Macau, China

⁎ Corresponding author. Tel.: +1 905 721 8668x3804E-mail address: [email protected] (L. Cao).

0047-2352/$ – see front matter © 2011 Elsevier Ltd. Alldoi:10.1016/j.jcrimjus.2011.11.006

a b s t r a c t

a r t i c l e i n f o

Available online 13 December 2011

Purpose: The present study tests the hypothesis that regime nature as a structural characteristic explainsvariations in public confidence in the police.Methods: Combining five sources of data from 50 nations with 69,309 respondents, the current article extends the extant research by using hierarchical logistic regression analyses with ample sample sizes atboth levels to test the hypothesis with a series of control variables.Findings: In addition to the largely consistent findings from the individual-level predictors, the results showthat that there is a U-shaper convex curvilinear relationship between the levels of democracy and confidencein the police. Residents in long-term stable authoritarian regimes as well as in long-term stable democraciesdisplay elevated levels of confidence in the police, whereas short-term or unstable authoritarian nations andnations in democratic transition have the lowest level of confidence in the police. Besides, confidence in thepolice is higher among citizens in nations with more government efficiency and is lower among residents ofcountries with higher homicide rates.Conclusion: Regime nature is important in understanding confidence in the police. In addition, governmentsshould make more efforts to promote their efficiency in order to win citizens’ support and they are expectedto reduce homicide rates.

© 2011 Elsevier Ltd. All rights reserved.

Introduction

A burgeoning literature on public confidence in the police hasemerged in the wake of the “third wave” of democratization(Huntington, 1991) and as the spread of community policing (Eck &Rosenbaum, 1994). Confidence in the police is often regarded as anembodiment of low policing (Brodeur, 1983) or democratic policing(Goldsmith, 2005; Skolnick, 1994), as an indicator of police effectiveness(Sampson & Jeglum-Bartusch, 1998), and as a gauge of police perfor-mance (Ren, Cao, Lovrich, & Gaffney, 2005). In addition, a proliferationof hierarchical linearmodelling techniques (HLM) has resulted in takinginto consideration of the impacts of structural characteristics on confi-dence in the police (Dai & Johnson, 2009; Huebner, Schafer, & Bynum,2004; Reisig & Parks, 2000; Sampson & Jeglum-Bartusch, 1998; Wu,Sun, & Triplett, 2009), including the recent expansion of such studiesin the comparative literature of confidence in the police (Ivković,2008; Jang, Joo, & Zhao, 2010).

The limited comparative literature on this topic has neglected threeissues. First, the existing comparative literature on confidence in thepolice has not paid enough attention to the nature of political regimeon confidence in the police. A detailed comparison of confidence in

; fax: +1 905 721 3371.

rights reserved.

the police between China and Taiwan shows that the regime nature isimportant in shaping public confidence in the police (Lai, Cao, & Zhao,2010). Confidence in the police is higher in China because the commu-nist regime has a tight grip over the media and over the public expres-sion of emotion. Jang et al.'s test (2010) of the impact of democracy onconfidence in the police, however, did not find the relationship signifi-cant. His thumbnail sample of fifteen nations is inadequate to drawany firm conclusion.Whether this thesis can survive a test with a largersample and more diversified pool of nations remains empirical.

Second, while individual level factors, such as socio-demographic,experiential (personal and vicarious experiences with crime), andattitudinal variables (Brown & Benedict, 2002; Cao & Burton, 2006;Cao, Stack, & Sun, 1998; Cao & Zhao, 2005; Correia, Reisig, &Lovrich, 1996; Hurst & Frank, 2000; Lai et al., 2010; Skogan, 2009;Weitzer & Tuch, 2006) are extensively studied, national characteris-tics on confidence in the police have been scant. Prior research withinthe U.S. has demonstrated the power of structural variables, such asconcentrated disadvantage, on confidence in the police. The thesis,however, has not been tested with international data. Neither of thetwo studies (Ivković, 2008; Jang et al., 2010) with multi-level dataexamines the effect of inequality on confidence in the police. Whilemost sources of public confidence in the police have been extensivelystudied by often excellent research at the individual level, additionalstudies, using more refined measures at the national/structurallevel, are warranted.

Page 2: Shades of blue: Confidence in the police in the world

41L. Cao et al. / Journal of Criminal Justice 40 (2012) 40–49

Third, the two studies (Ivković, 2008; Jang et al., 2010) with multi-level data are inadequate to draw any robust conclusion because thesamples are not large enough. Jang et al. (2010) used a sample ofonly 15 nations and Ivković (2008) relied on a sample of 28 countriesin Europe and North America. The smaller number of nations in thesamples tends to produce an unstable coefficient in hierarchical linearmodeling (Raudenbush & Bryk, 2002). All other studies comparedconfidence in the police either in two (e.g., Cao et al., 1998) or morenations (e.g., Cao & Burton, 2006; Cao & Zhao, 2005; Stack & Cao,1998), and the variable nations is the only structural predictor.While important insight can be drawn from more detailed studiescomparing two to three nations, results from a larger sample aremore robust and more generalizable. Accordingly, a new test with alarger sample and with more control variables is warranted.

This study, therefore, attempted to advance our understanding ofcomparative confidence in the police in three ways. First, macro-level data from four sources were combined with micro-level datafrom 2005 World Values Surveys consisting of 50 nations from fivecontinents and Oceania with approximately 69,309 respondents,resulting in a new data set with two levels of information. The nationsin the sample represent a much more diversified pool of states thanthose in prior studies. Second, the hypothesis that an importantsource of variations in confidence in the police lies in the differentregime nature is tested. Building on the insight of Sung (2006), wetest the convex curvilinear relationship between democracy and con-fidence in the police. Third, heeding to the call from Jang et al. (2010),this study examines the effect of the Gini Index on confidence in thepolice with hierarchical linear modeling (HLM) and with the effectof the volumes of homicide and government effectiveness index ascontrol variables.

Theory and literature review

Several overlapping theories provide the foundation for explana-tions of public confidence in the police. From the standpoint of instru-mentalism (Baumer, Messner, & Rosenfeld, 2003; Stack, Cao, &Adamzyck, 2007; Tyler & Weber, 1982), there is an associationbetween the crime rate and public opinions, such as confidence inthe police. High crime volume is frequently seen as promoting a prac-tical need for greater social control in general and a special demandfor tighter control of the lower class from the elite in particular. Theneed for control may emanate from correlates of high crime volumesuch as fear, victimization, and cries for retribution. Reisig and Parks(2000) called the “quality of life”model while Skogan (2009) dubbedthis relation between crime and confidence as “accountability modelof opinion about the police.” The police as the most visible social con-trol agency are at the receiving end of such demand. Extending theseperspectives, the quality of government has been used to predict pub-lic support for the police (Ivković, 2008).

In addition, social stratification has an impact on citizens’ view ofthe police because people at the bottom have the least ability to influ-ence the structures of power that constrains their lives (Sampson &Jeglum-Bartusch, 1998). As the group position perspective (Blumer,1958; Wu et al., 2009) proposes, police responses to incidents vary,often depending upon the location of the citizen in social space asmeasured by race, sex, age and social class. Blau and Blau (1982)argue that societies with greater economic inequalities generally fos-ter conflict and violence. Jacobs (1979) shows that the greater theeconomic cleavages the more likely will elites exercise coercive con-trol. Confidence in the police, therefore, is related to one's positionin the social stratification with the elite having higher confidence inthe police and the poor having lower confidence.

A third line of argument is that confidence in the police is related tothe form of government. In their classical work, Almond and Verba(1963) tested their proposition that there was a positive associationbetween public confidence in the police and the level of democracy in

a nation. They found that confidence in the police was higher in long-term stable democratic nations such as the U.S. and Britain than inMexico where the history of democracy was much shorter. When thisthesis was expanded in the comparative study of confidence in thepolice between authoritarian regime and democratic regime andbetween Muslim states and non-Muslim states, Cao and Burton(2006) observed that such research is complex. In an authoritariansociety, public opinion on politically sensitive topics is inaccuratebecause respondents are afraid of government persecution. Similarly,in a society with strong religious consensus, such as a Muslim state,expressed opinions may be less reflective of true public sentiment. Inaddition, public opinion obtained from anationwith a collectivist orien-tation should be interpreted differently from a society with an individ-ualistic orientation (Bierbrauer, 1994). Collectivism refers “to thetendency to bemore concerned about the consequences of one's behav-ior for in-group members and to be more willing to sacrifice personalinterests for the attainment of collective interests” (Leung, 1987: 899).All oriental and Muslim societies belong to the tradition with collectiveorientation. In contrast, most Western European and North Americannations are considered societies with more of an individualist orienta-tion (Bierbrauer, 1994; Leung, 1987) or expressive culture (Inglehart,1990). In individualistic societies, the distinction between out-groupsand in-groups is relatively unimportant. The individuals are not totallyfree from societal influence but rather are influenced by an abstract sys-tem of societal power and legal control.

Emphasis on individualism started in Protestantism states andspread to Catholicism and Orthodox Christian states (Lipset, 1994).These societies develop liberal democracy, which is “a system of gov-ernance in which those who exercise government power are subjectto the electoral control of citizens by majority vote” (Zimring,Hawkins, & Kamin, 2001: 183). In addition to periodic, multi-partyelections, liberal democracy is based on the principles of liberty andequality for all, and the rule of law (Fukuyama, 1992; Lipset, 1994).The rule of law recognizes certain individual rights or freedom fromgovernment control. In contrast, authoritarian and paternalistic soci-eties openly limit individual liberties in the interest of the commongood.

Historically, there is a difference between French model of highpolicing and English model of low policing (Brodeur, 1983; Cao &Dai, 2006; Chapman, 1970). High policing is an all absorbent policingparadigm that reaches out to potential threats in a systematic attemptto preserve the distribution of power in a given society. In the pursuitof order, high policing is ready to subdue the haughty by force. In con-trast, low policing limits its activities within the reinforcement ofcriminal law and is sensitive to public opinion (Brodeur, 1983). Thequest for order by peaceful means is the preferred style of low polic-ing. Contemporarily, low policing is dubbed as democratic policingand high policing undemocratic policing (Cao & Dai, 2006).

In a liberal democracy, the social order is maintained primarilythrough persuasion and through impartial reinforcement of law bythe police. Although the state makes its presence felt ubiquitously inthe form of police, the demand for “law and order” is a constitutionalimperative stemming from an unprecedentedly pervasive consensus.Policing by consent is the end and winning the “hearts and minds” ofresidents is the means to that end (Goldsmith, 2005). The police seekto activate popular support and constantly compete for such backing.In societies of authoritarian regimes, the police personify the statecontrol from the top. They are the repressive tool of the government,and the government is willing to use brute force to maintain a regi-me's power indefinitely. Police operation is coated in opaque and of-ficers are held accountable only to their superiors, not to the public orto the rule of law. The social order is largely maintained through dra-conian measures and the legal system is built on “repressed consen-sus,” not on consensus (Zhao & Cao, 2010). An elevated level ofconfidence in the police, therefore, is also expected from nationswith a long-term stable authoritarian government.

Page 3: Shades of blue: Confidence in the police in the world

42 L. Cao et al. / Journal of Criminal Justice 40 (2012) 40–49

As argued previously, a large body of research has been devoted toassess various factors that influence confidence in the police. The fol-lowing literature review focuses on two categories of variables: struc-tural variables and individual variables. First, we summarize findingsfrom the extant literature on effects of national/structural factors onconfidence in the police in the multi-level analyses. Second, we exam-ine the extensive research at the individual level. The advantage ofinternational study is to extend the thesis developed in one societyand to test whether it is applicable to the broader international cir-cumstances (Stack et al., 2007).

Macro level variables

Ever since Sampson and Jeglum-Bartusch's (1998) groundbreak-ing work, more research has followed the suite, paying attention tomacro characteristics and their impact on public opinions. In that re-search, Sampson and Jeglum-Bartusch (1998) found that structuralcharacteristics of neighborhoods in Chicago explained variations inresidents’ confidence in the police. Specifically, their study showsthat high concentrated disadvantage and violence crime rate signifi-cantly diminished residents’ confidence in the police, everythingelse being equal.

The replication of the study in other cities, however, did not pro-duce unequivocal results. Reisig and Parks (2000), for example,using data from Indianapolis and St. Petersburg, only found a signifi-cant effect from concentrated disadvantage, but not a significanteffect of homicide rates. Huebner et al. (2004) created a measurecalled “personal crime rate” in their multi-level modeling, and themeasure was not significantly related to either African Americans orWhites’ perception of police services. Schuck, Rosenbaum, andHawkins (2008), using Chicago data, also found a significant associa-tion between concentrated disadvantage and confidence in the police.Dai and Johnson (2009) employed neighborhood crime rates inCincinnati and found that they were not related to satisfaction withthe police while their measure of concentrated disadvantage wasnegatively related to satisfaction with the police. Wu et al. (2009),however, found a significant effect of violence crime rate but not asignificant effect of concentrated disadvantage with their data fromLouisville and Lexington: violent crime rate differentiates citizens’satisfaction with the police. Similarly, Sung (2006) discovered thatthe homicide rate dwindled confidence in the police effectivenesswhereas the effect of robbery rate was not significant. These studiesseem to suggest that crime is a prolific concept and may be measureddifferently. Residents are more concerned with homicide rates thanwith property crime or general crime. Homicide rates, therefore, aremore appropriate than general crime or property violation in thestudy of confidence in the police. For the concentrated disadvantage,its effects seem to depend on which city the data are collected andmore specifically, on the neighborhood racial composition (Wu etal., 2009).

Extending the multi-level research to the international arena, Janget al. (2010) used the data from 15,459 respondents nested in 15nations from 2000 World Values Survey and found that people livingin a country with a higher level of homicide reported lower levels ofconfidence in the police. In another comparative study, Stack et al.(2007) found a significant relationship between homicide rates andpublic attitudes toward law and order. In this study, homicide ratesare used because the definition of homicide is more consistent acrossnations than any other definitions of violent crime. Homicide isalways reported to the police (Pratt & Godsey, 2003; Stack et al.,2007) and it is viewed as a more accurate measure that is more com-parable across nations (He, Cao, Wells, & Maguire, 2003).

Although concentrated disadvantage as a measure of inequalityhas been tested extensively within the US (Dai & Johnson, 2009;Reisig & Parks, 2000; Sampson & Jeglum-Bartusch, 1998; Sampson,Raudenbush, & Earls, 1997; Schuck et al., 2008; Wu et al., 2009), the

measure with a component of percent blacks has a strong Americanflavor and is not applicable to the international arena where the con-flicts between races are either weak, or non-existing, and/or race orethnicity is not a dominant issue in those societies (Cao & Zhao,2005). In this study, the Gini Index is used as a measure of inequality.The Gini Index captures income inequality of a nation, and is the mostwidely used measure of income inequality in the cross-national re-search (He et al., 2003; Messner, 1989). Jang et al. (2010) specificallycalled for a test of the Gini Index on public confidence in the police.We hypothesize that the Gini Index is associated with confidence inthe police.

The extant studies indicate that confidence in the police is part of abroader attitudinal complex of larger legal and political systems(Albrecht & Green, 1977; Cao & Zhao, 2005; Fukuyama, 1995; Stack& Cao, 1998). In comparison of the US police and the police in LatinAmerican, Cao and Zhao (2005) found that the most importantexplanatory variable in their model of confidence in the police wastrust in the political and legal system. Extended this thesis, Ivković(2008) used the index of governance developed by the World Bankat the macro level to assess the relationship between quality of gover-nance and the levels of public support for the police, and she con-firmed that government effectiveness mattered in confidence in thepolice across the world. This thesis is tested in this study too.

The issue of regime nature is quite intricate. Jang et al. (2010) didnot find a relationship between democracy and confidence in thepolice. Their sample of fifteen nations was simply too small. Usingmore than 4,000 senior executives’ perception of police effective-ness as the dependent variable at the macro-level, Sung (2006),however, found that democracy measured by Freedom House andpolice effectiveness had a convex curvilinear relationship. Stable un-democratic countries and stable long-term advanced democracieshad the highest levels of perceived police effectiveness whereasmiddle range countries showed lower ratings of the police perfor-mance. We argue that this curvilinear relationship should alsoexist between democracy and confidence in the police because theelites’ perception of police effectiveness and public confidence inthe police should be largely consistent, capturing the similar under-lying concept. The current study, therefore, tests both the effect ofdemocracy and the curvilinear effect of democracy on confidencein the police with multiple levels of data.

Individual level variables

Most studies that examined the public confidence in the police,specifically the research that provided an in-depth exploration ofindividual level factors related to the police, have been conducted inthe United States (Ivković, 2008). A review of the comparative litera-ture revealed that prior studies have commonly used three groups ofvariables in the study of confidence in the police. These groups aresocio-demographic, experiential, and attitudinal variables. Limitedby our data and followed the footsteps of Jang et al. (2010), the cur-rent study focuses on socio-demographic and attitudinal variables toexplain confidence in the police.

First, social demographic variables concern the influences of age,gender, education, and marriage on public confidence in the police.Age, in general, is found to have a positive impact on public confi-dence in the police (Brown & Benedict, 2002; Correia et al., 1996;Dai & Johnson, 2009; Frank, Brandl, Cullen, & Stichman, 1996;Garcia & Cao, 2005; Ivković, 2008; Jang et al., 2010; Lai & Zhao,2010; Reisig & Parks, 2000; Ren et al., 2005; Sampson & Jeglum-Bartusch, 1998; Webb & Marshall, 1995; Wu et al., 2009). Agingtends to promote political conservatism and integration into the insti-tutional order (Cao et al., 1998). In contrast, younger individuals oftenperceive the police as an enforcer and restrict their freedom (Ivković,2008).

Page 4: Shades of blue: Confidence in the police in the world

43L. Cao et al. / Journal of Criminal Justice 40 (2012) 40–49

Less consistent results are found regarding the effects of gender,education, employment, and marriage. Many studies have notedthat female respondents tend to rate the police higher than theirmale counterparts (Brown & Benedict, 2002; Ivković, 2008; Lai etal., 2010; Lai & Zhao, 2010; Reisig & Parks, 2000) because maleswere somewhat more likely than females to be involved in situationsresulting in an arrest and/or the use of force (Brown & Benedict,2002). Some studies, however, do not find a significant associationbetween gender and the rating of the police (Cao et al., 1998; Cao &Dai, 2006; Cao & Zhao, 2005; Dai & Johnson, 2009; Reisig & Parks,2000; Ren et al., 2005; Schuck et al., 2008; Wu, 2010; Wu et al.,2009). Still others (Garcia & Cao, 2005; Correia et al., 1996;Sampson & Jeglum-Bartusch, 1998) report that females are less satis-fied with the police. For education, employment, andmarriage, the ef-fects are even more mixed with more studies showing no effects (Lai& Zhao, 2010). We control all these socio-demographic variables inour model.

The attitudinal variables are also found to be related to confidencein the police. Cao et al. (1998) revealed that higher scores in anomieindex reduced significantly confidence in the police whereas higherscores in satisfaction in life increased it. The finding is confirmed byJang et al.'s (2010) study. We hypothesize that anomie is negativelyrelated to confidence in the police.

In summary, using macro and micro level data from 50 nationswith 69,309 respondents (65,049 in regression analyses), the currentstudy attempts to extend our understanding of confidence in thepolice by testing the hypotheses that regime nature and incomeinequality have an impact on confidence in the police. Using hierar-chical linear modeling, we also control for the effects of homiciderates and government effectiveness at the macro level and a seriesof socio-demographic and attitudinal variables at the individual level.

Methods

The sample

The data for the current analyses are the combined results of fivesources. Data from the Gini Index for each nation in 2005 or closestto 2005 are taken from the Human Development Report 2006:Cultural Liberty in Today's Diverse World. This is an independent reportcommissioned by the United Nations Development Program (UNDP).Data of worldwide homicide rates in 2005 or closest to 2005 areretrieved from Homicide Statistics 2003–2008 collected by UnitedNations Office on Drugs and Crime (UNODC, 2009). Data on the qual-ity of governance are from Kaufman's index of governance. Data ofthe democracy index (2005) which is a measure of political rightsand civil liberties among independent countries are obtained fromFreedom House.

The individual-level data are from 2005 World Values Surveys(World Values Survey Organization, 2006). PreviousWorld Value Sur-veys have been utilized in numerous publications in more than 20languages. The survey is designed to facilitate cross-national andcross-cultural comparisons of a wide range of basic values and beliefsabout the political and social life across the globe. Each national ques-tionnaire is a translation from a standard questionnaire, which is inEnglish. The questionnaires are generally administered to about1,000 to 3,000 adult interviewees face-to-face. The samples are se-lected as close as possible following the random sampling methods.The current study uses the 2005 dataset, which is designed to givespecial attention to obtaining better coverage of non-Western socie-ties, especially Muslim nations.

Initially, the dataset contained 82,992 respondents from 57nations. For a variety of reasons, many items in the original versionhave not been used in some nations, which resulted in our droppingof these nations. Items of confidence in institutions (v131-v147) arenot included in the surveys in Egypt and Kyrgyzstan. Items in anomie

scale (v198-v210) are not used in Peru, Colombia, and Iraq. Employ-ment status is not asked in Jordan and New Zealand. After deletingabove countries, 50 out of 57 nations remain in our final sample.The sample size ranges from 1,000 cases such as in Chile, Guatemala,Poland, Ukraine, and Uruguay to 2,988 cases for South Africa (seeTable 1).

In order to maximize the number of nations in the sample, wedecide to drop some independent variables at the individual leveleven though previous studies have found them useful. For example,political conservatism is a significant predictor of confidence in thepolice, but we do not include it because China, Russia, and Iran donot have this item in their surveys.

Recognizing the fact that certain population in some countries areover- or under-represented and that the sample sizes are differentfrom one country to another, we apply the weight variable (s018)that was created in the data of the World Values Surveys to adjustfor the effect of unequal sample sizes. The weight variable incorpo-rates the national weights (for example, nationality, region, theurban/rural divide, town size, household size, sex, age, occupation,education, marital status, and economic activity) and at the sametime “gives each country the same weight to offset the fact thatsome countries used much larger samples than others, which coulddistort results when two or more samples are analyzed together”(Inglehart et al, 2004, p12). With the weight in the equation, ittakes into consideration of the national differences and transformsthe sample size to 1,000 for all countries in the study.

The dependent variable

In this study, the dependent variable, confidence in the police, ismeasured by a single item: “How much confidence do you have inthe police?” Respondents can choose from 1=a great deal,2=quite a lot, 3=not very much, 4=none at all. Although it is a sin-gle global item, it is reported that the global satisfaction is fairly stableover time, and it is highly correlated with more specific attitudes(Frank et al., 1996; Garcia & Cao, 2005). We collapsed the ordinalvariable into the dichotomized variable (1=“a great deal” or “quitea lot” and 0=“not very much” or “none at all”) for binary logisticregression analysis because a Brant test of parallel regression assump-tion needed for ordinal logistic regression was not met (Long &Freese, 2006). While we recognize that a composite scale is superiorto a single item measure when it is done within one language andwithin one culture, we believe that in the comparative literature, themore nuanced items you have, the more problematic the target lan-guage translation will be. The unaccomplished functional equivalencein translation sometimes can nullify the advantages gained by usingmultiple items. In contrast, simpler is better in comparative literaturebecause it is more straightforward and because it is easier to findtotal functional equivalence in translation. In addition, the global atti-tude was reported to be stable over the time and it was highly corre-lated with the composite index measure (Garcia & Cao, 2005). Werecoded 1 as “having confidence in the police” and 0 as “no or notmuch confidence in the police.”

The independent variables

The independent variables were grouped into two levels: individualand national levels. The six variables at the individual level were age,gender, marital status, educational attainment, employment, and ano-mie index. Age was measured in six categories: 1=15-24; 2=25-34;3=35-44; 4=45-54; 5-55-64; and 6=65 and older. Gender was abinary variable wheremale=0 and female=1. Educational attainmentwas assessed with eight orderly categories: 1=not completed elemen-tary school; 2=complete elementary education; 3=incompletesecondary school; 4=complete technical/vocational type; 5=university-preparatory type; 6=complete secondary; 7=some

Page 5: Shades of blue: Confidence in the police in the world

44 L. Cao et al. / Journal of Criminal Justice 40 (2012) 40–49

university without degrees; 8=university with degrees or higher.Marital status is one of the most important social institutions insociety and we dummy-coded the variable so that it is used tocompare with “all other non-marital statuses” (1=married, livingtogether as married, or widowed; and 0=all other non-marital sta-tuses). Employment statuses are coded into two dummy variables:employment v. unemployment, and no-working status (retired/housewife/student=1) v. unemployment. The unemployed is thereference group.

Finally, anomie was an index with four items led by the question:“Please tell me for each of the following statements whether youthink it can always be justified, never be justified, or something in be-tween”: (1) claiming government benefits which you are not entitledto; (2) avoiding a fare on public transport; (3) cheating on taxes ifyou have a chance; (4) someone accepting a bribe in the course oftheir duties. The responses range from 1=never justifiable to10=always justifiable. The index of anomie was calculated as thesum of scores of four items divided by four. A higher scores of theindex indicated that respondent would likely to accept the anomie.The Cronbach's alpha of this index was 0.79 with an Eigen value of2.46. Among individual level variables, this was the only variablemarked by a problem of skewness. It was treated with a log transfor-mation to address this problem accordingly.

There are five variables at the national level: democracy, democra-cy squared, the Gini Index, governance index, and homicide rate. Sinceour individual datawere collected in 2005, all data at the national levelare one year earlier or in the same year so that the causal order can befirmly established. Democracy was computed for each nation from theaverage score of political rights and civil liberties (e.g., rule of law) de-veloped by FreedomHouse (2005). Although there are various indica-tors of level of democracy developed by a number of scholars andinstitutions, Freedom House scores are the most frequently employedand appear to have the best validity of any indicator of democracylevel (McClintock & Lebovic, 2006; Zhang, Cao, & Vaughn, 2009). Free-domHouse assigns each nation a score (from1 to 7) for the level of de-mocracy in which 1 presents the highest level of democracy and 7represents the lowest level of democracy. For our analysis, the democ-racy index scores were reversely coded by applying a simple formula“democracy=8 - democracy score” to facilitate interpretation of theresults. Consequently, the number of 7 represents the highest levelof democracy, whereas 1 represents the lowest level of democracy.The relationship between democracy and confidence in the policecarries a complexity that cannot be adequately accounted for by sim-ple linear arguments, and following Sung (2006), a second-degreepolynomial (quadratic) function for the democracy index as democra-cy index squared was created to assess the empirical viability of thenon-linear proposition.

The Gini Index is a measure of economic inequality in a nation.It ranges from 0 to 100 with 0 representing perfect equality and100 representing perfect inequality. Quality of governance was cre-ated to tap the quality of public services in three areas: the qualityof the civil services and degree of independence from politicalpressure (GE), the quality of policy formulation and implementa-tion (RQ), and the credibility of the government's commitment tosuch policies (RL) (for further information, see Kaufman, Kraay, &Mastruzzi, 2007). The factor analysis shows that the five out ofthe original six measures loaded on one factor with an Eigenvalue of 4.43. The governance index, therefore, is created as thesum of scores of five items divided by five. The higher the score,the more positively the respondents evaluate the quality of govern-ment services in their home country (Cronbach's alpha=0.97).Data on homicide rate are used due to its relative high degree ofreliability and validity across nations (Jang et al., 2010; Stack etal., 2007). Because the distribution of homicide rates is highlyskewed, we use log transformation to correct the skewness. As aresult, homicide rate_Lg10 is used in final analyses. Both homicide

rate_ Lg10 and the Gini Index are hypothesized to have a negativeimpact on public confidence in the police.

Statistical analysis

To assess the influences of both the national level and the individ-ual level variables on confidence in the police, hierarchical linearmodeling (HLM) techniques are appropriate in this regard. Due tothe fact that the dependent variable is collapsed into a binary vari-able, we apply the Bernoulli analysis with logit link in HLM(Raudenbush & Bryk, 2002). The model of the current study can beexpressed in the following equations:

Individual Level Level� 1 Modelð ÞProb Having confidence in the police ¼ 1jβð Þ ¼ φij

Log φij= 1−φij

� �h i¼ β 0j þ β 1j Ageð Þij þ β 2j Genderð Þij þ β 3j Marriageð Þijþβ 4j Educationð Þij þ β 5j Employmentð Þijþβ 6j Anomie Lg10ð Þij

where age, education, and anomie scores are grand-mean centeredand the rest of dummy-coded variables are not centered. Thiswould facilitate the interpretation of results.1

At the national level, β0j is modeled as a function of the individuallevel predictors. We view the other individual level coefficients, β pj,p>0, as fixed.

National Level ðLevel� 2 ModelÞβ 0j ¼ γ00 þ γ01 Gini Indexð Þj þ γ02 Homicide rate Lg10ð Þj þ γ03 Governance indexð Þj

þγ04 Democracy indexð Þj þ γ05 Democracy index squaredð Þj þ uojβ pj ¼ γp0

This constitutes what Raudenbush and Bryk suggested as HGLM“conditional model” with link function at individual level (seeRaudenbush & Bryk, 2002, p298).2 Overall, the model tested boththe effects of individual level variables and the national level variableson confidence in the police.

Findings

Descriptive statistics for the variables used in the analysis areshowed in Tables 1 and 2, followed by the results from HGLM logisticregression analyses in Table 3. In Table 1, the reversed scores of thedemocracy index are reported. Nations, such as Andorra, Australia,Canada, etc. earn the highest possible score while China, Vietnamand Iran receive the lowest possible scores.

The mean scores of confidence in the police for each nation in2005 are presented. The original four-point scale is reversed so thatthe higher score presents a higher confidence in the police (1=notat all and 4=a great deal). The average mean of confidence in thepolice is 2.63 with a standard deviation of 0.90, suggesting thatthere is a wide range of variation in this dependent variable among50 nations. Among those nations, Vietnam has the highest mean ofconfidence in the police (3.54), followed by Rwanda (3.27), andFinland (3.25). The mean score of confidence in the police in theUnited States is 2.87, which is higher than the average mean of thisgroup. Nations with the lowest levels of confidence in the police areArgentina (1.89), Moldova (1.97), and Guatemala (2.00). An ANOVAis conducted to see if there is significant difference among these na-tions, and the results indicated that there is a statistically significantdifference among these nations (between groups) than within a na-tion (F=259.01, pb .001).

Table 2 presents descriptive statistics of variables at levels 1 and 2.The dichotomized confidence in the police shows that almost sixtypercent respondents have positive attitudes toward the police. Forthe independent variables at the individual level, the mean of respon-dents’ age is 3.12 at the range of 35 to 44 years old and about 48 per-cent of the respondents are male. Approximately sixty nine percent

Page 6: Shades of blue: Confidence in the police in the world

Table 22005 Participated Nations Descriptive Statistics

Variables Mean S.D. Min. Max.

Dependent VariableConfidence in the Police 0.58 0.49 0.00 1.00

Negative attitudes: 28,174 (41.7%)Positive attitudes: 39,415 (58.3%)

Individual Level (N=69,309)Age 3.19 1.61 1.00 6.00Gender (1=Female) 0.52 0.50 0.00 1.00Marital status (1=Married) 0.69 0.46 0.00 1.00

Others: 21,592 (31.3%)Married: 47,500 (68.7%)

Educational attainment 4.33 2.35 1.00 8.00Employed 0.53 0.50 0.00 1.00Retired/housewife/student 0.33 0.47 0.00 1.00Unemployed 0.10 0.30 0.00 1.00Anomie_Lg10 0.28 0.28 1.00 1.00

National Level (N=50)Gini coefficient 37.95 9.12 23.00 65.10Homicide rates_Lg10 0.50 0.51 −0.30 1.62Governance index 0.35 0.85 −1.05 1.87Democracy index 5.54 1.67 1.50 7.00Democracy index square 33.42 16.04 2.25 49.00

Note: Democracy index is inverted (by formula: X=8- original democracy scoreobtained from Freedom House) in statistical analyses to facilitate interpretation ofresults.

Table 1Mean Rating of Confidence in the police: One Way ANOVA Test (N=50)

Nation/Region Democracy index Mean of confidence S.D. N (Valid % )

Andorra 7 2.59 0.79 1,003(1.4)Argentina 6 1.89 0.76 1,002(1.4)Australia 7 3.04 0.70 1,421(2.1)Brazil 5.5 2.28 0.92 1,500(2.2)Britain 7 2.85 0.80 1,041(1.5)Bulgaria 6.5 2.54 0.92 1,001(1.4)Burkina Faso 3.5 2.52 0.95 1,534(2.2)Canada 7 3.06 0.73 2,164(3.1)Chile 7 2.62 0.87 1,000(1.4)China 1.5 3.01 0.72 2,015(2.9)Cyprus 7 2.88 0.92 1,050(1.5)Ethiopia 3 2.21 0.87 1,500(2.2)Finland 7 3.25 0.65 1,014(1.5)France 7 2.76 0.82 1,001(1.4)Georgia 4.5 2.40 0.81 1,500(2.2)Germany 7 2.83 0.73 2,064(3.0)Ghana 6 2.59 0.99 1,534(2.2)Guatemala 4 2.00 0.86 1,000(1.4)India 5.5 2.79 1.00 2,001(2.9)Indonesia 4.5 2.50 0.77 2,015(2.9)Iran 2 2.72 0.86 2,667(3.8)Italy 7 2.90 0.66 1,012(1.5)Japan 6.5 2.72 0.68 1,096(1.6)Malaysia 4 2.91 0.77 1,201(1.7)Mali 6 2.91 0.98 1,534(2.2)Mexico 6 2.09 0.91 1,560(2.3)Moldova 4.5 1.97 0.85 1,046(1.5)Morocco 3.5 2.75 0.97 1,200(1.7)Netherlands 7 2.58 0.71 1,050(1.5)Norway 7 3.05 0.60 1,025(1.5)Poland 7 2.44 0.78 1,000(1.4)Romania 5.5 2.29 0.84 1,776(2.6)Russia 2.5 2.11 0.88 2,033(2.9)Rwanda 2.5 3.27 0.73 1,507(2.2)S. Africa 6.5 2.63 0.93 2,988(4.3)S. Korea 6.5 2.59 0.70 1,200(1.7)Serbia 5.5 2.16 0.82 1,220(1.8)Slovenia 7 2.30 0.76 1,037(1.5)Spain 7 2.65 0.74 1,200(1.7)Sweden 7 2.87 0.63 1,003(1.4)Switzerland 7 2.98 0.61 1,241(1.8)Taiwan 6.5 2.29 0.75 1,227(1.8)Thailand 5.5 2.39 0.81 1,534(2.2)Trinidad and Tobago 5 2.13 0.81 1,002(1.4)Turkey 5 2.97 0.99 1,346(1.9)Ukraine 4.5 2.14 0.84 1,000(1.4)Uruguay 7 2.51 0.95 1,000(1.4)USA 7 2.87 0.72 1,249(1.8)Vietnam 1.5 3.54 0.66 1,495(2.2)Zambia 4 2.48 0.94 1,500(2.2)All nations 2.63 0.90 69,309

Note: 1.Mean square between group=175.73 and within group=0.678, F=259.01,pb .001.

45L. Cao et al. / Journal of Criminal Justice 40 (2012) 40–49

respondents are married or living together in these nations. Theaverage educational attainment is 4.3, indicating that respondentscompleted secondary school (technical /vocational type). Status ofemployment is grouped into three categories (dummy variables):the employed (53%), retired/housewife/student (33%), and the unem-ployed (10%). Finally, the mean of anomie_Lg10 is 0.28 with a stan-dard deviation of 0.28.

For the independent variables at the national level, the mean de-mocracy score of all fifty nations is 5.54 with a standard deviation of1.67 whereas the mean democracy squared is 33.42 with a standarddeviation of 16.04. The average Gini Index is 37.95 with a standarddeviation of 9.12. The mean of governance index is 0.35 with a stan-dard deviation of 0.85. The average homicide rate for fifty nations is6.56 per 100,000. This variable is positively skewed (2.40) with kurto-sis score of 5.39. A Q-Q Plot also showed a serious skewness, suggest-ing the necessity of data transformation (Mertler & Vannatta, 2005).

With logarithm (Lg10), the mean of homicide rate_Lg10 was 0.50with a standard deviation of 0.51. The new variable's skewness(0.54) and kurtosis (− 0.43) are both within the acceptable range.

The correlation between the governance index and the democracyindex are very high (r>0.7), which is a sign of multicollinearity.Entering both variables in the same model suppresses the effect ofone of the two variables. Accordingly, Model 3 and Model 4 are creat-ed to show the effects of the governance index and the democracyindex, respectively3 (see Table 3).

Table 3 presents the results of the HGLM- binary logistic analysesof confidence in the police. First, Model 1 shows the results from theunconditional model. This base model is estimated without any indi-vidual or national level predictors in the equation and is useful for es-timating the average log-odds of confidence in the police acrossnations (γ00) and for assessing the magnitude of variation betweennations in confidence in the police (τ00) (Raudenbush & Bryk,2002). The average log-odds of public confidence in the police acrossnations is 0.292 (SE=0.12). This translates into an odds ratio of1.339, suggesting that, for a nation with a “typical” confidence level,the expected log-odds of public confidence in the police is 1.339.The variance in nation average log-odds of confidence in the policeis 0.887 (X2=9227.291, df=49, pb .001). Knowledge of the variancebetween nations in national average log-odds of confidence in thepolice also made it possible to calculate the intra-class correction.This statistic represents the proportion of the variance in thebetween-group outcome.4 The intra-class correlation of the presentstudy is 0.212, which is quite substantial for a HLM model. It suggeststhat 21.2 percent of the variation in the odds of individual confidencein the police is explained by the national level factors. Importantly,this also suggests that the odds of confidence in the police vary signif-icantly across nations.

Model 2 reports the influences of individual level variables on con-fidence in the police. All of the variables have significant effects onconfidence in the police. The elderly, female, the married, theemployed, and the group of retired/housewife/student tend to in-crease the odds of confidence in the police but those with a higherscore of anomie tend to have significantly lower probability of confi-dence in the police. The odds ratios reported in Model 2 indicate that,

Page 7: Shades of blue: Confidence in the police in the world

Table 3Determinants of public confidence in the police (nation N=50; individual N=65,049)

Fixed effects Model 1 Model 2 Model 3 Model 4

Baseline Individual Variables Multilevel Variables Multilevel Variables

Coeff. SE Odds ratio Coeff. SE Odds ratio Coeff. SE Odds ratio Coeff. SE Odds ratio

Intercept 0.292 0.122 1.339* 0.052 0.117 1.054 0.039 0.114 1.140 0.057 0.101 1.059National LevelGini index 0.017 0.014 1.017 0.010 0.013 1.010Homicide rate_Lg10 −0.829 0.277 0.436** −0.698 0.217 0.498**Governance index 0.409 0.163 1.505* ————— ————— —————

Democracy index ———— ———— ———— −1.450 0.361 0.235***Democracy index squared ———— ———— ———— 0.153 0.037 1.165***

Individual LevelAge 0.031 0.010 1.032** 0.037 0.012 1.037** 0.034 0.011 1.035**Gender (1=female) 0.071 0.022 1.074** 0.083 0.026 1.086** 0.076 0.024 1.079**Marital status (1=married) 0.135 0.035 1.145*** 0.159 0.041 1.172*** 0.144 0.037 1.155***Educational Attainment −0.015 0.008 0.985 −0.017 0.010 0.983 −0.015 0.009 0.985Employed 0.103 0.034 1.109** 0.116 0.040 1.123** 0.106 0.038 1.112**Retired/housewife/student 0.173 0.042 1.188*** 0.197 0.049 1.218*** 0.181 0.046 1.198***Anomie_Lg10 −0.263 0.076 0.769** −0.309 0.087 0.734** −0.281 0.082 0.755**Random effectsVariance component u0 0.887 0.873 0.553 0.512

R 0.996 0.995 0.993 0.992X2 9227.291*** 8716.94*** 5180.701*** 5339.210***Explained Variance (%) ——— 1.6 36.7 41.4

Note: Only the unit-specific results with robust standard errors are reported.*pb .05, **pb .01, *** pb .001.

0

0.25

0.50

0.75

1.00

Con

fiden

ce in

the

Pol

ice

1.50 2.86 4.22 5.58 6.95

Democracy Index

Note: All other independent variables are held constant at meanThe nadir point X=-Bdemocracy index/ 2Bdemocracy index squared

The nadir point X=4.74

Fig. 1. U-shape Relationship between Democracy and Confidence in the Police.

46 L. Cao et al. / Journal of Criminal Justice 40 (2012) 40–49

among individual variables, anomie_Lg10 appears to be the strongestvariable to have a significant effect on the log odds of confidence inthe police, followed by retired/housewife/student status, marital sta-tus, employed, gender, and age. The anomie level is associated with23 percent ([1–0.769]×100) decrease in the odds of havingconfidence in the police. Being a retired/ housewife/student, theemployed, the older, female, and the married have higher probabilityof having confidence in the police (as compared to no or low confi-dence in the police) than the unemployed, younger, male, and the un-married. Overall, the variance component in this model is 0.873(X2=8716.94, df=49, pb .001), indicating that Model 2 explainsapproximately 2 percent of variance in confidence in the police acrossnations.

Compared with Model 2, Model 3 contains three national levelindependent variables, the Gini Index, the governance index, andhomicide rate_Lg10. Results show that the effects of all predictors atthe individual level remain largely consistent with results in Model2. Among the three national level variables, both governance index(coeff.=0.409) and homicide_Lg10 rate (coeff.=−0.829) have sig-nificant impacts on confidence in the police. As hypothesized, higherhomicide rates tend to diminish the odds that the respondent havingconfidence in the police whereas satisfaction with government effec-tiveness tend to boost the odds of having confidence in the police (ascompared to no or low confidence in the police). The Gini Index, how-ever, has no significant influence on confidence in the police. Theodds ratios reported in Model 3 also indicate that, among themacro-level variables, the effect of homicide_Lg10 is the strongest:as the effect of homicide rate increases by one unit, a respondent'slog odds of confidence in the police decreases significantly by 56percent([1–0.436]×100). At the same time, as the effect of gover-nance index increases by one unit, a respondent has 51 percent([1.505-1]×100) greater odds of confidence in the police. The vari-ance component drops from 0.873 in Model 2 to 0.553 in Model 3(X2=5180.70, df=46, pb .001), indicating that the full multilevelmodel explained approximately 37 percent of variance in confidencein the police across nations.

Finally, Model 4 introduces both democracy and the quadraticterm of the democracy index to the multilevel analysis and removesgovernance index. Compared with model 2 (the individual level

model) the variance components at the national level dropped signif-icantly from 0.873 in model 2 to 0.512 in Model 4(X2=5339.21,df=45, pb .001), and Model 4 explained approximately 41 percentof variance in confidence in the police across nations, which signifi-cantly improved the explained variance in Model 3. Results showthat all the effects of the individual level independent variables re-main consistent with results in Model 3. At the national level, homi-cide rate_Lg10 remains a significant predictor of confidence in thepolice (coeff.=−.698). With all other variables being held constant,both democracy (coeff.=−1.450) and the quadratic term of democ-racy (coeff.=0.153) have significant impacts on confidence in thepolice. The positive sign of the regression coefficient of the quadraticterm indicates that the function was a convex parabola (see Fig. 1).The probability of individuals having confidence in the police ishigher for authoritarian societies and for advanced democracies, butlower for unstable or transitional societies in between. The nadirpoint (of the convex parabola) that changed the direction of theeffect was estimated by applying the formula X=−Bdemocracy index /

Page 8: Shades of blue: Confidence in the police in the world

47L. Cao et al. / Journal of Criminal Justice 40 (2012) 40–49

2Bdemocracy index squared (Agresti & Finlay, 1997). The observed X (4.74)on the democracy axis is the nadir point at which the probability thatindividuals having confidence in the police is at its lowest level (seeFig. 1).

It is therefore possible to locate the relative position of each coun-try in relation to the two axes by examining this convex curvilinearfunction (Sung, 2006). Some stable authoritarian countries, such asChina, Vietnam, and Turkey, have very high level of confidence inthe police. Many transitional and/or unstable regimes, such as Argen-tina, Moldova, and Ukraine, experience crises in confidence in thepolice. This situation, however, may be ameliorated by the deepeningof democratic development. More advanced, stable and long-termdemocracies, such as Australia, Canada, and Finland, enjoy a higherlevel of confidence in the police.

Discussion and conclusion

Comparative study is difficult and challenging because nationshave noticeable differences culturally, politically, and religiously.Many times, these differences work intertwinedly, making it hard todisentangle them. Comparative study is important because such stud-ies can test the generalizability of theories developed in one society orin one region (Stack et al., 2007). On the surface, the police across theworld all wear blue and all have a similar task of preventing crimeand maintaining order. Beneath the superficial similarity, the shadesof blue differ and police differ fundamentally in their organizationalpriorities and in their means to achieve the priorities. By employinga multi-level data analysis design, the current study attempts to fillthe gap in the literature of cross-national citizens’ confidence in thepolice by focusing on the nature of regime. Using appropriate tech-nique of analysis (HGLM), our results show that indeed the regimenature has a significant impact on public opinion of the police andthe model with democracy and democracy square is the best in pre-dicting confidence in the police.

This is an important finding. Building on the insight of Almond andVerba (1963), Goldsmith (2005), and Sung (2006), the present studyconfirms that both democracy and authoritarianism leave a clear im-print in public opinions of the police. The findings provide a piece ofempirical evidence about determinants of confidence in the police,and particularly with respect of national differences in levels of publicconfidence. Our results show the level of confidence in the long-termand stable authoritarian states are similar to that of the long-term andstable democracies while unstable regimes, regardless of their gov-ernment form, have a lower level of confidence in the police. The rea-son for expressed higher level of confidence in the police in long-termand stable authoritarian states, however, is different from that inlong-term and stable democracies. Residents in long-term and stableauthoritarian states have mass media filtering of negative news aboutthe police, they have collective orientation cultures, and they areafraid of unpredictable hassles of capricious persecution. The confla-tion of these factors creates a different social reality for citizens ofthese nations, which colors their opinions about the police. Oncethese nations move away from authoritarianism and toward democ-racy, confidence in the police in these nations is expected to plunge.Under unstable regimes of both authoritarian and democracy, confi-dence is low. Eventually, as a thin blue line, police provide crucial ser-vices as it delivers authoritative intervention in crisis situations andsymbolic justice after the violation of behavioral norms (Bayley,1994; Manning, 1990; Sung, 2006). Confidence in the police wouldincrease gradually with time under democracy and the rule of law.Confidence in the police, therefore, is curvilinearly related to regimenature: confidence is artificially higher under stable authoritarian re-gime, lower under unstable regimes and governments in democratictransition, and higher again with a stable democracy. Confidenceeven in a long-term stable democracy, however, should not reach as

high as that of a long-term stable authoritarian regime because dem-ocratic culture creates “critical citizens” (Norris, 1999).

Contrary to our expectation, income inequality does not explainthe variance of confidence in the police in systematic ways. It is pos-sible that residents in many traditional and/or under-developed soci-eties do not feel strongly about income inequality because theyregard it as natural and inevitable human conditions (Durkheim,1984). Even though measured objectively, the perception of inequal-ity itself is symbolic and interactional in nature. More detailed theo-rizing and empirical verification is needed since ours is the firsttime that this thesis is tested with international data.

Consistent with the existing literature (Fukuyama, 1995; Ivković,2008), the government effectiveness is found to enhance confidencein the police, independent of other explanatory variables at nationallevels. This finding confirms that confidence in the police is part ofthe broader attitudinal complex of larger legal and political systems(Stack & Cao, 1998; Cao & Zhao, 2005) at both individual and nationallevels. In addition, higher homicide rates diminish confidence in thepolice, which is consistent with studies within the US (Sampson &Jeglum-Bartusch, 1998; Wu et al., 2009), and with comparative re-search findings at the international level (Jang et al., 2010; Stack etal., 2007). It is perhaps universal that residents hold the police atleast partially responsible for the control of crime. Contrasting 50nations grants greater generalizability to the findings and allows sta-tistical controls to avoid spurious relationships. Our results of govern-ment effectiveness and homicide rates are thus more robust than theprevious findings.

The findings from individual-level factors are largely consistentwith the existing literature. The elderly have higher confidence inthe police (Cao & Stack, 2005; Cao et al., 1998; Correia et al., 1996;Frank et al., 1996; Ivković, 2008; Jang et al., 2010; Lai & Zhao, 2010;Ren et al., 2005; Sampson & Jeglum-Bartusch, 1998). In addition,females (Cao et al., 1998; Correia et al., 1996; Ivković, 2008; Lai &Zhao, 2010; Reisig & Giacomazzi, 1998), the married (Cao & Zhao,2005), the employed (Cao & Zhao, 2005), the retired/housewife/student (Cao, 2001), and those who have lower levels of anomie(Cao et al., 1998; Jang et al., 2010) report higher levels of confidencein the police. Overall, these individual characteristics explain confi-dence in the police, but their impacts are small comparing with con-textual variables at the national/structural level.

This study as a whole has some interesting policy implications.First, the quality of governance is a robust factor in shaping respon-dents’ confidence in the police. Apparently, it is impossible to evalu-ate one part of the government while disregarding other segmentsof the same government. Citizens’ satisfaction with the police is close-ly related to their general perceptions of government effectiveness.Regardless of the nature of the regime, the government, therefore,should make more efforts to promote the governmental efficiencyto win citizens’ support (Ivković, 2008; Cao & Zhao, 2005; Zhao,Scheider, & Thurman, 2002). Second, programs of police agenciesthat attempt to reduce violent crime, inter alia, will make publicview the police more favorably because policing is seen as one facetof a culture broadly construed within sociocultural, historical, and in-stitutional environment (Loader & Walker, 2001; Manning, 1990).Strategies encouraging police officers to pay close attention to “qual-ity of life” issues as well as accountability, tend to boost confidence inthe police (Gibson, Zhao, Lovrich, & Gaffney, 2002; Reisig & Parks,2000; Ren et al., 2005; Skogan, 2009).

Along with its strengths, there are limitations of this study. Thenations in the current sample are not randomly selected, and inturn, they do not fully represent the entire world even though re-spondents in these samples are randomly selected. The dependentvariable is a single item tapping the global confidence of the policeand it may be too simplistic (Hurst & Frank, 2000). Some variablesthat are relevant to confidence in the police are not included. At theindividual level, experiential variables, such as contacts with police

Page 9: Shades of blue: Confidence in the police in the world

48 L. Cao et al. / Journal of Criminal Justice 40 (2012) 40–49

and victimization, are found to be significantly related to attitudes to-ward the police (Dai & Johnson, 2009; Ren et al., 2005; Reisig &Giacomazzi, 1998; Wu, 2010; Wu et al., 2009). At the national level,police size and percent of women in the police (Ivković, 2008) arenot controlled for. The cross-sectional design undermines the estab-lishment of the temporal order among analyzed factors intrinsic incausal inference. Despite the limitations, the study has providedclear evidence that regime nature as a contextual factor affects citi-zens’ confidence in the police. This finding alone significantly extendsour understanding of trust in general (Fukuyama, 1995; Goldsmith,2005) and confidence in the police in particular. Measuring and un-derstanding regime nature is pivotal to any analysis of the shades ofblue in the world. It is a social fact for which all future studies mustaccount.

Notes

1. When a value of zero is not meaningful for an independent variable, the re-searcher may want choose a proper centering method for the variable (Raudenbush& Bryk, 2002). A criterion for the choice among three different centering methods(grand-mean centering, group-mean centering, or centering at some theoretically cho-sen location) is the ease of interpretation of results. In the present study, the dummyvariables are not centered and all ordinal or interval variables are centered aroundthe grand mean(Xij-X.̅.).

2. Since our study is exploratory, it does not involve further cross-level interac-tional effects in the HGLM analysis.

3. When multicolinearity is concerned, it is often suggested not to correct the vari-ables if they turned out to measure different things in the model. Every remedy formulticollinearity has a drawback of some sort (Studenmund, 2006). As the governanceindex is highly correlated with the democracy index and placing both variables in themodel would suppress the effect of one of the two variables, model 3 and model 4 arebuilt to show the effect of governance index and democracy index respectively.

4. When a researcher used a HLM with non-linear link functions, the inter-classcorrelation traditionally has not been reported (i.e., the ratio of level 2 variance tothe total variation) because the level 1 variance for these functions is heteroscedastic(Raudenbush & Bryk, 2002; Wu et al., 2009). Snijders and Bosker (1999) provide a for-mula for calculating the intra-class correction when using a logit link. This formula isp=τ00 / ( τ00+π2/3) (see also, Altheimer, 2008; Kang, 2011).

References

Agresti, A., & Finlay, B. (1997). Statistical methods for the social sciences (3rd ed.). UpperSaddle River, NJ: Prentice-Hall.

Albrecht, S. L., & Green, M. (1977). Attitudes toward the police and the larger attitudescomplex. Criminology, 15, 67–86.

Almond, G. A., & Verba, S. (1963). The civic culture: Political attitudes and democracy infive nations. Princeton, NJ: Princeton University Press.

Altheimer, I. (2008). Do guns matter? A multi-level cross-national examination of gunavailability on assault and robbery victimization. Western Criminology Review, 9,9–32.

Baumer, E. P., Messner, S. F., & Rosenfeld, R. (2003). Explaining spatial variation in sup-port for capital punishment: A multi-level analysis. American Journal of Sociology,108, 844–875.

Bayley, D. H. (1994). Police for the future. New York, NY: Oxford University Press.Bierbrauer, G. (1994). Toward an understanding of legal culture: Variations in individ-

ualism and collectivism between Kurds, Lebanese, and Germans. Law and SocietyReview, 28, 243–264.

Blau, J. R., & Blau, P. M. (1982). The cost of inequality: Metropolitan structure and vio-lent crime. American Sociological Review, 47, 114–129.

Blumer, H. (1958). Race prejudice as a sense of group position. Pacific SociologicalReview, 1, 3–7.

Brodeur, J. P. (1983). High policing and low policing: Remarks about the policing ofpolitical activities. Social Problems, 30, 507–520.

Brown, B., & Benedict, W. (2002). Perceptions of the police: Past findings, methodolog-ical issues, conceptual issues and police implications. Policing, 25, 543–580.

Cao, L. (2001). A problem in no-problem-policing in Germany: Confidence in the policeGermany and USA. European Journal of Crime, Criminal Law, and Criminal Justice, 9,167–179.

Cao, L., & Burton, V., Jr. (2006). Spanning the continents: Assessing the Turkish publicconfidence in the police. Policing, 29, 451–463.

Cao, L., & Dai, M. (2006). Confidence in the police: Where does Taiwan rank in theworld? Asian Journal of Criminology, 1, 71–84.

Cao, L., & Stack, S. (2005). Confidence in the police between America and Japan: Resultsfrom two waves of surveys. Policing, 28(1), 139–151.

Cao, L., Stack, S., & Sun, Y. (1998). Public confidence in the police: A comparative studybetween Japan and America. Journal of Criminal Justice, 26, 279–289.

Cao, L., & Zhao, J. S. (2005). Confidence in the police in Latin America. Journal of CriminalJustice, 33, 403–412.

Chapman, B. (1970). Police State. New York, NY: Praeger Publishers.Correia, M., Reisig, M., & Lovrich, N. (1996). Public perceptions of state police: An anal-

ysis of individual-level and contextual variables. Journal of Criminal Justice, 24,17–28.

Dai, M., & Johnson, R. R. (2009). Is neighborhood context a confounder? Policing, 32,595–612.

Durkheim, E. (1984). The division of labor in society. New York: The Free Press.Eck, J. E., & Rosenbaum, D. P. (1994). The new police order: Effectiveness, equity, and

efficiency. In D. Rosenbaum (Ed.), The challenge of community policing: Testing thepromises (pp. 3–23). Thousand Oaks, CA: Sage.

Frank, J., Brandl, S. T., Cullen, F. T., & Stichman, A. (1996). Reassessing the impact of raceon citizens’ attitudes toward the police: A research note. Justice Quarterly, 13,321–334.

Freedom House (2005). Freedom in the world: Map of freedom in the world: Tablesand charts: Combined average ratings 2005. Retrieved October 14, 2010, fromhttp://www.freedomhouse.org/template.cfm?page=25&year=2005

Fukuyama, F. (1992). The end of history and the last man. New York: The Free Press.Fukuyama, F. (1995). Trust: The social virtues and the creation of prosperity. London:

Penguin.Garcia, V., & Cao, L. (2005). Race and satisfaction with the police in a small city. Journal

of Criminal Justice, 33, 191–199.Gibson, C., Zhao, J. S., Lovrich, N., & Gaffney, M. (2002). Social integration, individual

perceptions of collective efficacy, and fear of crime in three cities. Justice Quarterly,19, 537–564.

Goldsmith, A. (2005). Police reform and the problem of trust. Theoretical Criminology, 9,443–470.

He, N., Cao, L., Wells, W., & Maguire, E. R. (2003). Forces of production and direction: Anexpanded model of suicide and homicide. Homicide Studies, 7, 36–57.

Huebner, B., Schafer, J., & Bynum, T. (2004). African American and Whites perceptionsof police services: Within- and between-group variation. Journal of Criminal Justice,32, 125–135.

Human Development Report (2006). Beyond scarcity: Power, poverty and the globalwater crisis. Retrieved April 14, 2010, from http://hdr.undp.org/en/reports/global/hdr2006/

Huntington, S. P. (1991). The third wave: Democratization in the later twentieth century.Norman, OK: University of Oklahoma Press.

Hurst, Y. G., & Frank, J. (2000). How kids view cops: The nature of juvenile attitudestoward the police. Journal of Criminal Justice, 28, 189–203.

Inglehart, R. (1990). Culture shift in advanced industrial society. Princeton, New Jersey:Princeton University Press.

Inglehart, R., Basanez, M., Diez-Medrano, J., Halman, L., & Luijkx, R. (2004).World valuessurveys and European values surveys, 1999–2001: User guide and codebook (ICPSR3975). Retrieved March 30, 2011, from http://prod.library.utoronto.ca/datalib/codebooks/icpsr/3975/cb3975.pdf

Ivković, S. K. (2008). A comparative study of public support for the police. InternationalCriminal Justice Review, 18, 406–434.

Jacobs, D. (1979). Inequality and police strength: Conflict theory and coercive controlin metropolitan area. American Sociological Review, 44, 913–925.

Jang, H., Joo, H. -J., & Zhao, J. S. (2010). Determinants of public confidence in police: Aninternational perspective. Journal of Criminal Justice, 38, 57–68.

Kang, J. H. (2011). Participation in the community social control, the neighborhoodwatch groups: Individual- and neighborhood- related factors. Crime and Delinquency,doi:10.1177/0011128711398024.

Kaufman, D., Kraay, A., & Mastruzzi, M. (2007). Governance matters VI: Aggregate andindividual governance indicators 1996–2006. Retrieved April 14, 2010, fromhttp://info.worldbank.org/governance/wgi/pdf/governmatters6.pdf

Lai, Y. -L., Cao, L., & Zhao, J. S. (2010). The impact of political entity on confidence inlegal authorities: A comparison between China and Taiwan. Journal of Criminal Jus-tice, 38, 934–941.

Lai, Y. -L., & Zhao, J. S. (2010). The impact of race/ethnicity, neighborhood context, andpolice/citizen interaction on residents’ attitudes toward the police. Journal of Crim-inal Justice, 38, 685–692.

Leung, K. (1987). Some determinants of reactions to procedural model for conflict res-olution: A cross-national study. Journal of Personality and Social Psychology, 53,898–908.

Lipset, S. M. (1994). The social requisites of democracy revisited. American SociologicalReview, 59, 1–22.

Loader, I., & Walker, N. (2001). Policing as a public good: Reconstructing the connec-tions between policing the state. Theoretical Criminology, 5(1), 9–35.

Long, S., & Freese, J. (2006). Regression models for categorical dependent variables usingstata (2nd ed.). College Station, TX: Strata Press.

Manning, P. (1990). Symbolic communication. Cambridge, MA: MIT Press.McClintock, C., & Lebovic, J. (2006). Correlates of levels of democracy in Latin America

during the 1990s. Latin American Politics and Society, 48, 29–59.Mertler, C., & Vannatta, R. (2005). Advanced and multivariate statistical methods: Practi-

cal application and interpretation (3rd ed.). Glendale, CA: Pyrczak Publishing.Messner, S. (1989). Economic discrimination and societal homicide rates: Future evi-

dence on the cost of inequality. American Sociological Review, 54, 597–611.Norris, P. (1999). Critical Citizens: Global support for democratic governments. Oxford:

Oxford University Press.Pratt, T., & Godsey, T. (2003). Social support, inequality, and homicide: A cross-national

test of an integrated theoretical model. Criminology, 41, 611–643.Raudenbush, S. W., & Bryk, A. S. (2002). Hierarchical linear models: Applications and data

analysis methods (2nd ed.). Thousand Oak, CA: Sage.Reisig, M. D., & Giacomazzi, A. L. (1998). Citizen perceptions of community policing:

Are attitudes toward police important? Policing, 21, 547–561.

Page 10: Shades of blue: Confidence in the police in the world

49L. Cao et al. / Journal of Criminal Justice 40 (2012) 40–49

Reisig, M. D., & Parks, R. B. (2000). Experience, quality of life, and neighborhood con-text: A hierarchical analysis of satisfaction with police. Justice Quarterly, 17,607–630.

Ren, L., Cao, L., Lovrich, N., & Gaffney, M. (2005). Linking confidence in the police withthe performance of the police: Community policing can make a difference. Journalof Criminal Justice, 33, 55–66.

Sampson, R., & Jeglum-Bartusch, D. (1998). Legal cynicism and (subcultural?) toleranceof deviance: The neighborhood context of racial differences. Law and Society Re-view, 32, 777–802.

Sampson, R. J., Raudenbush, S. W., & Earls, F. (1997). Neighborhoods and violent crime:A multilevel study of collective efficacy. Science, 277, 918–924.

Schuck, A. M., Rosenbaum, D. P., & Hawkins, D. F. (2008). The influence of race/ethnicity,social class, and neighborhood context on residents’ attitudes toward the police. Po-lice Quarterly, 11, 496–519.

Skogan, W. (2009). Concern about crime and confidence in the police: Reassurance oraccountability? Police Quarterly, 12, 301–318.

Skolnick, J. (1994). Justice without trial. New York, NY: Macmillan.Snijders, T. A. B., & Bosker, R. J. (1999). Multilevel analysis: An introduction to basic and

advanced multilevel modeling. Thousand Oaks, CA: Sage Publications.Stack, S., & Cao, L. (1998). Political conservatism and confidence in the police: A com-

parative analysis. Journal of Crime and Justice, 21, 71–76.Stack, S., Cao, L., & Adamzyck, A. (2007). Crime volume and law and order culture. Jus-

tice Quarterly, 24, 291–308.Studenmund, A. H. (2006). Using econometrics: A practical guide (5th ed.). NY: Pearson

Education, Inc.Sung, H. -E. (2006). Police effectiveness and democracy: Shape and direction of the

relationship. Policing, 29, 347–367.Tyler, T., & Weber, R. (1982). Support for the death penalty: Instrumental response to

crime or symbolic attitude? Law and Society Review, 17, 21–45.United Nations Office on Drugs and Crime (UNODC) (2009). Homicide statistics, crimi-

nal justice sources-Latest available year (2003–2008). Retrieved February 6, 2011,from http://www.unodc.org/unodc/en/data-and-analysis/homicide.html

Webb, V., & Marshall, C. E. (1995). The relative importance of race and ethnicity on cit-izen attitudes toward the police. American Journal of Police, XIV, 45–66.

Weitzer, R., & Tuch, S. A. (2006). Race and policing in America: Conflict and reform. NewYork: Cambridge University Press.

World Values Survey Organization (2006). Introduction to the World Values Survey.Retrieved April 14, 2010, from. http://www.worldvaulessurvey.org

Wu, Y. (2010). Evaluation of police performance: An empirical comparison betweenChinese and Americans. Journal of Criminal Justice, 38, 773–780.

Wu, Y., Sun, I. Y., & Triplett, R. A. (2009). Race, class or neighborhood context: Whichmatters more in measuring satisfaction with police? Justice Quarterly, 26, 125–156.

Zhang, Y., Cao, L., & Vaughn, M. S. (2009). Social support and corruption: Structuraldeterminants of corruption in the world. The Australian and New Zealand Journalof Criminology, 42, 204–217.

Zhao, R., & Cao, L. (2010). Social change and anomie: A cross-national study. SocialForces, 88, 1209–1230.

Zhao, J. S., Scheider, M., & Thurman, Q. (2002). The effect of police presence on publicfear reduction and satisfaction: A review of the literature. The Justice Professional,15, 273–299.

Zimring, F. E., Hawkins, G., & Kamin, S. (2001). Punishment and democracy: Three strikesand you're out in California. Oxford: Oxford University Press.

LiInco

qun Cao is Professor of sociology and criminology at the University of Ontariostitute of Technology, Canada. His research interests include comparative sociology,nfidence in the police, criminological theory, gun ownership, police integrity, public

attitudes toward prostitution, race and ethnicity in criminal justice etc. His researchessays have appeared in Criminology, Journal of Criminal Justice, Justice Quarterly,Policing, and Social Forces. He is the author of Major Criminological Theories: Conceptsand Measurement (2004). His co-authored paper “Crime volume and law and orderculture” (2007) won the 2008 Academy of Criminal Justice Sciences’ Donal MacNamaraAward – the best article of the year.

Yung-Lien Lai, PhD, is an Assistant Professor in the Graduate School of Criminology atNational Taipei University, Taiwan. He earned his doctoral degree in criminal justicefrom Sam Houston State University in 2011. Prior to his academic career, he servedas a correction officer for 10 years at the Department of Corrections, Ministry of Justicein Taiwan, R.O.C. Dr Lai's primary research interests include community policing (po-lice-citizen relations), comparative correctional systems, and inmate misconduct. Hisrecent publications have appeared in Journal of Criminal Justice, Crime & Delinquency,and Journal of Crime and Justice.

Ruohi Zhao, Ph.D., is an assistant professor at the University of Macau. Her researchinterests include criminological theory, comparative criminology, and juvenile delin-quency. Her research has appeared in Social Forces.