growth is good for whom, when, how? economic growth and poverty reduction in exceptional cases

World Development Vol. 36, No. 11, pp. 2127–2143, 2008� 2008 Elsevier Ltd. All rights reserved.

0305-750X/$ - see front matter

doi:10.1016/j.worlddev.2007.10.020www.elsevier.com/locate/worlddev

Growth is Good for Whom, When, How?

Economic Growth and Poverty Reduction

in Exceptional Cases

JOHN A. DONALDSON *

Singapore Management University, Singapore

Summary. — Economic growth and liberal economic policies often help the poor, but what aboutthe numerous cases in which they do not? This article analyzes two types of cases: those in whichincome growth of the poor was significantly lower than expectations (‘‘negative exceptions’’) andthose in which income growth of the poor significantly exceeded expectations (‘‘positive excep-tions’’). Insights from these cases inform our theoretical understanding of poverty reduction. Inaddition, this article contributes a typology of strategies used in these cases, including alternativepathways to economic growth and neoliberal prescriptions for poverty reduction.� 2008 Elsevier Ltd. All rights reserved.

Key words — economic growth, poverty reduction, liberalization

‘‘It should come as no surprise that the general relation-ship between growth of income of the poor and growthof mean income is one-to-one.’’ (Dollar & Kraay,2000, p. 28)

‘‘Faster growth is normally better for the poor thanslower growth, and is not systematically offset byany change in distribution. But huge exceptions—andthe possibility of clusters of countries where growthis much better for distribution, or much worse—meanthat these findings are the beginning, not the end, ofthe inquiry. Residuals matter.’’ (Eastwood & Lipton,2001, p. 16)

* The author acknowledges a group of young scholars,

led by Swati Chaudhary, who helped research individual

or small sets of cases. This group consists of Anandiya

Bose, Chintan Rastogi, Lim Feng Ling, Malavika

Shanker, Meera Kanhere, Sa’adia Zaheer Baig, Sidd-

harth Poddar and Yoganand Chillarige. Others who also

helped instrumentally with other aspects of the research

include Hal Wolman, Kristopher Ramsay, Kazuhiro

Obayashi, Madhu Chaubey, Hing-man Leung, Yip Ch-

un Seng, Stephanie McNulty, Mark Teel, Tomoki Fujii,

Anthony G. Pazzanita, Chan Ying Xian and Kaleng

Wong. Remaining errors are solely those of the author.Final revision accepted: October 5, 2007.

1. INTRODUCTION

Economic growth often helps the poor, butwhat about the numerous cases when it doesnot? The widely-held belief that economicgrowth generally reduces poverty, encapsulatedby two World Bank economists in the above-quoted article entitled ‘‘Growth is Good forthe Poor’’ leaves many cases unexplained. Intheir article—one of the most influential toadvocate this argument—Dollar and Kraayanalyzed hundreds of cases (countries overperiods of at least five-years), concluding thateconomic growth and poverty reduction are re-lated on a one-to-one basis (Dollar & Kraay,

212

2002, p. 196). In addition, macroeconomic pol-icies associated with liberalization, such asreducing inflation, moderating the size of thegovernment, respecting the rule of law, openingthe economy to trade, and establishing a soundfinancial system, are good for both generatingeconomic growth and reducing poverty (p.218). Many scholars have criticized Dollarand Kraay’s research on methodological andtheoretical grounds (e.g., Danielson, 2001;Eastwood & Lipton, 2000; Ravallion, 2001;Rodrik, 2000). The present article adopts a dif-ferent tact, taking seriously Eastwood and Lip-ton’s criticism (quoted above) that the original

7

2128 WORLD DEVELOPMENT

work ignores cases that are exceptions to theeconomic growth-poverty reduction relation-ship—in short, that ‘‘residuals matter.’’ AsDollar and Kraay recognized on numerousoccasions in their article, even a one-to-onerelationship represents an average—a general-ization with numerous exceptions. We canlearn much from studying such exceptions.

This research paper applies Dollar and Kra-ay’s data to examine two types of exceptionsto this generalization. The first, ‘‘positiveexceptions,’’ are cases during specific time peri-ods in which the poor did much better than themodel’s prediction based on economic growthrates. The second type, ‘‘negative exceptions,’’includes cases in which the income of the poorincreased significantly less than was expected.Examining these positive and negative excep-tions is useful for illuminating pathways otherthan growth (or the absence of it) for povertyreduction (or the lack of it). Perhaps the posi-tive exceptions can discern policy approachesother than economic growth that are effectivefor reducing poverty, providing hope, andalternatives, to the myriad economies in whicheconomic growth has been either elusive ortoo slow in reducing poverty. Negative excep-tions, likewise, may teach us approaches toavoid. In either case, this analysis moves be-yond simple reliance on economic growth andlooks for potential systematic factors thatmight explain the exceptions. Although ‘‘onaverage’’ economic growth can be expected tohelp reduce poverty, as World Bank economistMartin Ravallion argues, ‘‘People are oftenhurting behind the averages. Panel data andobservations from the ground can reveal this,but the aggregate statistics cannot. It is impor-tant to know the aggregate balance of gains andlosses, but it will be of little consolation tothose suffering to be told that poverty is fallingon average,’’ (Ravallion, 2001, p. 1811). 1

In this way, this paper neither contradicts norconfirms the general relationship between eco-nomic growth and poverty reduction. It arguesinstead that there are multiple pathways topoverty reduction, of which Dollar and Kraayidentified but one—economic growth generatedthrough liberal economic policies. 2 As somehave argued (e.g., George & Bennett, 2005;Goertz, 2005), relying on aggregate statisticsis not the best way to identify the numerous di-verse paths to a result, such as the multipleways that countries have successfully, andunsuccessfully, addressed poverty. Instead,careful qualitative research can be more effec-

tive in identifying alternative, contingent path-ways. Else Øyen’s argument, more than adecade old, applies equally today, ‘‘Up-to-datedata are necessary to ensure that the poor andthe intensity of poverty are kept visible to thepublic eye, but it may still be wise to put some-what less energy into sheer measurement re-search, and instead turn to issues that yieldmore in poverty understanding,’’ (Øyen, 1996,p. 10).

2. IDENTIFYING EXCEPTIONS

Defining the poor as the lowest quintile interms of income, Dollar and Kraay compileda dataset from four different standard sourcesto produce 953 observations from 137 differentcountries or territories during 1950–99. To ren-der the sample more evenly distributed acrosscountries, Dollar and Kraay chose dates fromeach country that are spaced five years apart,starting with the first year available, resultingin 418 country–year observations of mean in-come of the poor. The authors further filteredthe sample to 258 observations from 92 coun-tries with at least two observations, also spacedfive years apart. 3 Dollar and Kraay, perform-ing an ordinary-least-squares regression analy-sis of economic growth against povertyreduction, found an R2 of 0.49 and a regressionline with a slope of 1.19. In spite of these re-sults, the authors realize (notwithstanding theiruncompromising title) that they cannot claimthat poverty rates vary lockstep with changesin the economy, cautioning, ‘‘Our findings donot imply that growth is all that is needed toimprove the lives of the poor. Rather, we sim-ply emphasize that growth on average doesbenefit the poor as much as anyone else in soci-ety, and so standard growth-enhancing policiesshould be at the center of any effective povertyreduction strategy,’’ (Dollar & Kraay, 2002, p.219). While Dollar and Kraay attempted to ad-just to the difficulties in comparing surveysacross countries, such as differing coverage,measures and units between observations, crit-ics argue that this dataset has numerous addi-tional methodological problems. Nevertheless,I adopt the same dataset because it provides acommon basis from which to identify and learnfrom exceptions to the relationship betweeneconomic growth and poverty reduction.

Although measurement errors and other dataproblems can be exacerbated when shiftingfrom aggregate data to individual cases, closer

GROWTH IS GOOD FOR WHOM, WHEN, HOW? 2129

scrutiny is also an effective way to identify sucherrors. For instance, such analysis revealed atleast two cases among the 27 cases that appar-ently contain serious data errors. The first isNepal. While Dollar and Kraay’s dataset re-ported that Nepal experienced a slight negativeannual growth rate (�0.15%) during 1977–84,the World Bank’s World Development Indica-tors reported positive economic growth acrossa variety of measures of GDP. Moreover, out-lying villages in Nepal—among the poorest inthe country—that were inaccessible by naviga-ble road were probably not counted in the sur-vey, and thus the data reflecting povertyreduction may be inaccurate. Second, the datafor Yemen are even more suspect. Given the ci-vil war, conflict with Eritrea, and extremely lowliteracy rates, it is unlikely—and little evidencecan be found—that the income of the poor in-creased 8% per annum, as Dollar and Kraay’sdata suggested.

Overall, Dollar and Kraay’s sample containsmuch more variation than an R2 of 0.49 im-plies. For instance, more than one in six ofthe authors’ 285 historical cases violate theexpectation that positive income growth im-proves the income of the poor (in 45 cases, in-comes of the poor declined while GDPincreased) or that negative income growth willreduce incomes of the poor (in six cases the in-come of the poor increased despite negativeGDP growth). Indeed, a number of data pointslie far enough away from the regression linethat the income of the poor is statistically unli-kely to be explained by economic growth alone.Instead of looking at the dataset in aggregate,this article uses the exceptional cases to identifyalternatives to economic growth for povertyreduction.

Exceptions are determined by computingresiduals, the vertical distance between theregression line and each data point, and calcu-lating the probability that the distance is due torandom variation. The further away any givenpoint is from the regression line, the less likelyit is that the model explains that point. Foreach data point, when the probability thatchanges in GDP explain changes in the incomeof the poor drops below 5% in either direction(equivalent to a 90% confidence interval), I la-bel that point an ‘‘exception.’’ This process gen-erates two types of exceptions (listed in Table1): (a) cases in which the increase in the incomeof the poor significantly outpaces expectationsbased on economic growth (‘‘positive excep-tions,’’—positive from the point of view of pov-

erty reduction), and (b) cases in which theincrease in the income of the poor was muchless than the expectations based on economicgrowth (‘‘negative exceptions’’). Specifically, Isubtract the change in poverty rates predictedby the model (column 4 in Table 1 below) fromthe value reported by the data (column 5) andrecord this ‘‘residual’’ value in column 6. Col-umn 7 (labeled ‘‘outlier’’) records the probabil-ity that the position of each case’scorresponding value for income of the lowestquintile is caused by random variation. This isestimated by calculating the P-values of theresidual’s Z-scores to determine the chance thatthe point’s distance from the regression line iscaused by random factors. For instance, as seenin Table 1, Finland during 1962–71 saw an an-nual growth rate of 3.99%, according to Dollarand Kraay’s data. Although the model predictsthat the income of the poorest quintile shouldhave increased 4.05%, the actual increase overthis period, according to the data, was 14.6%,a difference of 10.55 percentage points. Sincethe likelihood that this data point can be ex-plained by random error is approximately0.38%, Finland is considered a positive excep-tion. By contrast, China’s economy during1990–95 grew at an annual pace of 8.7%. How-ever, although the model predicts that thegrowth rate in the income of the bottom quin-tile would exceed 9.6%, the income of thatgroup grew annually by only 0.87%. Thechance that this data point is explained by ran-dom error is approximately 1.3%, meaning Chi-na during that time period qualifies as anegative exception.

These exceptions are not trivial. In the 13 po-sitive exceptions, the incomes of approximately33.5 million poor people grew faster than themodel’s predicted rate by an average of 9.1 per-centage points, while around 341 million poorpeople (101 million, excluding the case of Chi-na) in the 14 negative exceptions saw their in-comes rise by an average of 8.5 percentagepoints less than the model predicted. 4 More-over, these cases point toward alternative pathsto poverty reduction, recipes that do not in-clude growth, at least not as a prime ingredient.This paper examines the possibility that in theseexceptional cases, specific political, social oreconomic factors explained the particular pat-tern of economic growth and poverty reductionexperienced in the country. A complete analysisof each of the cases is an ambitious project,requiring greater time, space, and specific areaknowledge than is afforded here. Nevertheless,

Table 1. Exceptional cases

Country Time period Annualgrowth of

GDP/capita

Predicted annualchange in incomeof lowest quintile

Reported annualchange in incomeof lowest quintile

Residualreported –predicted

‘‘Outlier’’

Cases in which reported annual change in income of lowest quintile exceeded model’s predictions

Colombia 1964–70 2.33% 2.08% 17.16% 15.08% 0.01%Norway 1979–84 2.75% 2.58% 14.57% 11.99% 0.12%Finland 1962–71 3.99% 4.05% 14.61% 10.56% 0.38%Nepal 1977–84 �0.15% �0.86% 9.61% 10.47% 0.40%Honduras 1986–91 �0.25% �0.97% 8.62% 9.60% 0.75%Yemen 1992–98 0.28% �0.35% 8.00% 8.35% 1.70%Mauritania 1988–93 1.72% 1.36% 9.65% 8.30% 1.76%Peru 1971–81 0.91% 0.40% 8.51% 8.11% 1.98%Chile 1987–92 5.14% 5.41% 13.35% 7.94% 2.20%Norway 1989–95 2.67% 2.48% 10.26% 7.78% 2.41%El Salvador 1989–95 2.59% 2.39% 9.51% 7.11% 3.54%France 1975–81 2.19% 1.92% 9.01% 7.08% 3.59%Costa Rica 1977–82 �3.41% �4.72% 2.25% 6.97% 3.83%

Cases in which reported annual change in income of lowest quintile fell below model’s predictions

Ukraine 1988–95 �10.96% �13.66% �20.21% �6.55% 4.80%El Salvador 1977–89 �1.74% �2.73% �9.31% �6.58% 4.73%France 1956–62 3.84% 3.87% �3.07% �6.95% 3.87%Singapore 1978–83 5.83% 6.23% �1.28% �7.51% 2.82%Mali 1989–94 �2.62% �3.78% �11.39% �7.61% 2.67%Poland 1991–96 4.83% 5.04% �2.73% �7.78% 2.42%Estonia 1988–93 �8.40% �10.63% �18.41% �7.78% 2.41%Colombia 1970–78 3.35% 3.29% �4.79% �8.08% 2.02%Dominican Rep. 1984–89 2.38% 2.15% �6.45% �8.59% 1.46%Brazil 1986–93 �0.97% �1.83% �10.57% �8.75% 1.32%China 1990–95 8.71% 9.64% 0.87% �8.77% 1.30%Bulgaria 1989–94 �4.86% �6.43% �16.28% �9.85% 0.63%Puerto Rico 1963–67 6.08% 6.53% �4.80% �11.33% 0.21%Russia 1988–93 �6.43% �8.30% �20.88% �12.58% 0.07%

Sources: Dollar and Kraay (2002); author’s calculations.


an initial examination reveals that these excep-tional cases represent a range of approaches topoverty reduction and economic growth. Noneof these paths is new in that scholars have iden-tified and analyzed many of them long ago.Nevertheless, because they re-emerge usingDollar and Kraay’s own data, they underscorethat there are viable alternatives to Dollar andKraay’s conclusions about the importance andcentrality of economic growth and the need forimplementing a set of liberal policies to reducepoverty.

3. EXCEPTIONAL CASES

Three of the positive exceptions are Scandi-navian social democracies with broad-based so-cial programs designed to minimize poverty

rates (Gustafsson & Pedersen, 2000). Both Nor-way and Finland’s policies are characterized bylarge-scale government transfers, decommodifi-cation, unemployment programs and otherextensive social programs (Kenworthy, 1999).While Norway’s economy grew relativelyslowly during 1979–84 (2.8% each year) andagain during 1989–95 (per capita GDP declined0.2% on average each year), the income of thepoorest quintile grew rapidly, increasing14.6% and 9.6% on average each year overthose periods, respectively. The per capitaGDP of Finland grew on average nearly 4%each year during 1962–71, while the income ofthe bottom quintile increased 14.6% on averageeach year over that period. For Finland, thiswas a period of rapid poverty reduction, withthe number of Finns with income below theminimum national pension declining from


20% in 1966 to less than 6% in 1971, a resultthat the authors of one major study argue is ex-plained by the establishment of a welfare state,‘‘clearly the major, if not the only reason for thedeclined poverty rates in Finland,’’ (Gustafsson& Uusitalo, 1990, p. 256, 261). That Finlandwas a positive exception only in this period,and not in others, probably reflects increasingunemployment rates in the 1970s (from 1.9%in 1974 to 7.3% in 1978) and life-cycle changeswhich slowed gains for the poor (Gustafsson &Uusitalo, 1990). Neither of these countries di-rectly relied on economic growth, and bothimplemented policies that conflict with liberaleconomic policies, with the welfare state creat-ing conditions for rapid gains for those withlowest incomes while the economy as a wholegrew more modestly. The extensive welfare pro-visions and cradle-to-grave coverage, as well asthe large and progressive tax rates needed tosupport this system likely came at the expenseof growth rates, an exchange the country wasapparently willing to accept.

Even as these social democracies maintainedbroadly targeted programs to ensure the distri-bution of wealth throughout these societies,one positive exception reduced poverty throughnewly established, targeted social safety netsand complementary social programs. Duringthe oft-derided presidency of Giscard d’Esta-ing, the growth in income of France’s poorestquintile outpaced economic growth to such anextent that the country qualifies as a positiveexception. During this period Giscard, arguingthat France would ‘‘never be at ease with her-self until all the old inequalities had been re-moved,’’ established no fewer than sevenbenefit programs targeting and protecting thepoor. Prime Minister Chaban-Delmas solidifiedFrance’s welfare state by pursuing a ‘‘new soci-ety,’’ including vocational training, welfare forthe poor and elderly, and minimum wage in-creases. Meanwhile, oil shocks and skyrocket-ing raw material prices, combined with thecollapse of the Fordist system of manufacturingslowed the economy (Levy, 1999). As a result,the income of the bottom quintile increasedby an average of 9% each year, in spite of thecountry’s moribund overall economic perfor-mance of 2.2% each year during 1977–82. Con-servative leadership strengthened the welfarestate, helping to increase the incomes of thepoorest quintile, shielding them from factorsthat slowed the growth of the overall economy.

A number of positive exceptions from thedeveloping world achieved poverty reduction

through aggressive land reform and other redis-tributive policies. Colombia (1964–70) is onecase in which a large-scale program of land re-form, implemented during the presidency ofCarlos Lleras Restrepo (1966–70), helped in-crease significantly the income of the poor.During 1968–69 alone, some 60,000 land titlesinvolving 2.5 million hectares were issued topeasants and unemployed workers, a move thatwas vital to empowering the poor (Findlay,1972). 5 In addition, Restrepo’s policies helpedcurb inflation, diversified the economy awayfrom the waning coffee market, and improvedthe country’s balance of payments. Dollar andKraay’s raw data supported this view, report-ing that the Gini index for income in Colombiadeclined from 0.62 to 0.52 during 1964–70(although the Gini index rose again from 0.51to 0.57 during 1991–95 once land reform wasreversed). Thus, during a period in Colombia,which included several years of land reform(1964–70), this and other pro-poor policiesseem to be primary causes behind the incomegrowth of the poor, which averaged 17% eachyear. However, stimulating the rural economyrarely helps spur rapid economic growth, whichcame in at a modest 2.3% per year over the per-iod. 6

Redistributive policies are also responsiblefor the exceptional income growth of the poor-est quintile in some of the other exceptional La-tin American cases. For instance, the term ofPeru’s President Juan Velasco Alvarado wascharacterized by nationalization of the petro-leum and other industries, as well as compul-sory land reform policies, and broadereducation policies (Lopez & Valdes, 2000;McClintock, 1981). Although Peru’s economystagnated during 1971–81, growing less than1% on average as foreign debt mounted, the in-come of the poor in Peru increased by morethan 8% per year over that time as peasantsbenefited from land reform and other redistrib-utive policies. Moreover, the low growth waslikely directly linked to the government’s pov-erty reduction policies, suggesting povertyreduction came to some extent at the expenseof economic growth. During 1972–85, inequal-ity declined sharply with the share of the totalincome of the poorest 60% increasing from18% to 27%, according to one survey (Glewwe,1988). Similarly, the income of the poorestquintile of El Salvador (1989–95) increased9.5% on average, despite a relatively modest an-nual average economic growth of nearly 2.6%.The peace accords signed in January 1992 that


ended the civil war in that country coincidedwith policies designed to address staggeringinequality through land transfers, rural educa-tion, support for establishing microenterprises,and housing assistance, broadened povertyalleviation programs, and spending for socialsectors, policies intended to reintegrate com-batants into civilian life. These primarilybenefited lower income groups by enhancingnon-agricultural jobs, as well as improving edu-cation and opportunities for the poor (Boyce,1995; Marques, 2004). These progressive redis-tributive programs, along with remittancesfrom relatives overseas 7 and a period of rela-tive peace and stability, helped to make El Sal-vador a positive exception in which the poorreceived a disproportionate share of the bene-fits of economic growth.

Similarly, Chile’s first post-Pinochet govern-ment era, led by the reform-minded, center-leftgovernment led by Patricio Aylwin, wasmarked by gradual reform, redistribution andrelative empowerment of the poor, makingthe country a positive exception (1987–92).The repressive Pinochet dictatorship (1973–89), advised by the ‘‘Chicago Boys,’’ a groupof University of Chicago trained economistsinspired by Milton Friedman, was famouslyneoliberal (Hudson, 1994). Yet, the adminis-tration’s economic record was surpris-ingly mediocre, with per capita economicgrowth averaging a modest 1.9% each yearfrom 1974 to 1989, and regressive, leaving44.4% of all Chileans under the poverty linein 1987 (Ritter, 1992). Based on a credo of‘‘growth with equity,’’ the subsequent Aylwinadministration implemented an ambitious setof social programs. In the early 1990s, Aylwinincreased real health expenditures by animpressive 70% and expanded primary healthcare services in poor areas, both rural and ur-ban. The government strengthened labor un-ions, vigorously oppressed by Pinochet,carefully compromising with business to pro-tect workers without undermining the marketeconomy. New programs providing job train-ing and child care were also enacted, while pen-sions were made more equitable. All of this waspaid for by increases in the value-added tax.These policies were largely responsible for thestriking decline in poverty from 5.2 million in1990 to 800,000 in 1992, with a 50% declinein the poverty gap (Weyland, 1997). Further,Alwyin allowed a degree of decentralization,enhancing the autonomy and independentfunding of municipalities. Programs such as

those related to Chile’s Sectoral RegionalInvestment (ISAR), allowed regional govern-ments to request the transfer of up to 5% ofthe investments of Chile’s sectoral ministries.Projects transferred under the ISAR mecha-nism included rural and urban roads and pave-ments, neighborhood schemes, and potablewater. Decentralizing these schemes led to amore equitable distribution of resources amongthe municipalities, and local governments’ bet-ter understanding of local conditions furtherbenefited the poorest people in the country(Stewart & Ranis, 1994). Dollar and Kraay’sdata are consistent with the research of others(e.g., Larranga, 1994; Castro, 1994, cited inWeyland, 1997) that economic growth cannotfully explain the strong employment growthand rise in wages (a predicted 5.14% increasein income of the lowest quintile compared tothe 13.35% rise actually measured). DespiteAylwin’s reforms, the administration did notfundamentally alter the liberal market system,but promoted significant yet gradual changesthrough effectively increasing spending for so-cial programs and protections, and the poorbenefited disproportionately from the relativelyrobust growth.

Contrasting with this reform-minded govern-ment, one of the positive exceptions in LatinAmerica implemented structural reforms, indi-cating that liberal prescriptions can be a paththat helped cases become positive exceptions.Rodrigo Alberto Carazo Odio, Costa Rica’sconservative president during 1978–82, facedan emerging debt crisis and rapidly increasinginflation rates. In response, he implementedIMF-prescribed reforms and policies to liberal-ize the economy, with mixed results. During1977–82, Costa Rica’s economy shrank morethan 3% each year on average, increasingunemployment. Despite the recession, the in-come of the poor gained 2.25%, far exceedingthe predicted 4.72% loss, which rendered thecountry a positive exception during this period.

Just as countries have traversed a variety ofpaths to become positive exceptions, so toohave some countries become negative excep-tions through a range of policies. The mostcommon of these did so through rapid liberal-ization and reform. The ‘‘shock therapy’’ andprivatization policies implemented by the gov-ernments of Eastern Europe and the former So-viet Union in the wake of Communism’scollapse sharply contracted the economy, elim-inating millions of jobs and destroying the al-ready fraying social safety nets of the Soviet


system. This affected the poor especially badly(Bunce, 1999). While the gross domestic prod-ucts of Bulgaria (1989–94), Estonia (1988–93)and Ukraine (1988–95) declined annually byan average of 4.9%, 8.4% and 11%, respectively,over this period, the income of the poor fell farmore than expected, declining each year by anaverage of 16.3%, 18.4% and 20.2%, respec-tively. Russia’s experience (1989–93) is espe-cially dramatic. While the Russian economycontracted as per capita GDP declined 6.4%each year on average, the income of the poordeclined 20.9% each year, as workers were re-trenched, inefficient factories were closed, theagricultural sector weakened, and economicprotections were dismantled. These EasternEuropean countries implemented sudden tran-sitions from planned to market-based econo-mies, including the removal of price supports,elimination of subsidies for state-owned enter-prises, liberalization of foreign trade, and slash-ing of government expenditures (Derleth,2000), especially rapid versions of liberal re-forms that Dollar and Kraay advocated. Lack-ing economic and political institutions, neededas regulators and facilitating mechanisms inthe absence of the state and considered to beessential for the capitalist economy to moderatethe effects of these sudden shifts, the economycollapsed. GDP plummeted in these countrieswith the poor suffering more than others(World Bank, 2000).

Because Poland is often cited as defying thistrend, it is surprising to find it among the neg-ative exceptions. When Poland liberalized afterthe fall of communism, new-found freedom toparticipate and invest in the economy sparkedeconomic growth overall, making the countrythe only post-socialist state to achieve a positiveaverage GDP growth rate during 1990–97(Bunce, 1999) and the only European post-so-cialist economy to surpass its pre-transitionGDP within six years of the transition (Derleth,2000). However, like other Eastern Europeancountries, shock therapy also adversely affectedthe lowest quintile. Rapid privatization causedmass layoffs of workers, increasing unemploy-ment from nearly zero to more than 15% dur-ing 1989–93, without the corresponding safetynets found in most Western European countries(Kramer, 1995). The poor’s income declined byan average of 2.7% each year during 1991–96, 8

less than in other shock therapy countries, butenough to raise misgivings on Poland’s statusas a ‘‘model’’ example of post-Communist re-form (e.g., Sachs, 1995, p. 275), as far as pov-

erty is concerned. In Poland’s case (unlikemuch of the rest of post-Communist EasternEurope) liberalization spurred growth, but (likethe region’s other reforming states) theseimpressive gains came at the expense of incomegrowth among the poor.

At least two other countries became negativeexceptions because they scaled back or elimi-nated progressive social policies. Colombia,during 1970–78, directly after the period inwhich the country was a positive exception, be-came a negative exception. During this period,in order to liberalize and allow the market tomake distribution decisions, the governmentrolled back land reform and other progressivesocial policies that it had implemented duringthe previous period. In their place, the govern-ment implemented market-oriented policies,including opening the economy, providingincentives for foreign capital and eliminatingbarriers to free investment in the countryside(Molano, 2000). The government’s cancellationof land reform in 1971 and the subsequentintroduction of new crop varieties during thegreen revolution exacerbated the already ineq-uitable distribution of land (Findlay, 1972;Puyana, 2000). Ironically, though this is consid-ered a time of economic recovery, with annualper capita GDP growth in Colombia increasingduring this period by 3.3% on average, the in-come of the poor concurrently declined annu-ally by an average of 4.8%. Thus, while thecancellation of many of the pro-poor programslikely increased economic growth rates, com-pared to that of earlier periods, this growthcame at the expense of policies intended to re-duce poverty, and reduced the income of thepoor.

That China (1990–95) is a negative exceptionis surprising, since it is often cited as a typicalcase in which economic growth directly con-tributed to poverty reduction (e.g., Zhang,Huang, & Rozelle, 2003). One key reason wasthe shift from reforming the rural economy tostimulating urban development. For instance,subsidized loans, originally intended to reducerural poverty through subsidizing poor agricul-tural families’ investments in agricultural inputsand assets, were redirected during this period topromoting industry, retarding its effect on ruralpoverty reduction (Zhang et al., 2003). Despiterising overall GDP rates, this period saw adeclining share of GDP for agriculture (espe-cially between 1990 and 1993), which one recentWorld Bank study emphasized is crucial forpoverty reduction (Ravallion & Chen, 2007).


Finally, during 1988–95, education and healthcare costs took an increasingly high proportionof rural incomes of the rural poor, an effectwhich countered much of the benefits of eco-nomic growth for the poor (Gustafsson & Li,2004). According to Dollar and Kraay’s data,economic growth in China during 1990–95 in-creased by an average of 8.7%, even as the in-come of the poor increased on average lessthan 1% each year, a rate that is consistent withGustafsson and Li’s disaggregated income data(p. 295). Thus, while the income of the poorcontinued to rise, the growth lagged far behindin income in the rest of the population.

Debt, inflation, widespread corruption, andmisguided policies were primary factors forBrazil becoming a negative exception. Thecountry’s economic downturn of 1986–93, aperiod of mounting debt and rampant corrup-tion, was overseen by the Sarnay administra-tion. The poor were also hurt by a punitivepolicy of price and wage control and a 1987moratorium on foreign debt payments that iso-lated Brazil, weakening foreign credit andinvestment. In 1990, its first year in office, thesucceeding Collor presidency implemented far-reaching economic reforms, including deepen-ing price and wage freezes, confiscating savings,and deindexing the economy. While this suc-cessfully curbed inflation, real incomes plum-meted, especially for the poorest (Roett,1999). Meanwhile a sagging agricultural sectoraffected large numbers of rural residents, par-ticularly in northeastern Brazil. Abuses of thestate hamstrung national development (per ca-pita GDP declined about 1% each year overthat period), and the poor, whose income de-clined 10.6% per year, fared far worse thanthe population as a whole.

Chaos and violent disruption was the directcause of at least one negative exception. Theeconomic losses to the poor when death squadsterrorized El Salvador compounded the generalmisery of this period. Much of the period of the‘‘death squads,’’ which began with the right-wing military government seizing power in1979 and ended with peace accords in 1992(Boyce, 1995), overlaps with the 1977–89 peri-od in which the income of the poor shrank9.3% per year, compared to the annual declinein per capita GDP of 1.7%. (Subsequently, asdescribed above, the country’s recovery quali-fied it as a positive exception.) The civil warultimately claimed 70,000 lives and destroyedmuch of the nation’s infrastructure. While theeconomy as a whole suffered greatly, the severe

disruption in the education system and othersocial services further eroded the already pre-carious living conditions of the poor, dispro-portionately affecting poor people (Marques,2004).

Given the country’s relatively broad-basedpublic housing and education programs in-tended in part to moderate poverty, Singa-pore’s status as a negative exception is alsounexpected. Nevertheless, during 1978–83,when per capita economic growth averaged5.8% per year, the income of the lowest 20% de-clined 1.3% on average each year. During thisperiod, the government transformed the econ-omy to be more capital-intensive, encouragingthe automation of processes previously doneby non-skilled or semi-skilled physical labor(Chow, Lee, Hameed, & Cheong, 1988, p.178; Peebles & Wilson, 1996, p. 37). Since pub-lic assistance for the poor in Singapore is highlyrestricted, families avoid poverty primarilythrough employment (Lee, 2001). During thisperiod, economic growth remained moderatelystrong as productivity increased, but those withinsufficient human capital (the poor, dispropor-tionately) to meet new demands from the ser-vice sector suffered job losses and lowerincomes (Chow et al., 1988). Economic growthlikely came at the expense of the poor.

Focusing on exceptional cases can sometimesidentify two comparable countries that are rel-atively similar in many aspects, but are onopposite ends of the spectrum. Examining thesenaturally controlled cases allows us to focus oncontrasting policies and other factors thatmight cause the differences. This sample con-tains two neighboring African countries, onea positive exception (Mauritania), the other anegative one (Mali). The increase in the incomeof the poor in Mauritania (9.7% on averageduring 1988–93, despite growth of 1.7%) canbe explained by several major factors. Firstand most important was a return to stabilityin the wake of three factors: (a) recovery fromrepeated droughts in the 1980s, (b) the cessa-tion of the militarized conflict with Senegal,and (c) the end of the country’s isolation fromwestern powers that began when Mauritaniasupported Saddam Hussein during the firstGulf War. In the early 1990s, President OuldTaya introduced political and economic re-forms, granted amnesty to former militantsand political opponents, and held free elections.With his sweeping election victory, the presi-dent oversaw several years of relatively stableand effective government, and various types of

Table 2. Multiple paths toward/away frompoverty reduction

Positive exceptions

(1) Progressive redistributionColombia (1964–70) Peru (1971–81)El Salvador (1989–95) Chile (1987–92)(2) Social welfare programsFrance (1975–81)(3) Social democracyFinland (1962–71) Norway (1989–95)Norway (1979–84)(4) Liberal policies/structural readjustmentCosta Rica (1977–82)(5) Stability/opportunities accessible by the poorMauritania (1988–93)

Negative exceptions

(1) Regressive redistributionColombia (1970–78) China (1990–95)(2) Structural adjustmentMali (1989–94) Singapore (1978–83)(3) Violence/chaosEl Salvador (1977–89)(4) Corruption/debtBrazil (1986–93)(5) Shock therapyBulgaria (1989–94) Ukraine (1988–95)Estonia (1988–93) Russia (1988–93)Poland (1991–96)


non-governmental organizations and otherforms of civil society flourished (Pazzanita,1997). 9 Second, a large-scale mining sector—including recently established copper and goldindustries—added jobs that low-skilled poorpeople could perform (Coulombe & McKay,1996). Third, IMF programs that were sus-pended during the war with Senegal but rein-stated in 1992 also had the effect of reducingdomestic demand through fiscal consolidationand a conservative monetary policy—effectswhich likely constrained GDP growth, but didnot hurt as much the rural poor, who reliedon subsistence farming and were insulated toa certain extent from the formal economy.While the stability of this period helped boostthe economy as a whole, it likely favored therural poor who suffered greatly through thewar and drought. This factor, combined withnew opportunities and programs for the poor,explains the reported disconnect between eco-nomic growth and poverty reduction.

Mali, Mauritania’s neighbor, was also anexception during 1989–94, but a negative one.Just before this period, in a misguided attemptto respond to a series of droughts, the govern-ment liberalized and restructured the agricul-tural sector, flooding the market with cheaperimports and destroying the market for domesticagricultural production. Under IMF loan con-ditions, Mali slashed government employment,sold state assets, increased taxes, and enhancedcontrol over import and export duties. The sud-den and steep devaluation of the local currencyparadoxically made it more difficult and expen-sive for urbanites—the poor especially—to pur-chase food (Toulmin, Leonard, Hilhorst, &Diarra, 2000). The resulting economic reces-sion—per capita GDP declined 2.6% on aver-age each year during 1989–94—was especiallyharsh on the country’s poor, whose income de-clined 11.4% annually over that period.

4. ANALYSIS

As Table 2 summarizes, there are multiplepaths to reduce poverty and many ways toexacerbate it. Progressive redistribution policiesexplain at least three positive exceptions(Colombia, Peru, and El Salvador). The gov-ernment of one case (France 1975–81) strength-ened the nation’s social safety net throughextensive state-sponsored welfare programs.Three cases are Scandinavian social democra-cies. Structural readjustment policies based on

liberalization explain at least one positiveexception (Costa Rica), while stability and in-creased opportunities for employment of thepoor combine to create the final positive excep-tion examined here (Mauritania). In each ofthese cases, policies were implemented that in-creased the incomes of the poorest quintile toa greater extent than predicted by overallgrowth. Similarly, negative exceptions tra-versed a variety of pathways. Two (Colombia1970–78 and China 1990–95) rolled back pro-gressive policies. Violence and the destructionof civil war explain one negative exception (ElSalvador), while corruption, debt, and inflationexplain yet another (Brazil 1986–93). Theimplementation of structural adjustment poli-cies was the most common path taken by nega-tive exceptions. Two implemented structuraladjustments of various types—Mali an IMFprogram and Singapore an internally-deriveddevelopment strategy. Five Eastern Europeancases implemented policies identified with‘‘shock therapy,’’ that also involved wholesalestructural adjustment. Each case saw the in-come of the poorest drop far more than would


be expected by economic growth alone (only inone case did the income of the poor actuallygrow at all). This list, generated from a limitedsample, does not exhaust the range of possiblestrategies.

Many cases involve other factors that cutacross those in Table 2. First, many cases sup-port the finding that promotion of agriculture,especially in the developing context, can helpthe poor to a greater extent than economicgrowth alone (e.g., Dorward, Kydd, Morrison,& Urey, 2004; Mellor, 1995). While traditionalagriculture is often the main occupation of thepoor, it is often of low scale and lacking in tech-nology, and is thus not generally particularlyproductive. An expanding agricultural sectortherefore can contribute to poverty reductionwhile not substantially promoting economicgrowth in the short-term. This was seen in thepositive exceptions which implemented land re-form (e.g., Peru, Colombia 1964–70, and Ne-pal), as well as in negative exceptions inwhich the agricultural sector slackened (e.g.,Colombia 1970–78, China, and Mali). 10 Thus,agriculture—for good (if it is promoted) or ill(if it declines)—seems to be instrumental inexplaining the results of a number of excep-tional cases. Second, democratization is amixed bag regarding its impact on poverty, assome have argued (e.g., Leftwich, 2005). Whilethe positive exception of Chile (1987–92) is onecase in which democratization encouraged pro-poor policies, other cases cast doubt on the po-sitive role of democracy. For instance, theauthoritarian leader of Peru was able to usethe coercive power of the state to implement asweeping program, including forced land re-form and other programs that contributed toimproving the incomes of the poor. Moreover,the wave of democratization that swept acrossEastern Europe coincided with economic stag-nation that disproportionately hurt the poor.On the other hand, this likely resulted morefrom the breakdown of government institutionsbrought about through rapid reform than it didfrom democratization per se.

This research also supports the hypothesisthat ‘‘inequality is bad for the poor’’ (as oneadvocate, Ravallion (2005), puts it) and con-versely that lower Gini coefficients correlatewith pro-poor growth. Looking at the relation-ship between static inequality and povertyreduction on the extreme cases as a group, amixed pattern emerges. Consistent with Raval-lion’s hypothesis, eight of the 14 negativeexceptions had Gini coefficients above 45, the

Gini coefficient of all the negative exceptionsaveraged 44.9 and not a single negative excep-tion had a Gini coefficient lower than 30, Rav-allion’s informal standard for low inequality (p.11). Also consistent with Ravallion’s expecta-tions, four of the 13 positive exceptions (Ye-men, Norway 1989–95, Finland, and France)have Gini coefficients below 30 and two more(Nepal and Norway 1979–84) have Gini coeffi-cients under 31 (although Yemen and Nepalprobably have serious measurement errors, asnoted above). On the other hand, there are alsosix positive exceptions with Gini coefficientsabove 45 (mainly Latin American counties,plus Mauritania), and the Gini coefficients ofall the positive exceptions averaged above 38– not especially impressive. However, Raval-lion’s thesis would lead us to expect not onlythat low (high) Gini indices should induce amore (less) elastic relationship between eco-nomic growth and poverty, but also that adecreasing (increasing) inequality could explainpositive (negative) exceptions.

Observing the Gini index more dynamicallyreveals an even clearer pattern that is remark-ably consistent with Ravallion’s expectations.Of the positive exceptions, only two (El Salva-dor 1989–95 and Mauritania) show increasingGini indices—and El Salvador’s Gini index in-creased by less than a point over the entire per-iod. Moreover, eight of the positive exceptionsare among the ten cases which saw the greatestreduction of inequality. Moreover, every one ofthe negative exceptions showed worseninginequality; seven of these are among the top10 cases for increases in the Gini index. Theanalysis of the specific cases underscores therole that low inequality plays in some of the po-sitive exceptions, such as the social democraticNordic countries, as well as France. Highinequality is also related to some negativeexceptions, including Brazil, an example thatRavallion (2005) cites in his article. Moreover,redistributive policies can address or partiallyoverride the negative effects of inequality, as oc-curred for instance in Colombia and Peru.Thus, changes in inequality clearly played arole in the pathways of many of these excep-tional cases.

What of Dollar and Kraay’s second argu-ment (other than their contention that eco-nomic growth is good for the poor) thatliberal policies—which they measure throughlow inflation, low government consumption,high openness to trade, depth of the financialsystem, and establishment of rule of law—are


best for achieving the growth that reduces pov-erty? While this article does not address the roleof liberal policies for the cases nearer theregression line, the exceptional cases it doesstudy (some 10% of Dollar and Kraay’s sam-ple) provides little support for liberalization.Indeed, among the numerous cases in the sam-ple that liberalized, only two (Costa Rica andMauritania) were positive exceptions. WhileCosta Rica (1977–82) implemented an IMFprogram involving liberalization, this positiveexception provides only weak support becausethe country’s economy as a whole sufferedbadly over that five-year period. Indeed, CostaRica stands as one of the few countries, andone of three positive exceptions, in which theincome of the poor increased (by 2% per year)even as incomes as a whole declined (3.4% perannum), and the Carazo government’s eco-nomic management as a whole was considereddisastrous (e.g., Wilson, 1999). 11 While an-other positive exception, Mauritania, didimplement IMF policies linked to liberaliza-tion, these policies contributed to the countrybeing a positive exception in an unusual way.As argued above, these policies reduced aggre-gate demand, hampering economic growth, butdid not much affect the poorest of the ruralpoor. The poor meanwhile were greatly helpedby a more stable political environment andeffective government and the mining sector thathired many from their ranks. Thus, overall sta-bility and the opening of new opportunities thatcould be accessed by the poor were likely moreimportant in understanding the growth of poorincome. Meanwhile, several negative exceptionsadopted policies consistent with the liberalagenda by implementing shock therapy pro-grams (e.g., Post-Communist Eastern Euro-pean cases), restructuring their economies(e.g., Mali, the Dominican Republic, and Sin-gapore), or shifting sharply away from progres-sive to more liberal policies (e.g., Colombia1970–78). Moreover, a number of positiveexceptions implemented strategies in markedcontrast to liberal policies, including sometimesradical redistribution (e.g., Colombia 1964–70,Peru, and El Salvador 1989–95) and establish-ing social welfare programs of various types(e.g., France, Norway, and Finland). Thus,these exceptional cases suggest that the effectsof liberal policies on poverty and economicgrowth are at best contingent.

Analyzing four of Dollar and Kraay’s indica-tors of liberalization (due to data limitations,the fifth indicator, ‘‘rule of law’’, is constant

for all countries) for changes between the datesalso reveals mixed support for liberalization asa pathway explaining the exceptions. 12 Ifchanges in the degree of liberalism were to ex-plain the differences in residuals between thepositive and negative exceptions, we might ex-pect a systematic difference between these twotypes of cases in terms of changes in the fourindicators with which Dollar and Kraay mea-sure liberalism, and that such changes are sys-tematically related to the residuals. However,an ANOVA test indicates no significant differ-ence between the two groups of cases for anyof these indicators, and a regression analysisrun on these four variables against the residualsreveals a P-value of 0.932 for the entire sample.Thus, analyzed in aggregate, these indicators ofliberalization do not seem to explain the excep-tional cases.

Focusing on cases more specifically, we mightexpect that percentage changes in the four indi-cators of the positive exceptions would rankhighly compared to other cases. However, noneof the positive cases rank even in the top quin-tile of all cases for each of the indicators. ElSalvador (1989–95) comes closest, ranking 38of 231 for increasing trade openness, 5 of 248for lowering government expenditure, 17 of263 for lowering inflation and 66 of 232 fordevelopment of the financial sector. However,some of these changes (especially the first three)are likely to be caused by the end of the civilwar and the normalization of relations withmajor trading partners more than policies ofliberalization. Moreover, as argued above,since El Salvador’s government overtly redis-tributed wealth during this period, this case isnot a typical example of a liberalizing country,although this fact is not captured by these indi-cators. The only other positive exception thatapproximates conformity to liberalization isChile. However, while the country ranked 10of 184 for reductions in government expendi-tures and 46 of 175 for trade openness, Chiledid not rank even in the upper third for changesin the other two indicators during this period.Moreover, that Post-Pinochet Chile comes thisclose to representing a case of liberalization isironic, given the praise Milton Friedman andother liberal economists lavished on Pinochet’seconomic policies. Consistent with the analysisabove, other positive exceptions such as Hon-duras and Peru ranked near the bottom of lib-eralization. Ironically, all four indicators forCosta Rica, which implemented IMF liberaliza-tion policies, worsened with the changes in


indicators related to inflation and financial insti-tutions ranked near the bottom. To be sure,there are negative exceptions that score badlyon changes in liberalization—the authoritarianand closed governments of El Salvador (1977–89) and Brazil are not surprising in this regard.However, while Poland ranks near the bottomfor two indicators, lowered government spend-ing and development of the financial industry,Poland ranked 2 of 263 for increasing tradeopenness and 15 of 231 for lowered inflation,which is consistent with expert accounts thatthe country during this period typified liberal-ization. The negative exception of DominicanRepublic, while ranking poorly for trade open-ing and lowered inflation, ranked 13 of 248 forreduction of government expenditures and 34of 232 for its financial industry. Moreover, thenegative exception of Singapore ranks at thevery top for rule of law.

Two conclusions can be drawn from this re-view. First, this analysis underscores the weak-ness of these five commonly-used proxies ofliberalization. For instance, trade as a propor-tion of GDP can increase from liberalization,but it can also increase with circumvention ofsanctions, as is occurring with increasing ex-ports of raw materials between China and Afri-ca, or with the end of major conflict, as appearsto be the case for El Salvador. Rule of law canbe high in countries that are otherwise illib-eral—Singapore has high government con-sumption and a heavy reliance on state-ownedenterprises, but it ranks first for rule of law,for instance. Thus, one primary weakness isthat the variables, which purport to reflect theexpected results of liberalization, do not reflectwell the degree of liberalization of actual poli-cies implemented by a particular country. Sup-plementing these indicators, which emphasizethe effects of policies, with an analysis of thepolicies that were actually implemented, is war-ranted. Otherwise, based solely on the indica-tors of liberalization, one might conclude thatPost-Pinochet Chilean leaders actually liberal-ized, a conclusion with which Milton Friedmanmight have disagreed vehemently. Second,overall, there is little support for advocatingliberal policies irrespective of the political, eco-nomic, or social context. The state was active inmost positive exceptions (despite their differ-ences), implementing new policies and socialsupport program, strengthening institutions,and even applying violence on the poor’s be-half. While some also implemented some as-pects of a liberal framework (controlling

inflation, for instance), few of the positiveexceptions can truly be considered liberal eco-nomic systems. On the other hand, many casesinvolving states that were withdrawing from arole in the economic arena in a manner consis-tent with liberal approaches ended up beingnegative, not positive, exceptions.

This brief survey is incomplete for a numberof reasons. First, the database adopted forthese exercises, though covering the experiencesof an extensive range of countries over the pastseveral decades, omits many countries, includ-ing the poorest, for which comparable quanti-tative data are difficult to obtain. Whileadopting Dollar and Kraay’s data allows directcomparison with their analysis, in doing so thisstudy also adopts many of the data’s flaws.Moreover, limitations in space, time, andexpertise prevent the extensive exploration re-quired to establish strong causal linkages be-tween potential factors and changes inpoverty. Some exceptional cases were not ana-lyzed, while the analysis for others was neces-sarily brief. Despite the need for additionalwork to trace causal connections between gov-ernment policy and the economic effects, wecan nevertheless suggest that in many cases,government strategies and policies shaped andinfluenced the degree of economic growth andpoverty reduction, and the degree to which eco-nomic growth (or the lack thereof) influencedpoverty rates.

In addition, the factors that are identified inthis paper demand further attention. For in-stance, while some of the positive exceptionsare Scandinavian social democracies (Finlandand Norway), identifying them as such is insuf-ficient to explain these performances. Sincemany social democracies (e.g., Sweden andDenmark) and many time periods in Norwayand Finland were not exceptional under thisstudy’s strict criteria, further comparative re-search is needed to illuminate why some Scan-dinavian social democracies during certainperiods were exceptional, while others werenot. Similarly, some countries that restructuredtheir economies achieved strikingly different re-sults, with some (e.g., Costa Rica) becomingpositive exceptions and others (e.g., Mali)becoming negative ones. Identifying which fac-tors, in combination with restructuring, led tothese puzzling results requires a closer compar-ative review of the cases. Through qualitativemethods, we can discover the complex arrayof factors, whether political, economic, social,geographic, demographic, or otherwise, that


interacted with such policies to produce theseexceptional results.

One additional issue that should be addressedis the short-term nature of these cases (usually afive-year span) versus the sometimes long-termbenefits of economic growth. Those holdingthis view might argue that, valid as this article’sconclusions might be in the short-term, fewerexceptions might appear in the long-term be-cause the disjuncture between economic growthand poverty reduction narrows over time. Fol-lowing the principles behind Kuznets’s inverted‘‘U’’ (which predicts that, due to varied re-sponses from different economic sectors undereconomic growth, income inequality will ini-tially increase for a time as the economy grows,before subsequently declining), we might expectthat negative exceptions will occur whilegrowth is primarily benefiting the non-poor,until the gap closes over time as the benefitsof growth spread, possibly even creating posi-tive exceptions (Kuznets, 1955). While littleempirical evidence supports Kuznets’s inverted‘‘U’’ pattern (Kanbur, 2000), even without fol-lowing that strict pattern, growth can eventu-ally benefit the poor over the long-term.While this argument is plausible, many advo-cates of economic growth and liberal economicpolicies specifically reject ‘‘trickle-down’’growth as an important mechanism for povertyreduction, arguing that growth directly benefitsthe poor as much as other groups (e.g., Dollar& Kraay, 2002, p. 219). Moreover, few of thecases discussed here fall into a pattern consis-tent with Kuznets’s expectations. Colombia,for one, became a negative exception after—not before—it was a positive exception. El Sal-vador does show an inverted ‘‘U’’ pattern, butthis is explained by peace and redistribution,not by the dynamics that Kuznets predict. Inmany countries, such as China which in the1980s showed an upright ‘‘U’’ pattern, inequal-ity continues to expand, despite decades ofgrowth. While the image of a generation sacri-ficing for the benefit of their children is evoca-tive, in many cases, a generation of poorpeople toil for national development (or theenrichment of a narrow class), the benefits ofwhich are skewed away from the poor, whetherof their generation or the next. While economicgrowth may someday trickle down to the poor,the vague promises of such future gains may bea cold comfort.

Finally, we should consider the possibilitythat some of these cases might be explained be-cause poverty reduction policies came at the ex-

pense of economic growth. As discussed withinsome of the cases, this ‘‘trade-off’’ seems to ap-ply to a number of positive exceptions, includ-ing the Scandinavian social democracies, aswell as Peru, which implemented direct redis-tributive policies that alleviated poverty but re-duced productivity and slowed the growth rate.There appears to be no trade-off in other cases,including that of Chile (which grew faster in thepost-Pinochet era than it did before), CostaRica (which did not implement major policiesto help the poor that would conflict withgrowth), and possibly France (which faced anumber of external pressures on growth, suchas rising prices of oil and raw materials).Among some negative exceptions, economicgrowth sometimes came at the expense of theincome of the poor. Singapore’s stimulationof higher-order manufacturing, which requiredmore human capital than many poor peoplewere endowed with, increased growth at the ex-pense of the income of the poor. Poland,among the shock therapy countries, achievedpositive economic growth that traded off withjobs held by the poor. Colombia’s cancellationof many pro-poor policies increased its rate ofgrowth compared to the previous period (whenit was a positive exception), while the income oflowest quintile dropped precipitously. Many ofthese exceptions, therefore, do suggest that insome cases economic growth indeed comes atthe expense of poverty reduction, and some-times efforts to reduce poverty slows economicgrowth.

5. CONCLUSIONS

This research project seeks to identify path-ways to poverty reduction other than economicgrowth by focusing on exceptional cases. Manysocial scientists using regression and othertypes of quantitative methods often discardresidual cases as random anomalies that, whileinconvenient, do not detract significantly from,or contribute to, the overall conclusions.Granted, some exceptional cases will be ex-plained by important but uninteresting (to so-cial scientists at least) factors, such as weatheror natural disasters, that are largely out of hu-man hands. Data errors can create other appar-ent exceptions, such as the cases of Nepal andYemen. Caution is warranted, in any case,when applying insights from such exceptionsto other cases, as much damage has beencaused by blindly applying models generated


in certain spheres even when situations are dis-similar—advice that applies to policies derivedfrom all ideologies.

Despite these concerns, it remains possiblethat these unusual patterns are caused by socialactors who did something within the political,social, or economic realms to achieve excep-tional results. This suggests a general approachthat applies to more than the question of therelationship between economic growth andpoverty reduction: through disaggregatingaggregate statistical models, we can identifypotentially puzzling cases. Studying such anom-alies—explaining and understanding them—furthers theory, a key mission for social scien-tists. Studying exceptions opens up theoreticalspace, allowing us to seek other factors thatcontributed (in this case) to reducing poverty.What is more, analyzing such exceptions deep-ens our understanding of the connection be-tween economic growth and poverty reductionand the conditions under which this connectionapplies. Within the variation in Dollar and Kra-ay’s data lie clues to alternative paths to effec-tive poverty reduction—empirical models thatpolicymakers seeking to reduce poverty cancautiously emulate, carefully adapt, or con-sciously avoid. Studying such exceptional casesincreases our knowledge about what (besidespromoting economic growth) has worked ornot worked for reducing poverty. As two econ-omists argued:

‘‘. . . the general assertion that ‘growth is good for thepoor’ is not the most interesting way to interpret thisfinding. What is interesting is to identify commonfeatures of positive (negative) residuals—cases wheregrowth leads to substantially better (worse) povertyoutcomes than predicted from global regressions. Ifthese features can be linked to policies, there is a casefor switching toward policies that connect the poormore to growth. . .’’ (Eastwood & Lipton, 2000, p.40)

This paper answers this call. Its conclusionsdo not undermine a basic conclusion thateconomic growth and poverty reduction are of-ten related—to assert that they do fall into thetrap of radical falsificationism in which one orsome exceptions undermine otherwise generallyapplicable theoretical statements. Yet innumerous cases involving tens of millions ofpoor people, economic growth is divorced frompoverty rates. For negative exceptions, headyeconomic growth helped the poor less thanexpected or recessions hurt the poor signifi-cantly more than it did others. For positiveexceptions, the income of the poor increased

despite negative growth or poverty rates de-clined greatly in spite of modest economicgrowth.

The pathways identified herein are not new, asscholars have long cited many of these positiveexceptions for their unexpected accomplish-ments in reducing poverty to a greater extentthan economic growth alone could. Neverthe-less, they reinforce the view that while a com-plete list of viable pathways to povertyreduction should include economic growth(achieved through a number of ways) and liberaleconomic policies (which have reduced povertythrough growth and independently of growth),numerous other strategies that have been usedby countries to reduce poverty should also be in-cluded. It is true that none of these paths are‘‘systematic’’ mechanisms for poverty reductionin the sense that they do not apply ‘‘on average’’across all cases. However, the quest for system-atic solutions to poverty comes dangerouslyclose to a search for a panacea, part and parcelof the discredited attempt to confront complexand multifaceted social issues with a coveringlaw. There is no miracle lever—not economicgrowth, not the market, not the state—to reducepoverty, for poverty reduction is contingentupon the type of development that has beenachieved or policies implemented, and the im-pact of these on the livelihoods of the poor. Eco-nomic growth sometimes reduces poverty;sometimes, when poorly distributed, it leavespoverty untouched. Other times, when eco-nomic growth shifts market forces, capital,and property rights against the interests of thepoor, it can undermine the positions of the mostvulnerable of society, exacerbating poverty.When economic growth does benefit the poor,often it is insufficient or takes a long time incoming. As Nobel Lauriat Amartya Sen arguedlong ago (it is equally true today), ‘‘Not merelyis it the case that economic growth is a meansrather than an end, it is also the case that forsome important ends it is not a very efficientmeans either. . . It might well be the case that‘money answereth all things,’ but the answercertainly comes slowly’’ (Sen, 1983, p. 754).

More than one billion people still live on lessthan one dollar a day, and annual poverty-re-lated deaths exceed 18 million (Pogge, 2005).Given the contingent and sometimes long-de-layed benefits of economic growth for the poor,and decades of fruitless search for a magic bul-let, the imperative of discovering alternativepathways that work to reduce poverty remainsurgent.


NOTES

1. Ravallion (2001) notes that with a 95% confidenceinterval, a 2% growth rate in average household incomeswill reduce the poverty rate anywhere between 1% and 7%.

2. Dollar and Kraay’s conclusions are quite limited.Even if accepted at face value, Dollar and Kraay’s resultsdo not indicate that economic growth reduces eitherabsolute poverty, since they define poverty relativelywithin each country, or income gaps, since positiveoverall income growth that varies one-to-one with theincome of the poor actually increases the gap between richand poor (Eastwood & Lipton, 2001; Ravallion, 2001).

3. For a more thorough explanation of the dataset andtheir analysis, see Dollar and Kraay (2002). Because theresulting periods do not correspond to changes inadministrations or policies, the particular patterns ofyears impede analysis of the political causes of changes.This problem affects the authors’ attempts to assignvariables to factors such as degree of democracy andtrade openness, as well as my attempt to analyze whysome cases were exceptions.

4. Because Dollar and Kraay’s data exclude many poorcountries, there may be more exceptions.

5. The evidence related to the effects of land reform onpoverty in Colombia is mixed. Some analysts, notingthat the Gini coefficient for land holdings declined onlymodestly from 0.87 to 0.84 during the 1960–90s (Dein-inger, 1999), argue that land reform failed to make landholdings more equitable, in part because most of thesenew land titles were taken not from large farms but frompublic land (Dorner & Felstehausen, 1970). However,the date range of these analyses may be too broad: landreform was rolled back in the early 1970s. Subsequentresearch shows that the increase in the concentration ofland holdings, as Deininger (2004) subsequently noted,occurred after this rollback, primarily during mid-1980–90s. For instance, the share of land controlled by largerfarms increased from 46% to 54% during 1984–97.

6. Urban-to-rural migration increased substantially,probably contributing to these rates of growth. How-ever, the rate of increase in the urban population, whichgreatly outpaced expansion of urban jobs, limited thebenefit of migration for the income of the poor (Dorner& Felstehausen, 1970).

7. In 1989, remittances from relatives made up 3.5% ofthe country’s GDP. After the peace accord was signed,remittances increased rapidly, reaching US$686 million,or 8.1% of GDP, in 1992, and US$870 million, or 9.7%of GDP, in 1994 (Wood & Segovia, 1995).

8. This is consistent with poverty rates published by theWorld Bank that suggest that Poland’s poverty rateswere higher in 1998 than they were in 1991 (World Bank,2000).

9. This stability was short-lived, as President OuldTaya favored specific tribes and regions to a greaterextent than other presidents of the country had, a factthat contributed greatly to subsequent dissatisfactionwith the regime, and tribal-based corruption increasedsteadily over time.

10. Nepal’s government over the period in question forthe first time focused on reducing poverty by supportingagriculture. Starting in the Sixth Five-Year Plan (1980–85), Nepal spent liberally on promoting agriculture,placing emphasis on food production and developingcash crops, such as tobacco and sugar cane (Regmi,1997). Moreover, foreign aid from Japan and Europefocused on the agriculture sector and improving irriga-tion, watershed management, primary education andraising livestock. While there is doubt about Nepal’sstatus as a positive exception (its low growth rate islikely underestimated, as discussed above), these pro-grams likely helped significantly improve the income ofthe poor.

11. This result contrasts with that of other studies,which indicate that poverty in Costa Rica increasedduring 1977–83 from 16.1% to 30.5% (Rodriguez &Smith, 1994). This does not necessarily contradict Dollarand Kraay’s data since the poorest quintile’s incomecould increase even as the poverty rate, which exceeds20%, rises.

12. Unlike the case of inequality, where both theabsolute value of the Gini index and changes in thatindex should influence the formation of exceptions, itmakes more sense to study these four indicatorsdynamically, as such changes should represent theprocess of liberalization.

REFERENCES

Boyce, J. K. (1995). Adjustment toward peace: Anintroduction. World Development, 23(12), 2067–2077.

Bunce, V. (1999). The political economy of postsocial-ism. Slavic Review, 58(4), 756–793.


Chow, K. B., Lee, K. B., Hameed, A., & Cheong, B. C.(1988). Singapore business cycles. In Business cyclesin five ASEAN countries, India and Korea(pp. 175–192). Tokyo: Institute of Developing Econ-omies.

Coulombe, H., & McKay, A. (1996). Modeling deter-minants of poverty in Mauritania. World Develop-ment, 24(6), 1015–1031.

Danielson, A. (2001). When do the poor benefit fromgrowth, and why? Background paper to Sida’spoverty project. Working paper. Department ofEconomics, Lund.

Deininger, K. (1999). Making negotiated land reformwork: Initial experience from Colombia, Brazil andSouth Africa. World Development, 27(4), 651–672.

Deininger, K., Castagnini, R., & Gonzalez, M. A.(2004). Comparing land reform and land marketsin Colombia: Impacts on equity and efficiency.Working paper. World Bank, Washington, DC.

Derleth, J. W. (Ed.) (2000). The transition in Central andEastern European politics. Upper Saddle River, NJ:Prentice Hall.

Dollar, D., & Kraay, A. (2000). Growth is good for thepoor (preliminary and incomplete). Washington, DC:World Bank.

Dollar, D., & Kraay, A. (2002). Growth is good for thepoor. Journal of Economic Growth, 7(3), 195–225.

Dorner, P., & Felstehausen, H. (1970). Agrarian reformand employment: The Colombian case. InternationalLabor Review, 102(33), 221–241.

Dorward, A., Kydd, J., Morrison, J., & Urey, I. (2004).A policy agenda for pro-poor agricultural growth.World Development, 32(1), 73–89.

Eastwood, R., & Lipton, M. (2000). Pro-poor growthand pro-growth poverty reduction: Meaning, evi-dence and policy implications. Asian DevelopmentReview, 18(2), 22–58.

Eastwood, R., & Lipton, M. (2001). Pro-poor growth andpro-growth poverty reduction: What do they mean?What does the evidence mean? What can policymakersdo? (no. 19). Manila: Asian Development Bank.

Findlay, R. W. (1972). Ten years of land reform inColombia. Wisconsin Law Review, 3, 880–923.

George, A. L., & Bennett, A. (2005). Case studies andtheory development in the social sciences. Cambridge:MIT Press.

Glewwe, P. (1988). The distribution of welfare in Peru in1985–86 (no. 42). Washington, DC: World Bank.

Goertz, G. (2005). Social science concepts: A user’sguide (book manuscript), unpublished manuscript.

Gustafsson, B., & Li, S. (2004). Expenditures oneducation and health care and poverty in ruralChina. China Economic Review, 15(3), 292–301.

Gustafsson, B., & Pedersen, P. J. (Eds.) (2000). Povertyand low income in the Nordic countries. Ashgate:Aldershot.

Gustafsson, B., & Uusitalo, H. (1990). The welfare stateand poverty in Finland and Sweden from the mid-1960s to the mid-1980s. Review of Income andWealth, 36(3), 249–266.

Hudson, R. A. (Ed.) (1994). Chile: A country study (3rded.). Washington, DC: Library of Congress FederalResearch Division.

Kanbur, R. (2000). Income distribution and develop-ment. In A. B. Atkinson, & F. Bourguignon (Eds.).Handbook of income distribution (Vol. 1,pp. 791–842). Amsterdam: Elsevier Science B.V.

Kenworthy, L. (1999). Do social-welfare policies reducepoverty? A cross-national assessment. Social Forces,77(3), 1119–1139.

Kramer, M. (1995). Polish workers and the post-communist transition, 1989–93. Europe–Asia Studies,47(4), 669–712.

Kuznets, S. (1955). Economic growth and incomeinequality. American Economic Review, 45(1), 1–28.

Lee, W. K. M. (2001). The poor in Singapore: Issues andoptions. Journal of Contemporary Asia, 31(1), 57–70.

Leftwich, A. (2005). Democracy and development: Isthere an institutional incompatibility?. Democratiza-tion, 12(5), 686–703.

Levy, J. D. (1999). Vice into virtue? Progressive politicsand welfare reform in Continental Europe. Politicsand Society, 27(2), 239–273.

Lopez, R., & Valdes, A. (2000). Fighting rural poverty inLatin America: New evidence of the effects ofeducation, demographics, and access to land. Eco-nomic Development and Cultural Change, 49(1),197–212.

Marques, J. S. (2004). Operationalising pro-poor growth:A country case study of El Salvador. Washington,DC: World Bank.

McClintock, C. (1981). Peasant cooperatives and politicalchange in Peru. Princeton: Princeton UniversityPress.

Mellor, J. W. (1995). Agriculture on the road to indus-trialization. Baltimore: Johns Hopkins UniversityPress.

Molano, A. (2000). The Evolution of the FARC: AGuerilla Group’s long history. NACLA Report on theAmericas, 32(2), 22–31.

Øyen, E. (1996). Poverty research rethought. In E.Øyen, S. M. Miller, & S. A. Samad (Eds.), Poverty:A global review: Handbook on international povertyresearch. Oslo and Paris: Scandinavian UniversityPress.

Pazzanita, A. G. (1997). State and society in Mauritaniain the 1990s. The Journal of North African Studies,2(1), 16–39.

Peebles, G., & Wilson, P. (1996). The Singapore econ-omy. Cheltenham: Edward Elgar.

Pogge, T. (2005). World poverty and human rights.Ethics and International Affairs, 19(1), 1–7.

Puyana, A. (2000). Dutch disease, macroeconomicpolicies, and rural poverty in Colombia. InternationalJournal of Politics, Culture and Society, 14(1),205–233.

Ravallion, M. (2001). Growth, inequality and poverty:Looking beyond averages. World Development,29(11), 1803–1815.

Ravallion, M. (2005). Inequality is bad for the poor.Washington, DC: World Bank.

Ravallion, M., & Chen, S. (2007). China’s (uneven)progress against poverty. Journal of DevelopmentEconomics, 82, 1–42.

Regmi, S. K. (1997). Chapter IX: Nepal: Rural povertyalleviation under changing economic conditions. In


Regional expert meeting on capability-building toalleviate rural poverty. Beijing: United Nations Eco-nomic and Social Commission for Asia and thePacific.

Ritter, A. (1992). Development strategy and structuraladjustment in Chile: From the Unidad Popular to theConcertacion, 1970–92. Ottawa: North-South Insti-tute.

Rodriguez, A. G., & Smith, S. M. (1994). A compar-ison of determinants of urban, rural and farmpoverty in Costa Rica. World Development, 22(5),381–397.

Rodrik, D. (2000). Growth versus poverty reduction: Ahollow debate. Finance and Development, 37(4), 8–9.

Roett, R. (1999). Brazil: Politics in a patrimonial society(5th ed.). Westport, CT: Praeger.

Sachs, J. D. (1995). Shock therapy in Poland: Perspec-tives of five years. Salt Lake City: University of Utah.

Sen, A. (1983). Development: Which way now?. TheEconomic Journal, 93, 745–762.

Stewart, F., & Ranis, G. (1994). Occasional paper 14:Decentralization in Chile. <http://hdr.undp.org/

docs/publications/ocational_papers/oc14.htm#P2>Retrieved 30.12.06.

Toulmin, C., Leonard, R., Hilhorst, T., & Diarra, D.(2000). Mali poverty profile: IIED Drylands Pro-gramme.

Weyland, K. (1997). ‘‘Growth with Equity’’ in Chile’snew democracy?. Latin American Research Review,32(1), 37–67.

Wilson, B. M. (1999). Leftist parties, neoliberal policies,and reelection strategies: The case of the PLN inCosta Rica. Comparative Political Studies, 32(6),752–779.

Wood, E., & Segovia, A. (1995). Macroeconomic policyand the Salvadoran peace accords. World Develop-ment, 23(12), 2079–2099.

World Bank (2000). Making transition work for every-one: Poverty and inequality in Europe and CentralAsia. Washington, DC: World Bank.

Zhang, L., Huang, J., & Rozelle, S. (2003). China’s waron poverty: Assessing targeting and the growthimpacts of poverty programs. Journal of ChineseEconomic and Business Studies, 1(3), 301–317.

.sciencedirect.com
Available online at www
http://hdr.undp.org/docs/publications/ocational_papers/oc14.htm#P2

http://hdr.undp.org/docs/publications/ocational_papers/oc14.htm#P2

growth is good for whom, when, how? economic growth and poverty reduction in exceptional cases

Documents