Trade and Regional Inequality

Andrés Rodríguez-PoseDepartment of Geography

and EnvironmentLondon School of

EconomicsHoughton StreetLondon WC2A 2AEUKa.rodriguez-pose@lse.ac.uk

Key words:traderegional inequalitydeveloping and developed

countrieslow- and middle-income

countries

abst

ract This article examines the relationship between open-

ness and within-country regional inequality across 28countries over the period 1975–2005. In particular,it tests whether increases in trade lead to risinginequalities, whether these inequalities recede intime, and whether increases in global trade affect thedeveloped and developing worlds differently. Usingstatic and dynamic panel data analysis, I found thatwhile increases in trade per se do not lead to greaterterritorial polarization, in combination with certaincountry-specific conditions, trade has a positiveand significant association with regional inequality.States with higher interregional differences in sec-toral endowments, a lower share of governmentalexpenditures, and a combination of high internaltransaction costs with a higher degree of coincidencebetween the regional income distribution andregional foreign market access positions have expe-rienced the greatest rise in territorial inequality whenexposed to greater trade flows. Hence, changes intrade regimes have a more polarizing and enduringeffect in low- and middle-income countries whosestructural features tend to enhance the trade-inequality effect and whose levels of internal spatialinequality are, on average, significantly higher than inhigh-income countries.ecge_1147 109..136

109

ECO

NO

MIC

GEO

GR

APH

Y88(2):109–136.©

2012C

larkU

niversity.w

ww

.economicgeography.org

Acknowledgments

I am grateful to BenjaminFaber for his superb researchassistance and support withthe econometric analysis andto Vassilis Tselios and DiegoPuga for further econometricsuggestions. I am alsoindebted to Thomas Farole,Olivier Cadot, SouleymaneCoulibaly,Ana Paula Cusolito,Bernard Hoekman, andparticipants at a seminar atthe World Bank inWashington, D.C., forcomments on an earlierdraft.This article representsa contribution to a widerproject on “Trade andSubnationalCompetitiveness” by theWorld Bank’s InternationalTrade Department and hasgreatly benefited from thesupport of the EuropeanResearch Council under theEuropean Union’s SeventhFramework Programme(FP7/2007-2013)/ERC grantagreement 269868.Additional support from aLeverhulme Trust MajorResearch Fellowship andfrom the PROCIUDAD-CMproject is also gratefullyacknowledged.The usualdisclaimer applies.

The World Bank’s (2009) World DevelopmentReport (WDR), Reshaping Economic Geography, puttrade at the heart of the holy trinity of factors thatpromote growth. “Cities, migration, and trade havebeen the main catalysts of progress in the developedworld over the past two centuries [and] these stories arenow being repeated in the developing world’smost dynamic economies” (World Bank 2009, 20).Although promoting trade is acknowledged to lead togreater territorial disparities, it may not matter in themedium- and long term, since “evidence from today’sindustrial countries suggests that development haslargely eliminated rural-urban disparities” (WorldBank 2009, 62). Hence, from this perspective, thebest way to deal with territorial inequality is notthrough “spatially balanced growth,” which has been a“mantra of policymakers in many developing coun-tries” (World Bank 2009, 73), but through the promo-tion of growth resulting from increases in trade andeconomic integration.

This approach to promoting economic develop-ment rests on three assumptions for which the scho-larly literature has provided no firm answer, namely,that (1) increases in trade lead to rising territorialinequalities; (2) these inequalities subsequentlyrecede as a country develops; and (3) the emergenceof spatial disparities does not represent a threat tofuture development, implying that developing coun-tries should be more concerned about the promotionof growth, rather than worry about inequalities (WorldBank 2009, 12). However, and despite the surgeof attention on the relationship among globaliza-tion, the rise of trade, and inequality, whether theseassumptions hold remains unanswered.

Most of the work conducted so far on the linkbetween trade and inequality has been concerned withthe impact of the increasing integration of globalmarkets on interpersonal income inequality, both inthe developed and the developing worlds (e.g., Wood1994; Ravallion 2001; Alderson and Nielsen 2002;Williamson 2005). The spatial dimension of inequa-lity, in contrast, has attracted far less attention. Thus,as Kanbur and Venables (2005) underlined, boththe theoretical and empirical relationship betweengreater openness and spatial inequality remainsvague (see also Brülhart 2011). Almost as manystudies have pointed to a link between trade andspatial convergence as have identified spatial diver-gence (Brülhart 2011), and the direction and dimen-sion of this relationship is far from uniform and vary

ECONOMIC GEOGRAPHY

110

from one country to another and according to the data and methods that have beenused.

Although the number of single-country case studies that have delved into this issue hasgrown significantly in recent years, scant cross-country evidence exists of a general causallinkage between greater trade openness and market integration, on the one hand, andintranational spatial inequality, on the other.1 This may be the case because the literatureon the evolution of within-country spatial inequalities has tended—following the pathopened by Williamson (1965) in his account of the relationship between spatial disparitiesand the stage of economic development—to focus on the internal, not the external, forcesof agglomeration and dispersion. From this perspective, economic development mattersfor the evolution of spatial inequalities, which tend to wane as a country develops. Hence,the factors that explain the evolution of regional inequality have been considered to beinternal to the country itself, while external factors have been regarded, at best, as playinga supporting role in this process. And when external factors have been taken intoconsideration, the outcome has been inconclusive. As Milanovic put it (2005, 428),“country experiences differ, and . . . openness as such may not have the same discernableeffects on countries regardless of their level of development, type of economic institu-tions, and other macroeconomic policies.” Moreover, a large percentage of the literatureon the relationship between trade and spatial inequality has concentrated on developedcountries—particularly on the spatial effects of integration into the EuropeanUnion (EU) (e.g., Niebuhr 2006; Barrios and Strobl 2009)—meaning that thefindings, as inconclusive as they are, may be irrelevant to middle- and low-incomecountries.

Finally, it is far from certain that the supposed temporality and benign long-termimplications of any potential growth in within-country regional disparities resulting fromchanges in trade patterns will materialize. In particular, any rise in inequalities may riskbecoming permanent, especially when increasing polarization takes place during periodsof low growth or in countries with an already high level of territorial disparities. In thesecases, increases in trade may come to reinforce preexisting social, political, cultural, orethnic divides. Under these circumstances, greater regional inequality may lead to anenduring fragmentation of internal markets and to social, political, and/or ethnic tensionsthat may threaten the very growth and prosperity that greater trade is supposed to bringabout.

This article delves into the assumptions about the link between trade and regionalinequality present in the WDR (World Bank 2009) and for which the literature has offeredno conclusive answers. More specifically, its aim is to determine whether changes in tradematter for the evolution of spatial inequalities, whether openness to trade affects deve-loped and developing countries differently, and whether there is a dynamic element tothis association. The analysis covers the evolution of regional inequality across 28countries—including 15 high-income and 13 emerging countries—over the period 1975–2005.

To achieve this goal, the article combines the analysis of internal factors (in thetradition of Williamson 1965) with that of change in real trade as a potential externalfactor that may affect the evolution of within-country regional inequality. The internalfactors that are considered include both Williamson’s (1965) level of development and aseries of other factors identified by the new economic geography theory (NEG) as drivers

1 Brülhart (2011) limited the number of cross-country analyses to 11, virtually all using data on urbanprimacy (e.g., Ades and Glaeser 1995; Nitsch 2006; Brülhart and Sbergami 2008), rather than regionaldata.

Vol. 88 No. 2 2012

111

TR

AD

EA

ND

REG

ION

AL

INEQ

UA

LITY

of spatial inequality. I conducted the analysis by running unbalanced static panels withcountry and time fixed effects to address whether changes in trade patterns have beenconnected to changes in spatial inequalities in the short term and whether this process hasaffected the developed and emerging worlds differently. The second part of the analysisconsists of a dynamic panel estimation, differentiating between short-term and long-termeffects, as a way to assess whether this relationship changes with time.

The article is structured in five additional sections. The second section provides anecessarily brief overview of the theoretical and empirical literature. The third sectionpresents the data and its main trends, followed by the fourth section, which outlines thetheoretical framework and presents the variables that were included in the analysis. Thefifth section reports the results of the static and dynamic analyses, distinguishing betweenthe differential effect of trade on regional inequality in developed and developingcountries, and presents a series of checks for robustness. The conclusions are presented inthe final section.

Trade and Regional Inequality in the LiteratureAs I mentioned in the introduction, the link between changes in trade and the evolution

of regional disparities has hardly captured the imagination of geographers and econo-mists. In contrast with the spawning literature on trade and interpersonal inequality, therehas been, until recently, a dearth of studies on the within-country spatial consequences ofchanges in trade patterns. The emergence of NEG theory has somewhat alleviated this gapin the literature, especially from a theoretical perspective. A string of NEG modelsconcerned with the spatial implications of economic openness and trade (e.g., Krugmanand Livas Elizondo 1996; Monfort and Nicolini 2000; Paluzie 2001; Crozet and Koenig-Soubeyran 2004; Brülhart, Crozet, and Koenig 2004) have been published in recentyears. In this literature, the causal effect of globalization on the national geographyof production and income is conceptualized in terms of changes in cross-borderaccess to markets that affect the interplay between agglomeration and dispersionforces, which, in turn, determine the dynamics of industrial location across domesticregions.

Because of the two-sector nature (agriculture and manufacturing) of most of thesemodels, the central question has been whether increasing cross-border integration leads toa greater intranational concentration of manufacturing activity and thereby growingregional inequality. However, as a consequence of the use of different sets of assumptionsand of the particular nature of the agglomeration and dispersion forces included in themodels (Brülhart et al. 2004), contradictory and/or ambiguous conclusions have beenderived from this type of analysis (e.g., Krugman and Livas Elizondo 1996 versus Paluzie2001).

Empirical studies have not contributed to resolving this conundrum. Most of theempirical analyses have concentrated—in part as a result of the scarcity and lack ofreliability of subnational comparable data sets across countries—on single-country casestudies. Two countries have been featured prominently in empirical approaches. The firstis postreform (post-1978) China, where an expanding number of studies have focused,inter alia, on the trade-to-gross domestic product (GDP) ratio and/or inflows of foreigndirect investment (FDI) to explain either overall regional inequality or the growingcoast-inland divide (Jian, Sachs, and Warner 1996; Yang 2002; Zhang and Zhang 2003;Kanbur and Zhang 2005). These studies have tended to run time-series ordinary least-squares (OLS) regressions with the measure of provincial inequality on the left-hand sideand openness to trade and/or investment, among a list of variables, on the right. Most of

ECONOMIC GEOGRAPHY

112

these analyses have found a significant positive effect of the rise of trade on regionalinequality. The second country is Mexico. Using a number of measures, ranging fromchanges in trade ratios (Sánchez-Reaza and Rodríguez-Pose 2002; Rodríguez-Pose andSánchez-Reaza 2005), sometimes controlling for location and sector (Faber 2007), FDI(Jordaan 2008a, 2008b), retail sales (Adkisson and Zimmerman 2004), or retail trade(Ford, Logan, and Logan 2009), studies of Mexico have generally reported that increasesin trade and greater economic integration in the North American Free Trade Agreement(NAFTA) have resulted not just in important differences in the location of economicactivity, but in significant rises in wages (Chiquiar 2008) and income inequality betweenborder regions and the rest of Mexico. Other interesting single-country research hasfocused on the Canadian case, finding that rising competition in imports from low-incomecountries has had an important effect on wage inequalities, especially in Québec and thePrairie provinces (Breau and Rigby 2010).

Cross-country panel data analyses examining the link between changes in tradepatterns and the evolution of regional disparities have been rarer. A large number ofthese studies has been concerned with the impact of European integration on tradepatterns and how these patterns, in turn, influence regional inequality. Among thesestudies, Petrakos, Rodríguez-Pose, and Rovolis (2005) and Barrios and Strobl (2009) canbe highlighted. Petrakos et al. (2005) used a measure of relative intra-European integra-tion for a sample of eight EU member countries, measured as national exports plusimports to and from other EU countries divided by total trade, rather than overalltrade-to-GDP ratios. Running a system of seemingly unrelated equations, they foundmixed explanatory results for this variable and concluded that European integrationaffects countries differently. Barrios and Strobl (2009) ran fixed effects OLS analyses forthe EU15 over the period 1975–2000. Their aim was to explain how a measure of regionalinequalities within each country is influenced by the trade-to-GDP ratio, as well as bytrade over GDP in terms of purchasing power parities (PPP). For the latter, they found asignificant positive effect on regional inequalities among EU15 countries over the periodunder study.

Fewer studies have covered a more diverse sample of countries, both developed anddeveloping. Two such studies were Milanovic (2005) and Rodríguez-Pose and Gill (2006).Milanovic addressed the evolution of regional inequalities across the five most populouscountries of the world—Brazil, China, India, Indonesia, and the United States—overvarious time spans from 1980 to 2000. The results of his static fixed-effects and dynamicArellano-Bover panel analyses pointed to the absence of a significant causal relationshipbetween openness and regional inequalities. Rodríguez-Pose and Gill (2006) mapped twosets of binary relationships, first between nominal openness to trade and regional inequal-ity and second between a trade composition index and regional inequality, for eightcountries—Brazil, China, Germany, India, Italy, Mexico, Spain, and the UnitedStates—over various time spans between 1970 and 2000. They concluded that it is nottrade openness per se that has any bearing on the evolution of regional inequality, but itscombination with the evolution of the manufacturing-to-agriculture share of exports, thatinfluences which regions gain and which lose from greater economic integration overtime. As trade shifts from the primary sector to manufacturing, because manufacturing ismore geographically concentrated, especially in emerging countries, than agriculture ormining, within-country regional disparities tend to increase and do so at a faster pace inthe developing than in the developed world. Rodríguez-Pose and Gill (2006) foundindicative support for this hypothesis on the basis of the coincidence between changes inthe evolution of their trade composition index and changes in regional inequalities acrosscountries.

Vol. 88 No. 2 2012

113

TR

AD

EA

ND

REG

ION

AL

INEQ

UA

LITY

Given the diversity of results in both theoretical and empirical analyses, one would behard pressed to generalize from the literature. The relationship between trade and regionalinequalities thus remains wide open, from both a theoretical and an empirical perspective.

Overall Trade and Regional Inequality: Empirical EvidenceThis article revisits the question of the link between trade and regional inequality using

an unbalanced panel data set comprised of 28 countries from 1975 to 2005. The 28countries included 11 countries in Western Europe, 6 in Central and Eastern Europe, 4developed countries outside Europe, and 7 countries in the developing world (seeTable 1). The criterion for the choice of countries was based exclusively on the existenceof reliable and comparable regional GDP per capita data series. This criterion tended tobias the sample toward developed countries, in general, and developed countries in theEU, in particular, raising potential questions about whether the results based on such asubset of countries can be generalized. Although this is a reasonable concern, the databaseI used is significantly more comprehensive than those used in previous studies of regionalinequality from a comparative perspective.2 In addition, with 13 of the 28 countries in thesample among the ranks of emerging countries—9 are still classified as low- or middle-income countries by the World Bank, and 4 others (the Czech Republic, Hungary,Poland, and Slovakia) have only recently been reclassified as high-incomecountries—it could be claimed that emerging economies are well represented in thedatabase.

Another factor that limited the choice of countries is that the number and nature ofregional units should not vary over time. In countries with a certain tradition of politicalor fiscal decentralization, I chose the subnational units with the greatest level of

2 Barrios and Strobl (2009), for example, limited their analysis to European countries, while Rodríguez-Poseand Gill (2006) covered four developed and four emerging countries.

Table 1

Increasing Versus Decreasing Regional Inequality

Increasing Regional Inequality Stable Regional Inequality Decreasing Regional Inequality

Australia (1990–2005) Austria (1988–2004) Belgium (1977–1996)Bulgaria (1995–2004) Canada (1981–2005) Brazil (1989–2004)Czech Republic (1995–2004) China (1978–2004) South Africa (1995–2005)Finland (1995–2004) Italy (1995–2004)France (1982–2004) Japan (1975–2004)Greece (1979–2004) Netherlands (1986–2004)Hungary (1995–2004) United States (1975–2005)India (1993–2002)Indonesia (2000–05)Mexico (1993–2004)Poland (1995–2004)Portugal (1995–2004)Romania (1998–2004)Slovakia (1995–2004)Spain (1980–2004)Sweden (1994–2004)Thailand (1994–2005)United Kingdom (1994–2004)

ECONOMIC GEOGRAPHY

114

autonomy, including states in countries, such as Austria, Brazil, India, Mexico, and theUnited States; provinces in Canada, China, and South Africa; and regions in France, Italy,and Spain. In countries with no tradition of decentralization, I chose administrativedivisions—generally NUTS2-level regions for EU countries. In a few cases in which thenumber of regions was limited, I resorted to smaller territorial units.3 In all cases, thenumber of territorial units remained stable during the period under analysis, eitherdisregarding the creation of new states or provinces, such as the provinces of Gorontalo,West Sulawesi, and West Papua in Indonesia, or, more exceptionally, extrapolating theGDP per capita of new states to periods prior to their creation, as in the cases of therecently created Indian states of Chhattisgarh and Jharkhand. Keeping the number ofterritorial units stable, however, limited the time span of the analysis in countries that haveundergone a complete overhaul of their territorial systems, such as most Central andEastern European countries after the fall of the Iron Curtain or South Africa after thedemise of apartheid. Changes in regional accounting systems provide an additionalbarrier for the length of the data series I considered. The number and denomination ofterritorial units for each country is included in Table A1 in the appendix. Table 1 groupsthe countries in the sample according to whether they experienced increasing, stable, ordecreasing spatial disparities, using the evolution of the population-weighted coefficientof variation,4 over the time span covered by the data.

The majority of the countries in the sample experienced a rise in regional disparitiesover the period of analysis. In 18 of the 28 countries, spatial inequalities increased, while7 countries witnessed relative stability,5 and only three—Belgium, Brazil, and SouthAfrica—saw a reduction in disparities. The rate of change varied enormously acrosscountries (see Figure 1). Countries like Bulgaria, Hungary, India, Poland, Romania, theSlovak Republic, and China witnessed a rapid rise in disparities from the early 1990sonward, whereas the rate of increase was more moderate in such places as Australia,Spain, the United Kingdom, and the United States. Rates of decline in inequalities alsovaried widely, with Belgium and Brazil experiencing the strongest decline in territorialinequalities. During the period under study, there was also no apparent difference betweenthe trajectories of developed and emerging countries. Some of the low- and middle-income countries in the sample (Bulgaria, India, Indonesia, Mexico, and Thailand) sawspatial disparities increase, but this was not the case in Brazil and South Africa (seeFigure 1). However, the level of territorial inequalities differed widely among countries,and especially between countries in the developed and developing worlds. Regionaldisparities in Thailand were 19 times higher than those in Australia (see Figure 1). Theorder of magnitude was 10 to 1 between China and Mexico, on the one hand, and

3 One such example is Belgium, where the 3 regions (Brussels, Flanders, and Wallonia) were substituted by10 provinces and the city of Brussels.

4The population-weighted coefficient of variation, c, is defined by the equation c

p xi i=−( )μ

μ

2

where xi and pi are, respectively, the per capita income and the population share of region i (i = 1, 2, . . . , n)

in a given year, while μ ==

∑ p xi i

i

n

1

.

5 It is often the case that overall trends in stability during the period of analysis hide significant variations inthe evolution of regional inequality. Two such cases are Canada and China. In both countries, albeit fordifferent reasons, regional disparities decreased during the 1980s, but tended to grow and, in the case ofChina, particularly rapidly—from the early 1990s.

Vol. 88 No. 2 2012

115

TR

AD

EA

ND

REG

ION

AL

INEQ

UA

LITY

Australia, on the other, and 7 to 1 in the case of Brazil and India relative to, once again,Australia.

Is there a general relationship between the evolution of openness to trade and spatialinequalities? To assess whether this is the case, I performed a simple binary associationbetween annual measures of real openness to trade and regional inequality for eachcountry. Figure 2 plots the regression coefficient of the log Gini index of regional GDP per

Figure 1. Evolution of regional inequality in a selected sample of countries (measured by thepopulation-weighted coefficient of variation).

ECONOMIC GEOGRAPHY

116

capita6 on the log of the share of exports plus imports in GDP adjusted to PPP by country.In Figure 3, the same regression coefficients are presented, having replaced the annualmeasures by three-year averages, since multiannual averages may be better than annualdata at picking up any potential lagged effects, thus correcting for yearly fluctuations.

Figures 2 and 3 show no dominating pattern. There is a huge diversity in both the signand the dimension of the coefficient, with some countries sporting a positive relationshipbetween trade and the evolution of regional disparities and others a negative one. Thereconsequently seems to be, as Milanovic (2005) and Rodríguez-Pose and Gill (2006)indicated, no evidence of the presence of a simple linear relationship between trade andregional inequality that holds across different types of countries. A more subtle observa-tion concerns the sequence of countries from left to right. On the whole, wealthiercountries (Finland, Sweden, Canada, the Netherlands, and Japan) tend to be located on theleft-hand side of both figures, displaying a negative association between increases in tradeand regional disparities, while poorer countries (India, Romania, and Poland) tend to befound toward the right-hand side of the figures. This relationship is far from linear,however, with some high-income countries (Spain, Italy, United Kingdom, and Greece)displaying a positive binary association between trade and spatial inequality.

Theoretical Model and Empirical VariablesThere are limitations, however, in what can be inferred from simple binary associa-

tions, since these associations offer only limited information about the mechanisms at

6 The Gini index is defined as follows: G

p p x xi j i j

j

n

i

n

=−

==∑∑

11

2μ, where x

iand p

iare, respectively, the per

capita income and the population share of region i, while μ ==

∑ p xi i

i

n

1

.

Figure 2. Regression coefficients of regional inequality on real trade openness.

Vol. 88 No. 2 2012

117

TR

AD

EA

ND

REG

ION

AL

INEQ

UA

LITY

play. Many other factors may affect the evolution of within-country regional disparities.To address this issue, I present a formal econometric specification with additional controlsand conditioning variables to test whether there is a significant association betweenopenness and spatial inequality and whether this association—if it exists—changes withtime and affects developed and developing countries differently.

The Basic ModelWith few exceptions (e.g., Milanovic 2005), the bulk of studies on the determinants of

regional inequalities have been based on static one-year specifications. However, regionalinequality is bound to be a time-persistent phenomenon with a high degree of inertia,which makes overlooking time considerations problematic. Theory, however, provides noclear (if any) insights into the temporal dimension of internal spatial adjustments tochanges in access to external markets. Hence, rather than guess an appropriate adjustmenttime frame, I tackled potential inertia by formulating a dynamic model with past levels ofspatial inequality on the dependent variable side. The use of dynamic panels, whichcomplement static panels, has the advantage of introducing the distinction betweenshort-term and long-term effects.

In view of the foregoing discussion, the following general model is formulated:

Inequality*it it= + ∑ +α β εxit (1)

where Inequality*it is the level of inequality in country i at time t and xit is a vector of

independent variables conditioning the spatial distribution of income in any given countryi at time t. This model does not contemplate inertia in the system. Using Brown’s (1952)classical habit persistence model, equation (1) is transformed into equation (2):

Inequality Inequality Inequality* Inequalityit − = −(− −it it it1 1λ )) < <, 0 1λ (2)

Figure 3. Regression coefficients of regional inequality on openness for three-year averages.

ECONOMIC GEOGRAPHY

118

where the actual observed change of the spatial configuration (Inequalityit - Inequalityt-1)is a fraction l of the adjustment that would have taken place under instantaneousadjustment.

Parameter l ranges between 0 and 1 and represents the speed of adjustment. If l is closeto 1, then the adjustment is almost instantaneous and the relationship between thetheoretical determinants xit and the actual observed spatial outcomes Inequalityit is static.If l is less than 1, then the difference between the observed spatial outcomes and theirinertia-free theoretical counterpart Inequality*

it becomes significant, creating the need tocontrol for partial adjustment in a dynamic model. Rearranging and substituting forInequality*

it, one obtains:

Inequality Inequalityit it it= + ∑ +( ) + −( ) < <−λ α β ε λ λxit 1 0 11, (3)

Equation (3) presents the basic specification followed in the dynamic panel regressions.On the left side of the equation is the dependent variable, representing the observedinequality. On the right are the theoretical determinants of the inertia-free spatial con-figuration plus the previous period’s value of the dependent variable. The latter effectivelycontrols for potential inertia and partial adjustment. By fixing the previous spatialoutcome Inequalityit-1, the short-term effect of any independent variable xit is given by itsrevealed regression coefficient when running equation (3). Conceptually, this coefficientrepresents the product lb. The assumption for the long run is that a country’s spatialconfiguration reaches a stable equilibrium, making the current and previous year’sinequality levels close to identical. Setting Inequalityit-1 equal to Inequalityit in equation(3), the long-term effect of any independent variable on the spatial configuration can thusbe obtained by dividing the observed regression coefficient lb by the speed of adjustmentparameter l.

The Conditioning VariablesHaving set the basic model, the next task was to identify an appropriate set of

conditioning variables that capture the relationship between openness to trade andinternal spatial inequality in the form of equation (1). I did so in two stages. The first onedraws on recent NEG models, while the second reaches beyond a purely market access-driven framework.

In an NEG core-periphery framework and as a consequence of NEG’s basic two-sectorassumption and the absence of intraindustry linkages, distinguishing whether or notgreater accessibility to foreign markets promotes economic growth is tricky. The intro-duction of cross-border intraindustry linkages and a multisector industrial scenario givesrise to additional pull factors toward highly accessible regions once trade is liberalizedand allows export market potential, intraindustry supply potential, and import competi-tion to affect domestic sectors differently, depending on the comparative advantagesrevealed by market integration (Faber 2007). Sectors that are characterized by a revealedcomparative advantage and/or cross-border intraindustry linkages will thus grow faster inregions with good access to foreign markets, whereas import-competing sectors gain inrelative terms in regions with higher “natural protection,” in the form of poor access tomarkets. Faber (2007) found empirical support for this trade-location linkage across 43industrial sectors in post-NAFTA Mexico from 1993 to 2003.

The implications of this possible divergence of sectoral location patterns under cross-border market integration are important to understand whether and how accessibility tomarkets affects regional performance. Regions with high relative access to foreign

Vol. 88 No. 2 2012

119

TR

AD

EA

ND

REG

ION

AL

INEQ

UA

LITY

markets will attract the winners of integration and shed declining sectors, resulting inhigher medium- to long-term regional growth rates than in regions with limited and/orconstrained access to foreign markets.

In conditions of increasing trade and economic integration, two additional country-specific factors may determine the evolution of regional inequalities. The first is thedegree of variation of the accessibility of foreign markets among regions within any givencountry. If, given the foregoing discussion, it is assumed that relative access of foreignmarkets drives regional attractiveness for expanding sectors, the locational pull will be thestrongest in countries that are characterized by stark regional differences in cross-borderaccessibility of markets. The strength of the variation in accessibility is further condi-tioned by a second factor, namely the degree of coincidence between the existing regionalincome distribution and the distribution of relative access to foreign markets. Whenrelatively wealthy regions are also those with a greater degree of accessibility, increasesin trade are likely to exacerbate previously existing inequalities. In contrast, when poorerregions have a market-accessibility advantage relative to better-off regions, the netoutcome of increases in trade is likely to be a reduction in regional disparities andwithin-country territorial convergence. Hence, it can be safely assumed that greater tradeopenness will have a more polarizing effect in countries that are characterized by higherdifferences in accessibility to foreign markets among its regions and a high degree ofcoincidence between the regional income distribution and accessibility to foreignmarkets. The presence of a strong coincidence between regional income distribution andaccessibility to foreign markets is a sufficient, rather than a necessary, condition forgenerating greater inequality because openness to trade may also exacerbate previouslyexisting inequality even in cases in which wealthier regions have less accessibility toforeign markets than do poorer regions. Differences in endowments or in adaptivecapacity between rich and poor regions may more than compensate for differences inaccessibility.

Stepping outside the NEG framework, other factors may come into play in determiningthe link between trade and regional inequality. Among these factors, differences in thedistribution of human capital and skills and infrastructure affect trade patterns as well aseconomic growth. It can therefore be envisaged that the greater the regional differences inendowments and sectoral specialization, the greater the spatial impact of openness totrade.

The role of governmental policies may also enhance or attenuate the spatial effects oftrade. Governments with a greater social and territorial redistributive capacity throughpublic policies will be in a better position to counter any potential tendency of increasesin trade patterns to lead to greater geographic polarization. Budgetary or regional policytransfers from prosperous to lagging regions will offset rises in regional inequality,making the effect of openness to trade on spatial inequality likely to be more severe incountries with a weaker redistributive capacity by the central government and/or withfewer provisions for interregional transfers.

The fourth conditioning factor concerns the degree of labor mobility, especiallywithin-country mobility. Depending on the conditions of any particular country, themobility of interregional workers may contribute either to greater agglomeration, asworkers concentrate in core areas that offer higher salaries or greater job opportunities, orto greater territorial cohesion, if workers follow firms that are seeking lower costs inperipheral areas (Puga 1999). Hence, the effect of trade on regional inequality will dependon the degree of interregional labor mobility and the specific conditions of the country.

The final factor is the quality of institutions, which vary significantly from one regionto another. Poorer and/or lagging regions are likely to suffer the most from inadequate

ECONOMIC GEOGRAPHY

120

institutions. Problems of institutional sclerosis, clientelism, corruption, and pervasiverent seeking by durable local elites, which beset many lagging areas, are likely tocontribute to trade bypassing these regions in favor of those with more “appropriate”institutions. “Informal institutions in these places are often similarly dysfunctional,resulting in low levels of trust and declining associative capacity, and restricting thepotential for effective collective action” (Farole, Rodríguez-Pose, and Storper 2011,1098). “Inappropriate” institutions will thus represent an important barrier to trade,leading to a more severe spatial effect of trade in countries with a significant gap ininstitutional capacity among their regions.

Unfortunately, because of the lack of comparable and reliable data on interregionallabor mobility and institutions across the 28 countries covered in the analysis, the lattertwo hypotheses could not be tested. It therefore has to be assumed that labor mobility andinstitutions are not systematically correlated with any of the other regressors, implyingthat there is no omitted variable problem in leaving out these conditioning interactions.

There is also a need to control for other factors that may affect the relationship betweentrade and spatial inequality. The key element in this realm is related to Williamson’s(1965) classical account of the linkage between spatial disparities and the stage ofeconomic development. In Williamson’s account, within-country spatial inequalities arefundamentally the result of the level of national economic development (proxied in thiscase by real GDP per capita and its growth). As countries prosper, inequalities tend todiminish, making economic growth a primary driver of changes in spatial inequalities.Williamson’s theory is built into the WDR (World Bank 2009), which stated that not onlyhas “development . . . largely eliminated rural-urban disparities” (62) but also that “highurban shares and concentrated economic density go hand in hand with small differencesin rural-urban well-being on a range of indicators” (62). Since economic growth is alsolikely to be correlated with changes in trade (Sachs and Warner 1995), a control for realGDP per capita and its interaction with the country’s developmental stage is included inthe analysis.

The Empirical Model, Data, and MethodThe foregoing discussion leads to the transformation of equation (1) into the following

empirical specification (4). Table A1 in the appendix presents the actual values of thestructural conditions across the 28 countries.

ln ln lnInequality* GDPcap GDPcap Developmenit it it= + ( ) + ( ) ⋅α β β1 2 ttTrade Trade DevelopmentTrade

i

it it

i

[ ]+ ( ) + ( ) ⋅[ ]+

β ββ

3 4

5

ln lnln tt i it i( ) ⋅ ( )[ ] + ( ) ⋅ ( )[ ]

+ln ln ln

lnGovernment Trade Sectors

Trβ

β6

7 aade MarketAccess Coincidenceit i i it( ) ⋅ ( ) ⋅ ( )[ ] +ln ln ε

(4)

whereInequalityit represents the level of within-country regional inequality in country i in

year t, measured using the Gini index of regional GDP per capita.GDPcapit denotes real GDP per capita in PPP in constant US$ (2000) for country i in

year t.Developmenti is a dummy variable that takes the value of 1 if country i is a developing

or transitional economy and 0 otherwise. The categories were assigned on the basis ofhistorical World Bank classifications. Each country was assigned to its most frequentclassification over the period covered in the data set. This variable was, in turn, subdividedinto three components:

Vol. 88 No. 2 2012

121

TR

AD

EA

ND

REG

ION

AL

INEQ

UA

LITY

1. High-incomei is another dummy variable that takes the value of 1 if country i has beenmost frequently classified as a high-income country and 0 otherwise.

2. Middle-incomei is a dummy variable that takes the value 1 of if country i has been mostfrequently classified as an upper-middle-income country and 0 otherwise.

3. Low-incomei is a dummy variable that takes the value of 1 if country i has been mostfrequently classified as a low- (or lower-middle) income country and 0 otherwise.

Tradeit represents the total imports and exports in current U.S. dollars divided by GDPin PPP current U.S. dollars for country i in year t.

Sectorsi is a variable aimed at capturing the degree of interregional sectoral differencesthat exist across countries, proxied by the standard deviation of the share of agriculture inregional GDP, averaged across the periods under study for country i.7

Governmenti denotes the size of the government in country i, proxied by the share ofnonmilitary governmental expenditures in the total GDP averaged across the periodsunder study. It is assumed that interregional transfer programs and social expenditures arelinearly related to the level of governmental expenditure in the total GDP and that, in mostcountries, there will be a certain progressiveness built into the territorial distribution ofinvestment.

MarketAccessi denotes the degree of interregional differences in access to foreignmarkets across countries. Taking into account existing data constraints in the countriescovered in the sample, I used two alternative measures of access to markets. The firstvariable (Surfacei) is each country’s surface area in square kilometers. However, surfacearea is a rather crude measure of market access, especially in view of the huge diversityin population density among countries. Hence, I constructed an alternative compositemeasure of internal market access polarization (MAPolarizationi). The MAPolarizationi

index has the following form:

MAPolarization Surface Index Infrastructure Indexi i i= + , (5)

where

Surface Index surface maximum surface in samplei i= ( )100 (6)

and

InfrastructureIndexroad railway

population

density

ii i

i

=+

⎡

⎣

⎢⎢⎢⎢⎢

⎤

⎦

⎥⎥⎥⎥

100 (7)

where surfacei represents the surface area in square kilometers of country i, roadi

represents the kilometers of paved road in country i, railwayi represents the kilometers ofrailway lines in country i, populationi represents the total population of country i, anddensity represents the mean population density of the countries in the sample.

7 Ideally, a finer sectoral disaggregation to capture in a more precise way the variation of modern sectorendowments among domestic regions should have been used, perhaps including the subsectors of theservice sector. But given the diversity of countries that are included in the panel, the share of agriculture inregional GDP over time was the best comparable indicator available.

ECONOMIC GEOGRAPHY

122

The logic behind the use of the MAPolarizationi variable is that both the level ofabsolute internal distances (Surface Indexi) and the population density-adjusted infra-structural endowments (Infrastructure Indexi) determine the degree of interregionalvariation in access to foreign markets. The first concerns the internal transportationdistances, and the second proxies for the average transportation costs of a country. I chosea one-to-one weighting under the assumption that the proxy for the quality and quantityof the transportation infrastructure will reflect not only average transportation costs butthe number and availability of international transhipment and customs facilities along acountry’s coasts and borders.

Coincidencei reflects the degree of coincidence between relative regional positions inaccess to markets and regional levels of income per capita across countries. I used twoalternative measures of coincidence between both factors. The first (Coincidence25i) isthe ratio of the average GDP per capita levels of the regions in the top 25 percent in termsof access to foreign markets over average regional GDP per capita. The second(Coincidence50i) calculates the same ratio on the basis of the regions in the top 50 percentin terms of relative access to foreign markets. To ensure consistency with the dependentmeasure of regional inequality, which treats each region as one observation, I alsocomputed the coincidence ratios disregarding regional population sizes.

The question, of course, is how to determine relative positions in access to markets. Inthe absence of adequate and comparable data sets of regional transportation costs to anequivalent selection of international trade points—ports, airports, and the main terrestrialborder crossings—in each country, I first identified the trade entry points that wereaccountable for at least 70 percent of the country’s total trade, as well as the top quarteror half of the regions in terms of the locations of borders or coasts in closest proximity tothe main trade routes. When two regions were close in the accessibility of their borders ofcoasts to the main trade routes, I chose the region with the higher number of internationalports or border crossings. The main trade-entry regions for each country are presented inTable A1 in the appendix.

This geography-based construction of the coincidence measures also addresses a potentialendogeneity issue.Assuming that perfect data about each region’s access to foreign marketsin terms of the weighted market potential of actual transportation costs are available, it ishighly likely that high degrees of regional inequality are associated with higher degrees ofcoincidence because regional prosperity tends to be a driver of access to markets whenmeasured in terms of human-built infrastructure. Relying on physical proximity and thelocation of borders or coasts instead is not subject to this potential endogeneity issue. As inthe case of the previous structural conditioning variables, I averaged the coincidencemeasures across periods for each country. The data sources for each of the variables arepresented in Table A2 in the appendix. Finally, e represents the error term.

To assess the original questions of whether trade and the remaining variables includedin equation (4) affect regional inequalities and whether this relationship changes overtime, I ran both static OLS with country and time fixed effects, as well as dynamic panels.The aim of the static analysis was to discover the association (or lack of it) between tradeand the evolution of regional disparities. In the case of the dynamic regressions, I appliedgeneral method of moments (GMM) estimation, following Arellano and Bond (1991),Arellano and Bover (1995), and Blundell and Bond (1998), to distinguish between short-and long-term effects. The problem with running OLS on panels that include the laggeddependent variable is that it will be correlated with the error term even after theunobserved country heterogeneity is omitted. To adjust for this bias, Arellano and Bond(1991) proposed a first-difference GMM estimator using lagged values of the dependentand predetermined variables and differences of the strictly exogenous ones as instru-

Vol. 88 No. 2 2012

123

TR

AD

EA

ND

REG

ION

AL

INEQ

UA

LITY

ments. The system GMM estimator involves variables in levels instrumented with lags oftheir own first differences to exploit additional moment conditions (Arellano and Bover1995; Blundell and Bond 1998).

Another potential problem is that the interaction specifications are estimated withoutincluding the full set of main effects; that is, the models are nonsaturated. Estimatingnonsaturated models has the risk of leading to coefficients that may just be attributable tosome omitted levels or lower-order interactions. However, there are two reasons forrunning models that do not include an individual parameter for all the possible valuestaken by the explanatory variables. The first reason is related to the structure of the data.Some of the variables included in the analysis, that is, the dependent variable (inequality)and the main independent variable of interest (trade) and GDP per capita, vary over time.Other variables, however, because of theoretical reasons or as a result of the problems oftrying to gather comparable data for 28 countries over a long period, are time invariant.These variables include the development and income dummy variables, polarization,coincidence, and the government variables. Time variation is thus achieved through theinteraction of time-variant and time-invariant variables and the estimated models arealmost as close to saturated as they can be, given the data that are available. The otherreason is that “saturated models generate a lot of interaction terms, many of which maybe uninteresting or imprecise” (Angrist and Pischke 2009, 38), making it sensible to omitsome of the interaction terms. In the case of the current analysis, the only way to make themodel more saturated is to introduce the interaction of GDP per capita with all the othervariables. Doing so does not alter the coefficients significantly but makes the model lessparsimonious and dilutes its link to the theoretical discussion.

The Impact of Trade on Regional InequalityStatic Analysis

This section presents the results of running the different specifications of equation (4).Table 2 summarizes the results for the static OLS with country and time fixed effects. Thestandard errors are clustered by country. Given that all unobserved invariant country andtime heterogeneity was eliminated from the model, the coefficients can be interpreted asthe partial effects that annual variations of independent variables around the country meanhave had on annual variations of spatial inequality around the country mean.

When I considered trade as a free-standing variable (Table 2, Regression 1), I found noassociation between changes in trade patterns and the evolution of regional disparities.This finding coincides with the results of other studies that have looked at the simpleassociation between trade and regional inequality (e.g., Rodríguez-Pose and Gill 2006).This lack of association changes when, as specified in the diverse hypotheses, trade isconsidered in interaction with a series of country-specific factors. Here, the results of thestatic panel highlight the presence of a weak, but positive and highly significant, effect ofthe dimension of real trade on spatial inequality when pooled across all 28 countries.AfterI controlled for the internal growth effect and its different slopes across developed anddeveloping countries, I found that a 1 percent increase in real trade openness is on averageassociated with a 0.17 percent increase in the Gini index of regional inequality (Table 2,Regression 2). The results also indicate that this effect is significantly stronger indeveloping countries than in developed ones (Table 2, Regression 3), and that, amongemerging countries, the higher the aggregate level of wealth, the weaker the effect of tradeon interregional disparities (Table 2, Regression 3).

Regressions 4 to 9 go beyond the simple binary relationship between trade andinequality and introduce the conditioning structural variables that were identified in the

ECONOMIC GEOGRAPHY

124

Tabl

e2

Stat

icPa

nelw

ithCo

untr

yan

dTi

me

Fixe

dEf

fect

s

Dep

ende

ntva

riab

le:G

inic

oeffi

cien

t1

23

45

67

89

10

GD

Pcap

.168

0.2

433*

*.2

766*

*.2

657*

*.3

049*

**.1

799

.179

1.2

251*

*.2

418*

*.3

607*

**G

DPc

ap*D

evel

opm

ent

-.12

23**

-.17

21**

-.15

23**

-.19

92**

*-.

0540

**-.

0404

**-.

1025

**-.

0998

**-.

2363

***

Trad

e.0

725

.172

8**

.104

2**

-.48

40.8

620*

*1.

7055

**1.

770*

*1.

1955

**1.

2968

**2.

1162

***

Trad

e*D

evel

opm

ent

.123

7**

.116

0Tr

ade*

Gov

ernm

ent

-.33

37**

-.09

32Tr

ade*

Sect

ors

.208

1***

.235

8***

Trad

e*C

oinc

iden

ce50

*MA

Pola

riza

tion

.788

8***

Trad

e*C

oinc

iden

ce25

*MA

Pola

riza

tion

.888

9***

Trad

e*C

oinc

iden

ce50

*Sur

face

.154

4***

Trad

e*C

oinc

iden

ce25

*Sur

face

.135

1***

.127

2***

Con

stan

t-1

.510

-3.6

31-3

.811

-3.7

29-3

.968

-3.2

97-3

.317

-3.6

99-3

.841

-4.5

92R2

(with

in)

0.00

30.

227

0.23

270.

2527

0.25

770.

2503

0.26

220.

2775

0.28

850.

359

Obs

erva

tions

435

435

435

435

435

435

435

435

435

435

Fte

stfo

rco

untr

ydu

mm

ies

Prob

>F

Prob

>F

Prob

>F

Prob

>F

Prob

>F

Prob

>F

Prob

>F

Prob

>F

Prob

>F

Prob

>F

=0.

640

=0.

000

=0.

000

=0.

000

=0.

000

=0.

000

=0.

000

=0.

000

=0.

000

=0.

000

*,**

,***

corr

espo

ndto

10,5

,and

1pe

rcen

tsi

gnifi

canc

ele

vels

,res

pect

ivel

y,co

mpu

ted

with

hete

rosk

edas

ticity

adju

sted

stan

dard

erro

rs.

Vol. 88 No. 2 2012

125

TR

AD

EA

ND

REG

ION

AL

INEQ

UA

LITY

previous section. All the coefficients have the expected sign—rises in trade are associatedwith lower regional inequalities in countries with large governments and with higherinequalities in cases of strong interregional sectoral differences, when there are importantdifferences in access to markets and when these differences coincide with geographicdisparities in income per capita, and all are significant at the 1 percent level. Poorercountries with lower governmental expenditures, higher variations in regional sectoralstructures, and a spatial structure dominated by high internal transaction costs coupledwith a higher degree of coincidence between prosperous regions and access to foreignmarkets are thus bound to experience greater rises in regional inequality when opening toforeign trade. This effect rises with the overall level of trade and declines as the wealth ofthe country increases.

It is interesting that when all the conditioning interactions are added together (Table 2,Regression 10), the binary Development dummy variable interaction effect becomesinsignificant. The same is the case for the Government expenditure interaction. Thesechanges could simply be the result of collinearity between the Development dummyvariable and the Government variable. But this is not the case. The Government variableremains significant once the Sectors interaction is dropped, meaning that the problem ofcollinearity arises between the Government and Sectors interactions, but not betweenDevelopment and Government. This finding suggests that the proposed structural vari-ables account to a great extent for the apparent differences in the association betweentrade and within-country spatial inequalities across developed and developing countries.

To test whether the weak binary Development dummy variable interaction of the impactof trade also holds at a less aggregate categorical level, I divided the panel into high-, middle-and low-income countries, according to the World Bank’s classification, using the high-income group as the reference category. Table 3 presents the results of this type of analysis.

Adding greater nuance to the developed/developing country division leads to anincrease in the significance of development dummy variable interactions (Table 3,Regression 2), in comparison to those reported in Regression 3 in Table 2. The resultssuggest that variations in levels of trade openness have a significantly higher association

Table 3

Trade Effect in Developed and Developing Countries (Time and Country Fixed Effects Included)

Dependent variable: Gini coefficient 1 2 3 4

GDPcap .2766** .4628*** .1427 -.0954GDPcap*Development -.1721** -.3489*** -.2438*** .3507*Trade .1042** -.0587 .9534** 2.8924***Trade*Development .1237** -3.2878***Trade*GDPcap -.0814** -.2888***Trade*GDPcap*Development .3508***Trade*Middle Income .3963***Trade*Low Income .3523***Constant -3.811 -5.027 -2.262 -1.951R2 (within) 0.2327 0.2968 0.2347 0.2681Observations 435 435 435 435F test for country dummies Prob > F Prob > F Prob > F Prob > F

= 0.000 = 0.000 = 0.000 = 0.000

*, **, *** correspond to 10, 5, and 1 percent significance levels, respectively, computed with heteroskedasticity adjustedstandard errors.

ECONOMIC GEOGRAPHY

126

with average variations in spatial inequality in middle- and low-income countries than inhigh-income ones in the short term. In contrast, there tends to be less difference betweenthe impact of changes in trade on spatial inequality between low- and middle-incomecountries (Table 3, Regression 2).

When instead of testing for different slopes of the trade effect on spatial inequalityacross groups, I examined the effect of trade changes as countries grow—by interactingtrade openness with the countries’ real GDP per capita (Table 3, Regression 3)—theresulting coefficient points, as was shown in Table 2, toward a weakening of the positiveassociation between increases in trade and within-country spatial inequalities as countriesbecome wealthier. Overall, Table 3 confirms the differential impact of trade on regionalinequalities in developed and developing countries. Trade has had a greater impact onspatial inequality in developing countries, and this effect tends to diminish with economicdevelopment at a slower pace than in developed countries.

An important final point concerns the striking difference between the coefficient resultsfor the internal GDP per capita determinant of spatial inequality, in the tradition ofWilliamson (1965), and the external trade-induced factor. The negative and frequentlysignificant coefficient of the interaction term is particularly surprising. It suggests thatreal trade openness appears to have had a more polarizing effect in developing countriesthan has had wealth. The important question in this context, of course, is this: what are theunderlying structural factors behind the observed differences in the trade effect?

These results highlight that, as was inferred by Rodríguez-Pose and Gill (2006), theroom for growth in spatial inequalities is much greater in the developing than in thedeveloped world because (1) developing countries tend to be characterized by structuralfeatures that potentiate the polarizing effect of trade openness; (2) they already have muchhigher existing levels of spatial inequality; and (3) their level of trade openness is, onaverage, still only lower than that among developed countries.

Dynamic AnalysisTable 4 presents the short- and long-run results of the dynamic panel regressions. The

results were computed using the xtabond2 command in STATA (Roodman 2006). Theycorrespond to the first difference Arellano-Bond GMM estimation because the usuallypreferred Arellano-Bover system GMM was repeatedly rejected by the Sargan test ofoveridentification, indicating that its additional assumptions on the data-generatingprocess did not hold.

Since this section of the article is fundamentally interested in the medium- to long-termimplications of trade for regional disparities, the explanation focuses on the long-termparameters. The first result was expected. When I switched to dynamic panels with thelagged level of inequality included on the right-hand side, I found that most of thedifferences in current levels of within-country spatial inequality are explained by previouslevels of within-country inequality. The high degree of inertia inferred from the coeffi-cient of the lagged level of regional inequality is normal for this type of analysis andrenders the effect of trade openness on regional inequality less relevant than in the staticanalysis (Table 4, Regression 1). The same is the case for the binary Development dummyvariable interaction term in Regression 2 (Table 4). The speed of adjustment parameter isabout 0.3, suggesting the presence of a strong difference between short- and long-termeffects of all the included independent factors (see Table 4). Regional inequality tends tobe path dependent and does not change radically from one year to the next.

Regressions 3 to 9 (see Table 4) introduce the structural conditions in the dynamicmodel. Here, some of the partial effects of the static fixed-effect model are confirmed in

Vol. 88 No. 2 2012

127

TR

AD

EA

ND

REG

ION

AL

INEQ

UA

LITY

Tabl

e4

Dyn

amic

Pane

lwith

Firs

t-Diff

eren

ceAr

ella

no-B

ond

GM

M(T

ime

Fixe

dEf

fect

sIn

clud

ed)

Dep

ende

ntva

riab

le:G

inic

oeffi

cien

t

Sho

rt-t

erm

para

met

ers

12

34

56

78

910

Lagg

edin

equa

lity

.713

2***

.718

8***

.691

7***

.691

7***

.712

6***

.715

4***

.711

2***

.709

0***

.709

9***

.691

7***

GD

Pcap

-.01

02.0

002

.006

.021

6-.

0165

-.01

06-.

0168

-.01

37.0

040

.003

7G

DPc

ap*D

evel

opm

ent

.030

3.0

243

.014

1-.

0038

.028

9.0

261

.033

8.0

311

.016

6.0

133

Trad

e.0

158

.020

0-.

2429

**.2

631*

*-.

1196

-.08

03.0

862

.118

7.0

232

.117

2Tr

ade*

Dev

elop

men

t-.

0116

-.04

86Tr

ade*

Gov

ernm

ent

-.13

84**

-.06

36Tr

ade*

Sect

ors

.072

6**

.059

6Tr

ade*

Coi

ncid

ence

50*M

APo

lari

zatio

n-.

0110

Trad

e*C

oinc

iden

ce25

*MA

Pola

riza

tion

.069

4Tr

ade*

Coi

ncid

ence

50*S

urfa

ce.0

009

Trad

e*C

oinc

iden

ce25

*Sur

face

.017

4Tr

ade*

Coi

ncid

ence

25*D

evel

opm

ent

.721

0**

.589

8*

Obs

erva

tions

379

379

379

379

379

379

379

379

379

379

Sarg

ante

stPr

ob>

chi2

Prob

>ch

i2Pr

ob>

chi2

Prob

>ch

i2Pr

ob>

chi2

Prob

>ch

i2Pr

ob>

chi2

Prob

>ch

i2Pr

ob>

chi2

Prob

>ch

i2=

0.93

55=

0.94

07=

0.88

94=

0.91

47=

0.94

93=

0.94

84=

0.95

41=

0.94

61=

0.95

30=

0.93

95

Seco

nd-o

rder

auto

corr

elat

ion

Pr>

z=

Pr>

z=

Pr>

z=

Pr>

z=

Pr>

z=

Pr>

z=

Pr>

z=

Pr>

z=

Pr>

z=

Pr>

z=

0.50

320.

4920

0.52

620.

5343

0.50

110.

4886

0.53

330.

5252

0.48

770.

4958

Lo

ng-t

erm

para

met

ers

12

34

56

78

910

GD

Pcap

-.03

56*

.000

7.0

195

.070

1-.

0574

*-.

0372

*-.

0582

-.04

71*

.013

8.0

120

GD

Pcap

*Dev

elop

men

t.1

056

.086

4.0

457*

*-.

0123

.100

6.0

917

.117

0.1

069*

.057

2.0

431

Trad

e.0

551

.071

1-.

7879

.853

4*-.

4161

-.28

22.2

985

.407

9.0

800*

*.3

801*

Trad

e*D

evel

opm

ent

-.04

13-.

1576

Trad

e*G

over

nmen

t-.

4489

-.20

63Tr

ade*

Sect

ors

.235

5**

.193

3Tr

ade*

Coi

ncid

ence

50*M

APo

lari

zatio

n-.

0383

Trad

e*C

oinc

iden

ce25

*MA

Pola

riza

tion

.243

9**

Trad

e*C

oinc

iden

ce50

*Sur

face

.003

1Tr

ade*

Coi

ncid

ence

25*S

urfa

ce.0

598

Trad

e*C

oinc

iden

ce25

*Dev

elop

men

t2.

4853

***

1.91

31**

*

*,**

,***

corr

espo

ndto

10,5

,and

1pe

rcen

tsi

gnifi

canc

ele

vels

,res

pect

ivel

y,co

mpu

ted

with

hete

rosk

edas

ticity

adju

sted

stan

dard

erro

rs.

Trad

e,se

ctor

s,go

vern

men

t,an

dsp

atia

lvar

iabl

esen

tere

dth

ein

stru

men

tm

atri

xas

stri

ctly

exog

enou

s.

ECONOMIC GEOGRAPHY

128

the medium- and long term. This is the case of sectoral differences. The presence of stronginterregional sectoral differences not only affects regional inequality in the medium-term,but contributes to making trade significant and a key factor in increasing regionalinequalities (Table 4, Regression 4). Moreover, countries with a strong level of marketpolarization and where market access is concentrated in the richest regions are also likelyto witness an increase of regional inequality in the medium- and long run (Table 4,Regression 6). This effect is particularly strong in developing economies, where increasesin trade will have a significantly stronger effect in those countries where trade-entry pointscoincide with the richest regions (Table 4, Regressions 9 and 10). Since this is the case inthe majority of countries in the developing world, trade is likely to have a long-lastingeffect on the persistence and increase in regional disparities in the emerging world.8 Otherspatial variables, in contrast, keep the same signs of the static analysis but displayinsignificant coefficients.

Robustness testsTo check whether these results are robust to differences in specifications, I replaced the

Gini index of regional inequality is with alternative measures of inequality. Thus, I ran thespecifications in Tables 2 to 4, replacing the Gini coefficient of within-country regionalinequality as the dependent variable with two alternative measures: the Theil index andthe population-weighted coefficient of variation. The results are robust to the change inspecification and can be provided on request.

Another check of robustness, given the limited number of observations in a panelincluding 28 countries relative to the time of the analysis, is to use a bias-correctedleast-squares dummy variable (LSDV) estimator (Kiviet 1995; Bun and Kiviet 2003),instead of an instrumental variable GMM estimation. This approach also allows one toaccommodate for unbalanced panels (Bruno 2005). By resorting to this method, one cancheck whether the results from the Arellano-Bond GMM estimation in Table 3 proverobust to an alternative estimator. The results are displayed in Table 5. Standard errorswere derived by setting the number of bootstrap repetitions to 200.

Table 5 reveals that the size and sign of the coefficients of interest remain similar tothose presented in Table 4. The speed-of-adjustment parameter slightly decreases tobelow 0.25, as indicated by the higher coefficient of the lagged level of regional inequa-lity. However, none of the previously found significance levels is confirmed. This resultmakes it difficult to draw any firm conclusions on the dynamic adjustment processbetween openness and regional inequality from the data.

ConclusionThe aim of this article has been to improve our understanding of the relationship

between changes in trade patterns linked to global market integration, on the one hand,

8 In all the emerging world countries included in the sample, trade entry points coincide with some of therichest regions in the country. This is particularly the case in China, Indonesia, Brazil, and Thailand, whereBeijing or Shanghai, Djakarta, São Paulo or Rio de Janeiro, and Bangkok, respectively, represent the keyeconomic agglomerations and the main hubs for imports and exports. I observed a similar structure in manydeveloping countries outside the sample. In many sub-Saharan African countries, such as Senegal, Guinea,Ivory Coast, Ghana, Togo, Benin, Nigeria, Angola, Mozambique, and Tanzania, the richest cities are locatedin ports that channel trade flows to the rest of the country. Only in a limited number of cases (India’s largestport, Kandla, in Gujarat, or Mexico’s ports of Manzanillo and Lázaro Cárdenas, in Colima and Michoacán,respectively) trade points do not coincide with the richest and most dynamic regions. But it is extremely rareto see a major port, airport, or trade route in one of the poorest regions in a country. Gioa Tauro inimpoverished Calabria, in southern Italy, is one of these exceptions.

Vol. 88 No. 2 2012

129

TR

AD

EA

ND

REG

ION

AL

INEQ

UA

LITY

Tabl

e5

Dyn

amic

Pane

lwith

Bias

Corr

ecte

dLS

DV

(Are

llano

-Bon

das

the

Initi

atin

gEs

timat

or;T

ime

Fixe

dEf

fect

sIn

clud

ed)

Dep

ende

ntva

riab

le:G

inic

oeffi

cien

t1

23

45

67

89

10

Lagg

edin

equa

lity

.769

5***

.773

2***

.762

5***

.756

2***

.771

7***

.771

2***

.765

8***

.763

7***

.768

8***

.760

1***

GD

Pcap

-.00

43-.

0114

-.00

57.0

019

-.00

17-.

0033

-.00

07.0

003

-.01

02-.

0076

GD

Pcap

*Dev

evel

opm

ent

.044

7.0

553

.054

4.0

366

.039

4.0

413

.042

3.0

414

.054

0.0

507

Trad

e.0

073

.017

2-.

0514

.172

5-.

1524

-.09

48.0

582

.101

7.0

198

.341

5Tr

ade*

Dev

elop

men

t-.

0231

-.05

09Tr

ade*

Gov

ernm

ent

-.03

06.0

416

Trad

e*Se

ctor

s.0

4884

.069

7Tr

ade*

Coi

ncid

ence

50*M

APo

lari

zatio

n-.

0675

Trad

e*C

oinc

iden

ce25

*MA

Pola

riza

tion

.104

7Tr

ade*

Coi

ncid

ence

50*S

urfa

ce-.

0081

Trad

e*C

oinc

iden

ce25

*Sur

face

.014

4Tr

ade*

Coi

ncid

ence

25*D

evel

opm

ent

.569

9.5

615

Obs

erva

tions

379

379

379

379

379

379

379

379

379

379

*,**

,***

corr

espo

ndto

10,5

,and

1pe

rcen

tsi

gnifi

canc

ele

vels

,res

pect

ivel

y,co

mpu

ted

with

200

boot

stra

pre

petit

ions

.Tr

ade,

sect

ors,

gove

rnm

ent,

and

spat

ialv

aria

bles

ente

red

the

inst

rum

ent

mat

rix

asst

rict

lyex

ogen

ous.

ECONOMIC GEOGRAPHY

130

and within-country spatial inequalities, on the other, from both a theoretical and anempirical perspective. This is particularly relevant, given the recent emphasis of the WDR(World Bank 2009) that increases in trade may lead to greater growth at the expense ofincreases in territorial disparities, but that this is a temporary condition because greaterdevelopment would eventually weaken within-country spatial inequality.

The article is based on a model that combines regional spatial characteristics with aseries of country features. The spatial characteristics include the degree of interregionalvariation in access to foreign markets and whether these differences in foreign marketscoincide with differences in income. The conditioning country features include the degreeof interregional sectoral variation, the level of governmental expenditures, the degree oflabor mobility, and institutions. The lack of data on the two latter categories allows testingfor the former two conditions only. To examine whether development weakens spatialinequalities, my analysis also controlled for the internal growth effect and its interactionwith the country’s developmental stage. The influence of these variables on the evolutionof within-country regional inequality was then tested using both static and dynamicpanels.

The results show that trade—when considered in combination with country-specificfactors—matters for the evolution of regional inequalities. The weak association betweenboth factors in the static panel analysis, which improved significantly when the condi-tioning variables were included in the analysis, implying that while changes in trade makea difference for the evolution of spatial disparities, the impact of changes in trade is morepolarizing in countries with higher interregional sectoral differences, lower shares ofgovernmental expenditures, and a combination of higher internal transaction costs withhigher degrees of coincidence between wealthier regions and access to foreign markets.However, the spatial country variables ceased to be significant once the lagged levels ofinequality in dynamic panels were controlled, meaning that no firm conclusions can bedrawn regarding the dynamic time frame of spatial adjustments and the distinctionbetween short- and long-term effects of trade openness.

The key result is that changes in trade patterns seem to affect the evolution of regionalinequality in developing countries to a much greater extent than in developed ones. Thespatially polarizing effect of trade also decreases at a significantly slower pace indeveloping countries than in developed ones. And trade, in contrast to what was suggestedby Williamson (1965), seems to have a greater sway on the evolution of regionalinequality than does economic growth. That is, economic growth—whether directlyprovoked by changes in trade or not—cannot offset the potentially negative effects forterritorial equality of increases in trade in the developing world.

Policymakers in the developing world, as well as international organizations, may thusneed to tread carefully when thinking about the potential implications of greater marketopenness for their countries. While greater openness to trade is likely to yield rewards interms of growth and the absolute welfare of local citizens, it may also bring the unwel-come consequence of greater territorial polarization. Although, as the WDR (World Bank2009) pointed out, this situation may not necessarily be bad in the short term, enhancingterritorial inequality in countries with already high levels of spatial polarization andwhere territorial differences may pile on top of preexisting social, cultural, ethnic, and/orreligious grievances, can exacerbate tensions that could ultimately undermine the veryeconomic benefits that trade is supposed to bring about. Hence, it is convenient to bringthe territorial implications of trade into the trade policy equation. Doing so may implytrade policies aimed at promoting a more harmonious territorial growth not just thosefocused on generating greater agglomeration, since these latter policies can have unin-tended effects that may ultimately limit their influence on development.

Vol. 88 No. 2 2012

131

TR

AD

EA

ND

REG

ION

AL

INEQ

UA

LITY

Ades, A. F., and Glaeser, E. L. 1995. Trade and circuses: Explaining urban giants. QuarterlyJournal of Economics 110:195–227.

Adkisson, R., and Zimmerman, L. 2004. Retail trade on the U.S.-Mexico border duringthe NAFTA implementation era. Growth and Change 35:77–89.

Alderson, A., and Nielsen, F. 2002. Globalisation and the great U-turn: Income inequalitytrends in 16 OECD countries. American Journal of Sociology 107: 1244–99.

Angrist, J. D., and Pischke, J.-S. 2009. Mostly harmless econometrics: An empiricist’s compan-ion. Princeton, N.J.: Princeton University Press.

Arellano, M., and Bond, S. 1991. Some tests of specification for panel data: Monte Carloevidence and an application to employment equations. Review of Economic Studies58:277–97.

Arellano, M., and Bover, O. 1995. Another look at the instrumental-variable estimationof error-components models. Journal of Econometrics 68:29–52.

Barrios, S., and Strobl, E. 2009. The dynamics of regional inequalities.Regional Science andUrban Economics 39:575–91.

Blundell, R., and Bond, S. 1998. Initial conditions and moment restrictions in dynamicpanel data models. Journal of Econometrics 87:115–43.

Breau, S., and Rigby D. L. 2010. International trade and wage inequality in Canada. Journalof Economic Geography 10:55–86.

Brown, T. M. 1952. Habit, persistence and lags in consumer behaviour. Econometrica20:355–71.

Brülhart, M. 2011. The spatial effects of trade openness: A survey. Review of WorldEconomics 147:59–83.

Brülhart, M.; Crozet, M.; and Koenig, P. 2004. Enlargement and the EU periphery:The impact of changing market potential. World Economy 27:853–75.

Brülhart, M., and Sbergami, F. 2008. Agglomeration and growth: Empirical evidence.Journal of Urban Economics, 65:48–63.

Bruno, G. S. F. 2005. Approximating the bias of the LSDV estimator for dynamicunbalanced panel data models. Economics Letters 87:361–66.

Bun, M. J. G., and Kiviet, J. F. 2003. On the diminishing returns of higher order terms inasymptotic expansions of bias. Economics Letters 79:145–52.

Chiquiar, D. 2008. Globalization, regional wage differentials and the Stolper-Samuelsontheorem: Evidence from Mexico. Journal of International Economics 74:70–93.

Crozet, M., and Koenig-Soubeyran, P. 2004. EU enlargement and the internal geographyof countries. Journal of Comparative Economics 32:265–78.

Faber, B. 2007. Towards the spatial patterns of sectoral adjustments to trade liberali-sation: The case of NAFTA in Mexico. Growth and Change 38:567–94.

Farole, T.; Rodríguez-Pose, A.; and Storper, M. 2011. Cohesion policy in the EuropeanUnion:Growth,geography, institutions. Journal of Common Market Studies 49:1089–111.

Ford, T.C.; Logan,B.; and Logan, J. 2009.NAFTA or nada? Trade’s impact on U.S. borderretailers. Growth and Change 40:260–86.

Jian, T.; Sachs, J. D.; and Warner, A. M. 1996. Trends in regional inequality in China. ChinaEconomic Review 7:1–21.

Jordaan, J. A. 2008a. Intra- and inter-industry externalities from foreign direct investmentin the Mexican manufacturing sector: New evidence from Mexican regions. WorldDevelopment 36:2838–54.

ECONOMIC GEOGRAPHY

Ref

eren

ces

132

——. 2008b. Regional foreign participation and externalities: New empirical evidence fromMexican regions. Environment and Planning A 40:2948–69.

Kanbur, R., and Venables A. J., eds. 2005. Spatial inequality and development, Oxford, U.K.: OxfordUniversity Press.

Kanbur, R., and Zhang, X. 2005. Fifty years of regional inequality in China: A journey throughcentral planning, reform and openness. Review of Development Economics 9:87–106.

Kiviet, J. F. 1995. On bias, inconsistency, and efficiency of various estimators in dynamic panel datamodels. Journal of Econometrics 68:53–78.

Krugman, P., and Livas Elizondo, R. 1996. Trade policy and the Third World metropolis. Journal ofDevelopment Economics 49:137–50.

Milanovic, B. 2005. Worlds apart: Measuring international and global inequality. Princeton, N.J.:Princeton University Press.

Monfort, P., and Nicolini, R. 2000. Regional convergence and international integration. Journal ofUrban Economics 48:286–306.

Niebuhr, A. 2006. Market access and regional disparities: New economic geography in Europe.Annals of Regional Science 40:313–34.

Nitsch, V. 2006. Trade openness and urban concentration: New evidence. Journal of EconomicIntegration 21:340–62.

Paluzie, E. 2001. Trade policies and regional inequalities. Papers in Regional Science 80:67–85.Petrakos G.;Rodríguez-Pose, A.; and Rovolis, A. 2005.Growth, integration,and regional disparities

in the European Union. Environment and Planning A 37: 1837–55.Puga, D. 1999. The rise and fall of regional inequalities. European Economic Review 43:303–34.Ravallion, M. 2001. Growth, inequality and poverty: Looking beyond averages. World Development

29:1803–15.Rodríguez-Pose, A., and Gill, N. 2006. How does trade affect regional disparities? World Develop-

ment 34:1201–22.Rodríguez-Pose, A., and Sánchez-Reaza, J. 2005. Economic polarization through trade: Trade

liberalization and regional growth in Mexico. In Spatial inequality and development,eds. R. Kanburand A. J. Venables 237–59. Oxford, U.K.: Oxford University Press.

Roodman, D. 2006. How to do xtabond2: An introduction to “difference” and “system” GMM inStata. Working paper 103, Center for Global Development,Washington, D.C. Available online:www.cgdev.org/files/11619_file_HowtoDoxtabond8_with_foreword.pdf

Sachs, J. D., and Warner, A. M. 1995. Economic reform and the process of global integration.Brookings Papers on Economic Activity, No. 1:1–118.

Sánchez-Reaza, J., and Rodríguez-Pose, A. 2002. The impact of trade liberalization on regionaldisparities in Mexico. Growth and Change 33:72–90.

Williamson, J.G. 1965.Regional inequality and the process of national development: A descriptionof the patterns. Economic Development and Cultural Change 13(4):3–45.

——. 2005. Winners and losers over two centuries of globalization. Wider perspectives on globaldevelopment, ed. UNU-WIDER, 136–74. New York: Palgrave Macmillan.

Wood, A. 1994.North–south trade,employment,and inequality:Changing fortunes in a skill-driven world.Oxford, U.K.: Clarendon Press.

World Bank. 2009. World development report 2009: Reshaping economic geography. Washington,D.C.: World Bank

Yang, D. T. 2002. What has caused regional inequality in China? China Economic Review 13:331–34.Zhang, X., and Zhang, K. H. 2003. How does globalisation affect regional inequality within a

developing country? Evidence from China. Journal of Development Studies 39:47–67.

Vol. 88 No. 2 2012

133

TR

AD

EA

ND

REG

ION

AL

INEQ

UA

LITY

App

endi

xTa

ble

A1

Num

ber

ofRe

gion

san

dSt

ruct

ural

Cond

ition

s,by

Coun

try

Cou

ntry

Num

ber

ofre

gion

sD

evel

opm

ent

Hig

hin

com

eM

iddl

eIn

com

eLo

wIn

com

eG

over

nmen

tSe

ctor

sM

APo

lM

ain

Trad

eEn

try

Poin

tsC

oin2

5C

oin5

0

Aus

tral

ia8

stat

es0

10

00.

160.

0214

5.09

New

Sout

hW

ales

,V

icto

ria

11.

05

Aus

tria

9Lä

nder

01

00

0.18

0.02

83.7

2V

ienn

a1.

061.

07B

elgi

um11

prov

ince

s0

10

00.

20.

0187

.77

Brus

sels

,Ant

wer

pen,

Wes

tFl

ande

rs0.

951.

1

Bra

zil

26st

ates

10

10

0.17

0.07

182.

44Sã

oPa

ulo,

Rio

deJa

neir

o,Pa

raná

0.59

0.65

Bul

gari

a6

NU

TS2

regi

ons

10

01

0.14

0.06

98.8

3Yu

goza

pade

n,Se

vero

izto

chen

1.15

1.12

Can

ada

12pr

ovin

ces

01

00

0.2

0.03

174.

58O

ntar

io,B

ritis

hC

olum

bia

10.

91C

hina

31pr

ovin

ces

10

01

0.13

0.07

182.

86Sh

angh

ai,T

ianj

in,B

eijin

g,G

uand

ong,

Zhe

jiang

,Sh

ando

ng

1.73

1.32

Cze

chR

ep8

NU

TS2

regi

ons

10

10

0.2

0.03

95.4

2Pr

ague

0.88

1.15

Fin

land

5N

UT

S2re

gion

s0

10

00.

210.

0296

.04

Sout

hFi

nlan

d1.

181.

13Fr

ance

22R

égio

ns0

10

00.

20.

0257

.36

Ile-d

e-Fr

ance

,Pr

oven

ce-A

lpes

-Côt

ed’

Azu

r,H

aute

Nor

man

die

0.97

0.99

Gre

ece

13N

UT

S2re

gion

s0

10

00.

110.

0690

.3A

ttic

a,C

entr

alM

aced

onia

,Wes

tG

reec

e

0.93

1

Hun

gary

7re

gion

s1

01

00.

090.

0493

.96

Cen

tral

Hun

gary

1.1

0.76

Indi

a28

stat

esan

d4

unio

nte

rrito

ries

10

01

0.09

0.11

118.

73D

elhi

,Mah

aras

htra

,G

ujar

at,A

ndhr

aPr

ades

h,Ta

mil

Nad

u

1.17

0.97

Indo

nesi

a30

prov

ince

s1

00

10.

060.

1111

6.06

Jaka

rta,

East

Java

1.18

1.29

(Con

tinue

don

next

page

)

ECONOMIC GEOGRAPHY

134

Tabl

eA

1

Num

ber

ofRe

gion

san

dSt

ruct

ural

Cond

ition

s,by

Coun

try

(Con

tinue

d)

Cou

ntry

Num

ber

ofre

gion

sD

evel

opm

ent

Hig

hin

com

eM

iddl

eIn

com

eLo

wIn

com

eG

over

nmen

tSe

ctor

sM

APo

lM

ain

Trad

eEn

try

Poin

tsC

oin2

5C

oin5

0

Ital

y21

regi

ons

01

00

0.17

0.02

87.6

9La

zio,

Lom

bard

y,Li

guri

a,C

alab

ria

1.25

1.22

Japa

n47

pref

ectu

res

01

00

0.15

0.02

74.5

3To

kyo,

Chi

ba,K

anag

awa,

Osa

ka,A

ichi

,Hyo

go,

Fuku

oka

1.02

1.03

Mex

ico

31st

ates

and

Dis

tuito

Fede

ral

10

10

0.1

0.05

117.

73M

exic

oD

.F,Ve

racr

uz,

Col

ima,

Tam

aulip

as1.

411.

04

Net

herl

ands

12pr

ovin

ces

01

00

0.21

0.02

91.4

7N

orth

Hol

land

,Sou

thH

olla

nd1.

071

Pola

nd16

voiv

odes

hips

10

10

0.18

0.04

88.1

Maz

owie

ckie

,Pom

orsk

ie1.

061.

01Po

rtug

al5

cont

inen

talN

UT

S20

10

00.

160.

0796

.02

Lisb

oa,N

orte

1.41

1.13

Ro

man

ia8

NU

TS2

regi

ons

10

01

0.08

0.07

97.6

Buca

rest

,Sou

thEa

st0.

970.

95S

lova

kR

ep4

obla

sti(

NU

TS2

)1

01

00.

190.

0296

.4Br

atis

lava

1.85

1.33

So

uth

Afr

ica

9pr

ovin

ces

10

10

0.17

0.02

104.

42G

aute

ng,

Wes

tern

Cap

e,K

waz

ulu-

Nat

al1.

031

Spa

in17

auto

nom

ous

regi

ons

01

00

0.16

0.03

84.4

8M

adri

d,C

atal

onia

,Va

lenc

ia,A

ndal

usia

1.02

1.07

Sw

eden

8N

UT

S2re

gion

s0

10

00.

250.

0283

.1St

ockh

olm

,W

est

Swed

en0.

970.

95

Tha

iland

75pr

ovin

ces

10

01

0.08

0.13

104.

8Ba

ngko

k,C

honb

uri

1.92

1.46

Uni

ted

Kin

gdo

m37

NU

TS2

regi

ons

01

00

0.17

0.03

83.3

4Lo

ndon

,Ess

ex,

Ham

pshi

re1.

11.

05

Uni

ted

Sta

tes

50st

ates

and

Dis

tric

tof

Col

umbi

a0

10

00.

120.

0296

.43

New

York

,Cal

iforn

ia,

Texa

s,Lo

uisi

ana,

Geo

rgia

,Flo

rida

1.05

0.98

Vol. 88 No. 2 2012

135

TR

AD

EA

ND

REG

ION

AL

INEQ

UA

LITY

Table A2

Variables and Sources of Data

Variable Source of Data

Inequality National statistical offices and Eurostat Regio databaseGDPcap Word Development IndicatorsDevelopment Historical Series of World Bank classificationsHigh-income Historical Series of World Bank classificationsMiddle-income Historical Series of World Bank classificationsLow-income Historical Series of World Bank classificationsTrade UN Comtrade and World Development IndicatorsGovernment World Development IndicatorsCoincidence UN Comtrade,World Port Database, own calculations

ECONOMIC GEOGRAPHY

136