You are on page 1of 13

Economic Integration and Income Convergence: Not Such a Strong Link?

Author(s): Branko Milanovic Source: The Review of Economics and Statistics, Vol. 88, No. 4 (Nov., 2006), pp. 659-670 Published by: The MIT Press Stable URL: http://www.jstor.org/stable/40043026 . Accessed: 18/06/2013 12:52
Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at . http://www.jstor.org/page/info/about/policies/terms.jsp

.
JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of content in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new forms of scholarship. For more information about JSTOR, please contact support@jstor.org.

The MIT Press is collaborating with JSTOR to digitize, preserve and extend access to The Review of Economics and Statistics.

http://www.jstor.org

This content downloaded from 41.141.98.200 on Tue, 18 Jun 2013 12:52:49 PM All use subject to JSTOR Terms and Conditions

ECONOMIC INTEGRATION AND INCOME CONVERGENCE: NOT SUCH A STRONG LINK?


Branko Milanovic*
- We would expect thatthe process of globalizationbetween Abstract of the world economy 1870 and 1914 and subsequentdisintegration to incomeconvergence theinterwar and during periodwouldhave led first betweentheparticipating countries. But in fact thento incomedivergence evidence for income convergence we findstronger duringthe interwar the first the average level of Similarly, globalization. periodthanduring in the worldcannotbe shownto have either helpedor import protection is The evidence for trade-induced convergence. convergence hampered therefore weak.

states (Barro and Sala-i-Martin, 1992), among European Cannon & Duck, 2000, p. 418), (for regions example, (Goerlich& Mas, 2001), and so amongSpanishprovinces forth.1 In all such cases, greatereconomic integration or regionsor states)was shownto amongunits(countries - as we wouldexpect in incomeconvergence have resulted from economictheory. Morerecently, somewhat attention was paid to the greater I. EconomicIntegration and Income Convergence historical (Maddison, 1995, processof incomedivergence factdid notdetract from the in favor of economic 2001; Pritchett, of the main arguments 1997), butthat in mainstream belief a link causal between in to the fact it economic is addition that raises that, strong integration it helps the poorerones integration incomes of all the participants, and income convergence. This is because the more. This is the view that has informed Great named Kenneth proportionally Pomeranz)was Divergence(so by on income convergence much of the recentliterature due to the discretetechnological of the Inbreakthroughs of theconditional or of theunconditional It whether dustrial while the fact of income variety. Revolution, divergence is a view thathas a long and distinguished of theworldoverthelast20 years(see pedigreein amongthecountries and is supported economic theory, by a fair amountof Milanovic,2005, chapter 4; Kanburand Lustig,1999,table In evidence. increased trade raises 2) was explainedaway by theclaim thattheslow-growing theory, contemporary But access of a (or countries. real incomesof all participating were preciselythose thatdid not declining)countries to embodied in goods or technology poor country superior The onlyshadowwas cast by thosewho did not integrate.2 capital, or simply throughintellectualexchange allows regardtheGreatDivergenceas something thatoccurred in the that is further gains poorcountry, - in one greater productivity forwhatever institutional or geographical reasons frontier. the production Free capital away from possibility the world and then (the North) part (slowly) spreadto the in flowswill also help thepoor country more,by bringing and industrialization in the rest,but held thatthe growth new technology and by allowingit to tap into the larger Northwere linkedwiththedeclineand deindustrialization too should Finally, migration savingspool of a richcountry. in the South. Under the latterhypothesis, it is clearly in incomes,as people from to convergence contribute poor is that the cause of the South's decline and integration to the rich.Thus, greater countries migrate integration therefore of the divergence of incomes.3 That view is exin closer sharingof information and technology reflected in Krugman and was recently summarized (1991), pressed and moretrade, greater capitalflows, (knowledge spillover), Baldwin and Martin At a time beforethe (1999, 7): by p. labor migration should help reducethe gap betweenthe Industrial are Revolution, write, they "regions initially poor and therich. so thequestionwhichregiontakesoffis a matter This view is behind a score of empiricalpapers on identical, Whichever regionedges ahead initially, The earliestpapers on the conver- of happenstance. income convergence. call it finds itself in a virtuous circle.Higher incomes North, countries over the periodof a gence amongindustrialized to lead a local market in the North and this in turn larger in 1870 were by Baumol (1986) and century beginning attracts more investment to the North. Of course, relatively literature conBaumol and Wolff (1988). The convergence the investment rate leads to a market-size OECD countries higher withpaperson convergence growing tinued among ... As theNorth this experiences 1992) amongEuropeanCommu- gap and thecyclerestarts. (Barroand Sala-i-Martin, industrial Southern disrevolution, individual U.S. members stylized industry rapidly 1993), (Ben-David, among nity from Northern In appearsin thefaceofcompetition exports. Received for publicationDecember 12, 2003. Revision accepted for a self-generating process,theNorthspecializesin industry September13, 2005. publication and theSouthin primary goods." * Endowment WorldBank and Research
Carnegie Group, Development Peace. forInternational assistance.I am to PremSangraulaforexcellentresearch I am grateful to Michele Alacevich, Joe Ferrie,Mansoob Murshed,Martin grateful Thomas Pogge, BernardWasow, two anonyRavallion,Maurice Schiff, Economic at Northwestern and participants mous referees, University The views expressedin the forveryhelpfulcomments. seminar History to theWorldBank, paperare author'sown, and shouldnotbe attributed or its affiliated organizations. CarnegieEndowment, 1See also the reviewof in Barroand Sala-i-Martin (1995). findings 2 For themostrecent manifestation of such a view see theWorldBank's on globalization (2002). report 3 Even if the South's decline (see Bairoch, 1997, vol. 2, pp. 549, 576, 648; also Bairoch, 1989, p. 238) may notbe viewed as the cause of the Northern success. On a moreradical note,Frank(1998) arguesthatthe South's declinehelpedthe North'sadvance.

November and Statistics, TheReview 2006, 88(4): 659-670 ofEconomics Institute of Technology and Fellowsof Harvard 2006 by thePresident Collegeand theMassachusetts

This content downloaded from 41.141.98.200 on Tue, 18 Jun 2013 12:52:49 PM All use subject to JSTOR Terms and Conditions

660

THE REVIEW OF ECONOMICS AND STATISTICS

So, we see that it is at least possible for economic which widened (notice the causality) the gaps between to lead to a declinein incomesin a partof the nations.Foreman-Peck (1998, p. xxiii),in theintroduction integration foundaof textson historical worldand/or to divergence. The introduction of increasing to an excellentcompendium the interwar summarizes returns to scale in thecontext of neoclassicalor endogenous tion of globalization, period of and models(fora reviewsee Easterly & Levine,2001) deglobalization: "[b]etweenthe wars,trademigration growth becamefarmoreserious, movement makes this a more realisticpossibility. A similarpointis currency impediments Williamson in train and divergence." made by Rodriguez and Rodrik (2000), who, based on setting deglobalisation also "I will inference also writes, 278) (1996, suggest numerous evidence and reruns of by p. empirical equationsorigbecause and 1950 that between 1914 convergence stopped estimated inally by various authors, argue thateconomic and "an and implosioninto autarchy," and convergence are orthogonal, and find that cr of deglobalization integration converwhich intervened, stopped anticonvergence regime counconvergence amongthefuture EuropeanCommunity if we look at between and 1950" 1914 But, 281). (p. gence triescontinued theinterwar during period. the case? the was this data, really is not veryseriouslycontemHowever,thispossibility The finding of incomeconverplatedby manyeconomists. (western between II. What Happened to Income Convergence genceamongtheclub ofrichcountries Europeand - to use Maddison's terminology) its offshoots the 1919 1939? and during earlierperiodof globalization1870-1913 providesempiriof the cal support forthemainstream view.4 The well-documented Thattheperiod1919-1939 is one of disintegration well known and world or is economy, deglobalization, post-WorldWar II convergencein incomes among the economic documented. affected all Disintegration OECD countries [Barro and Sala-i-Martin (1992, p. 244) amply and its sharein GDP flows:real volumeof tradestagnated, more and Serrano Li and, Maudos, Pastor, (2000), recently, trade barriers decreased from the proliferprewarvalues; and Papell (1999), de La Fuente(1998), and Tsangarides flows driedto a trickle and labor international ated; capital a further corroboration of thehypothesis. (2001)] presents 2000; Lewis, 1949; Bairoch, 1993; League of theseresults and theoretical we (see Crafts, Then,following predictions, - theperiodof retreat Nations,1927, 1936, 1939, 1945).6 would expecttheperiod 1919-1939 of the world Such a violentprocess of disintegration - to be characterized fromglobalization inby increasing in growth, to a slowdown is to lead expected come gaps betweenthecountries. And indeed,Lindert and economy - to affect - whatis important forour purposes and poorer Williamson(2001, p. 13) write:"Real wages and living Did then incomes reallydicountries disproportionately. standards[my emphasis]convergedamong the currently years?We calculatethreemeaduringthe interwar industrialized countries between 1850 and WorldWar I," verge - the Gini, sures of income (=GDP per capita) inequality and then, fortheinterwar was no period the period,". . . there across variation the coefficient of and Theil, twenty when divergence was more 'big time.' betweencountries major westerncounties (denoted WENAO, for western We do notyetknowhow muchof thisshouldbe attributed America,and Oceania)7 as well as across a Europe,North to thegreat twoworldwars,anti-global policies morerestricted depression, of western Europe sample of 17 countries and otherforces"(p. 19).5 Lindertand Williamsonneatly and its offshoots.8 inThese are measuresof intercountry summarize their resultsin a table wherethe period 1914- come We same. the each with counting country inequality from as theperiodofretreat 1950 is described globalization, calculate annual values for the Gini measure,the Theil of variation and thecoefficient measure, (or sigmaconver4 See Williamson (1998, figure 1), Lindertand Williamson (2001), gence,as itis called in theconvergence acrossthe literature) O'Rourke and Williamson(1999). 1870-1950 for the entire countries two of 5 In a different period groups that his finding Williamson (1991, p. 34) mentions paper, 1 and 2). The dataon GDP percapita(expressed in richcountries between 1914 and 1945 is contra- (see figures of wage divergence dicted by the findings of Baumol, Blackman, and Wolff(1989) and in 1990 dollars) are takenfromMaddison Geary-Khamis of GDP percapitaor Abramowitz (1986, table 1,p. 391) that convergence unabatedexceptfortheperiodof World (2003). GDP per worker-hour continued theGini,theTheil,orthe is never resolved,nor is it paid much War II. But the contradiction usingeither Figure1 showsthat almostall theliterature coefficient to.This is strange forthree reasons.First, attention we findthatWENAO incomesdid of variation,
is couched in terms of GDP per capita. Why should we on convergence forthe interwar use a different criterion of convergence period?Second, less so than are probably theGDP per capita data,howeverproblematic, of average wages of five the real wage data, which are a potpourri in Sydney(Australia), industries average (Belgium),realwage of laborers (Argentina), average daily wage of laborersin wage in manufacturing a country, thedefinitions trades (Canada), and so on. Even within building of wages used by Williamson changequite a lot over time(see appendix 1 in Williamson,1991). Third,GDP per capita is surelya much better so wherewage indicator of livingstandards thanwages, and particularly as was thecase in most earners accountfor50% or less of thelaborforce, of thesecountries the periodunderstudy. during 6 For moredetailsee section2 oftheWeb version available of thispaper, 1.pdf . at http://econ. worldbank.org/files/22948_wps294 7 Australia, Finland,France,GerAustria, Belgium,Canada, Denmark, New Zealand, Norway, many,Greece, Ireland,Italy,the Netherlands, Portugal,Spain, Sweden, Switzerland,the United Kingdom,and the UnitedStates. 8 Thatis, thesame WENAO countries minus footnote from theprevious is almost and Spain. "Western Greece,Portugal, Europeand itsoffshoots" the same group of countriesas "the Atlanticeconomy" as dubbed by Williamson.

This content downloaded from 41.141.98.200 on Tue, 18 Jun 2013 12:52:49 PM All use subject to JSTOR Terms and Conditions

ECONOMIC INTEGRATION AND INCOME CONVERGENCE


- Gini, Theil, and Coefficientof Variation of GDP per Capita: WENAO Countries, 1870-1950 Figure 1.

661

Figure 2.- Gini, Theil, and Coefficientof Variation of GDP per Capita: Restricted Sample, 1870-1950

of economic activityin a not divergeduringthe interwar period.9On the contrary, wroughta massive disruption In 1919,theGini,theTheil,and the numberof continental there was convergence. Thus between European countries. coefficient of variationwere respectively 21.0, 7.0, and 1939 and 1945, Germany'sGDP per capita decreased by and 0.38; in 1939, theywererespectively 17.8, 5.5, and 0.32. It 23%, France'sby almost50%, Greece's by two-thirds, is the Second World War that "created" divergence.It so on. On the other hand, the United States, Canada, and Australia 78%, 50%, and surgedahead (by respectively in thus differences GDP 9 The annual values of the threecoefficients 18%), widening per capita and withtheirstandard errors thedivergence. can be obtainedfromthe author. The intercountry Giniwentup from creating

This content downloaded from 41.141.98.200 on Tue, 18 Jun 2013 12:52:49 PM All use subject to JSTOR Terms and Conditions

662

THE REVIEW OF ECONOMICS AND STATISTICS

Figure 3.- Gini Coefficients for Restricted Sample of WENAO Countries' GDPs per Capita and Their 95% Intervals in 1870, 1913, 1919, and 1939

Note:Ginicoefficient shown as fraction of 20%). Number of countries is 17 in each year. (0.2 instead

theoutbreak of theWorldWarII to 32 same period,going from5.8 to 2.8. This is all the more just under18 before at itsend; thecoefficient of variation almostdoubled,from interesting in that itis with to thisgroupthat Lindert respect 0.32 to 0.59. Of course,incomedivergence is notuniqueto and Williamsonclaim thatthe disintegration of the world the Second WorldWar.The same divergence in incomes, economy led to income divergence.As can be easily albeitof a smaller theFirst World War checked,theirmistakestemsfroma comparison of 1913 size, occurred during 1 and 2). We shallreturn to theroleof thewars and 1945. Indeed,incomedifferences in 1945 weregreater (see figures below. thanin 1913,butthat was entirely due to thehugedifference and contrary to expectations, theevidencefor in fortunes Moreover, the Second World War. Ascribingthe during incomeconvergence theglobalization in than in 1945 1913 to thedevelopments during period 1870- higher inequality Therewas an important 1913 is weaker. ofincome during theinterwar would be incorrect. shrinking years differences between 1890 and 1895 duringthe economic In figure 3, we focus on the Gini values in fourcrucial that thethree of inequality butafter measures crisis, display years: 1870, 1913, 1919, and 1939. We see no statistically no trend untiltheoutbreak of WorldWar I. difference betweenthe first threevalues. The significant Even more dramaticand tellingis the example of the situation, is different fortheyear 1939,wherethe however, morerestricted Figure2 showsthat Giniis less and is statistically sampleof 17 countries. (at the5% level) significantly the Gini coefficient of GDPs per capita of the restricted lowerthantheearlierGinis.10 The results are thesame for between1919 and theTheil index(not shownhere).Appendix A provides sampledeclinedalmostuninterruptedly the 1939. The interwar veryfastincomecon- results periodwitnessed of univariate of the three indexes testing inequality theera of (Gini, Theil, and coefficient muchfaster thananything seen during vergence, of variation)over the 1870The negativeslope of the Gini, Theil, and 1939 period.Our objectiveis to check whether globalization. theseries thanduring variation linesis muchsteeper the are stationary coefficient-ofand/orwhetherthere is a level or trend in The Gini coefficient heydayof the first globalization. 1919 was 18; on theeve of the Second WorldWar,it was 10 Note thatthe 1939 Gini is slightly less than(or equal to, in thecase more thanhalved over the of 1914) thelowerboundof the95% interval fortheprevious three Ginis. only 12. The Theil coefficient

This content downloaded from 41.141.98.200 on Tue, 18 Jun 2013 12:52:49 PM All use subject to JSTOR Terms and Conditions

ECONOMIC INTEGRATIONAND INCOME CONVERGENCE

663

fortheentire change between the globalization and interwar periods. shownin table 1. Considertheresults sample The results imply either (a) that the data-generating of WENAO countries(top panel). The coefficients on be shown to have been different cannot between GDP are in but the process lagged per capita always negative, thatthe globalization pooled and IV regressions the two periods,and in particular theyare not statistically signifias far as conver- cant, and we findno evidence for convergence and the interwar in either period did not differ are concerned, or (b) thatwhenevidence for period. In the fixed-effects gence trends regression, though,we find it is is for evidence for in theinterwar convergence present unambiguously stronger statistically strong convergence the interwar and no effect for the R2is quite The period. period, pre-1914 period. We can also testforconvergence in usingstandard regres- low in all cases exceptforthe interwar years the fixedsionstests. As is conventionally thegrowth effects formulation. The results fortherestricted done,we regress sampleare rateof GDP per capita (change in incomelogs) on initial stronger of table the coefficient on income (bottom 1): panel and t is negative and statistically level of income (yit for the interwar ,-j), wherei indicatescountry, significant time: in two out of three formulations. The coefficient on period income for the globalizationperiod is never statistically - In _, = + Pi Inyu- i In yit (Jo beyu significant. Figure 4 displays the simple relationship tween rates and initial income: it convergrowth highlights + P2 ln Zit+ Ui+ vt+ eih gence in the interwar period relativeto its near-absence and In Zit = In (nit+ g + 8), wherenitis the population before. We can testthesensitivity of ourresults the by dropping rate,g the rate of labor-augmenting technological growth a between the World War I and prioriperiodization prerate(all derivedfromthe and 8 thedepreciation progress, the extentof periods.Instead,to reflect directly textbook Solow modelof economicgrowth), and uitvt, and interwar economic or we use an estimate integration protectionism, time- and both countryand timeeit are the countryof the world tariff rate calculated average by Coatsworth terms.11 All GDP percapitavaluesaretaken error dependent and Williamson The results where we introduce the (2002). and the growthrate (the dependent at 5-year intervals, world tariff rate and its interaction with rateover each averageweighted variable)is the average annualizedgrowth income (in orderto check whether a more protectionist with 1870.12 5-yearperiodstarting overall environment has different effects on poorer and suffer from a number of Equationssuchas (1) potentially richer in are shown table 2.15 We note thatno countries) econometric The most obvious are the omitted problems. variable is for either the WENAO statistically significant variablebias whererelevant information is country-specific or the restricted The coefficient on income is not sample. where the dependentand not included,and endogeneity different from in 0 of the formulations any variablesarejointlydetermined.13 We thusrun significantly independent positive as negative. The average of equation (1): pooled regression, in- and is as frequently threeformulations rateand its interaction withincomeare likewisenot strumental variable (IV) regression, and fixed-effects re- tariff significant. Consequently,the level of proformulation combinesindiscriminatelystatistically gression.The first tectionism the 1870-1939 during periodcannotbe shown in thesecondformutimeand cross-sectional observations; to have been either or associatedwith negatively positively lation,we addresspotential endogeneity by instrumenting variablesby their theright-side lagged values; in thethird convergence. we adjustforunobserved efformulation, country-specific fectsby estimating a fixed-effects model.14 The results are A. Crisesand Wars income 11 produced The sum of g and 8 is assumedto be 0.05 (5 %). 12 WENAO countries. But also the 1890 because of thewar,and thesampleends in divergence The year 1915 is omitted among than a 1938 (so the last period gives a three-year and 1929 crisesled to a remarkable average rate rather reduction in inequality average). five-year betweenthecountries (table 3). 13 a formulation such as equation (1) suffers Besides these problems, Bliss The weakness of the from Galton'sfallacy(see Quah, 1993; 1999). is emphasizedin Wodon and Yitzhaki formulation. testsforp convergence Note also thatin theStatasoftware, fixed-effects empirical regressions can be observedwhenone moves do have an intercept because insteadof the constraint of intercept term, (2002). They show thatP-convergence in time,as well as whentheunderlying distribution equal to 0, thesoftware forward or backward constraint imposesa different (equally arbitrary) of the (the sum of country dummiesis 0). See http://www.stata.com/support/ diverges,converges,or stays the same. Even the interpretation .html . obtainedresultsis questioned(Quah, 1996). We use this formulation f aqs/stat/xtreg2 15 used in the(immense) and themostcommonly The tariff rateis measured as import because it is thesimplest dutiesdividedby import values. Our view is thatdirecttestsof unconditional Coatsworth and Williamson literature on convergence. estimate annualvalues forboththeweighted of inequality statistics like the and the unweighted rate.The weightedaverage is (as impliedin thecalculation average world tariff convergence to testsbased on regression obtainedby usingcountries' are superior Gini coefficient) & (see Coatsworth analysis. exportsharesas weights 14 one (because relevant Williamson, a verystripped-down The modelis by necessity 2002, p. 2). We use five-year averages(the mean of annual ratesand educationlevels are world average tariff variablessuch as investment rates forthe 1870-1875 period,and so on) in the country-level forthisis to controlfor regressions. We also experiment withtheunweighted unavailable).One of the ways to tryto correct averageworldtariff as is done herein thefixed-effects rate.The results unobservable are basicallyunchanged. factors, country-specific They are given in appendixB. We have notedthat thetwoworldwars

This content downloaded from 41.141.98.200 on Tue, 18 Jun 2013 12:52:49 PM All use subject to JSTOR Terms and Conditions

664

THE REVIEW OF ECONOMICS AND STATISTICS


Table 1.-Convergence in the Two Periods 1870-1910and 1920-1938 PooledRegression (lj 1870-1913 (2) 1918-1939 IV Regression (3) 1870-1913 (4) 1919-1939 FixedEffects (5) 1870-1913 (6) 1919-1939

A. WENAO Countries In yit-] ln(*,,+8+\) Constant No. ofobs. R2 F-value F-test of excluded instruments -0.0004 (0.879) 0.008 (0.329) 0.041 (0.279) 159 0.01 14 -0.009 (0.213) 0.025* (0.016) 0.163* (0.022) 91 0.031 -0.001 (0.74) 0.009 (0.30) 0.05 (0.23) 156 0.01 13 >3000 -0.003 (0.68) 0.02* (0.02) 0.11 (0.13) 89 0.03 895 -0.004 (0.467) -0.002 (0.868) 0.039 (0.522) 152 0.004 0.29 -0.079** (0.006) 0.047* (0.031) 0.806** (0.003) 91 0.13 5 -

B. The Restricted Sample \nyit-i ln(/i,,+5+\) Constant No. ofobs. R2 F-value F-test of excluded instruments -0.003 (0.363) 0.008 (0.312) 0.061 (0.114) 142 0.01 1 -0.025* (0.013) 0.032** (0) 0.319** (0) 80 0.10 11 -0.003 (0.382) 0.009 (0.292) 0.063 (0.114) 140 0.01 1 2634 (0) -0.014 (0.153) 0.029** (0.001) 0.219* (0.013) 79 0.09 6 421 (0) -0.004 (0.526) -0.002 (0.876) 0.036 (0.555) 142 0.01 0.3 -0.063* (0.021) 0.043* (0.032) 0.664** (0.008) 80 0.12 4 -

variable: annualized GDP percapitagrowth overthe5-year interval. Dependent at 1% levelandless Coefficients between as that Note:Growth rateof population is calculated in exactly thesamefashion, andoverthesameperiod, ofGDP percapita, significant parentheses. p-valuesshown R2 to correct for standard error 7 and 8 above.All regressions runwith robust at 5% level,byone asterisk. Forthecomposition of thesamples, see footnotes indicated (White) heteroskedasticity. bytwoasterisks; forfixed-effects is R2within. regression

to"the in 1945than in 1914canbe assigned 5 showstheratio between end-and initial-year great variously Figure GDP percapita the crises and war as function two world wars, (for policies and anti-global periods) depression, due to the ofinitial GDP percapita. richer countries lost otherforces."In effect, On average, the gap is entirely more than more than the others the and War. Second World crises, during gained policies,the Despite "anti-global" theothers the The of income differwars. and 1939. between 1919 to shrink income continued shrinking during gap in the severe decline encesin 1890-1895was driven the standard by Moreover, regressions convergence though intheworld), whose Australia therichest of the income the acceptance (then country per on balance lean toward 16 the convergence for the interwar capitaincome dropped by morethan20%. During period,they hypothesis GreatDepression, theUnitedStatesplayedtherole of showthatconvergence absentfortheprewas entirely theopposite 1914period.17 Australia. The twoworldwarshad exactly leveloftariff theaverage that We also find effect on the crises.In both,richcountries did much protection for in theworldseemsnotto have mattered In theFirst World better than theless rich. War, widening convergence. ofthe income differences werecreated seemto that on twokeypoints bythefast growth castdoubt Theseresults UnitedStates,theUnitedKingdom, New Zealand,and havebecomewidely to themany thanks largely accepted in 1914),and in the contributions Australia four countries (therichest that the Williamson. First, pre-1914 byJeffrey World WarII, by that of theUnited Switzerland,globalization States, bothabsolute witnessed convergence period in (realwage,realrent) three countries and New Zealand (again,therichest factor andrelative convergence price 1939). countries rich Williamson, (see ratios) among (wage/rental GDPs rich countries' of well as as 1998, 68), convergence p. III. ConcludingComments and Williamson's In O'Rourke (1999, p. 167) percapita. an offers Atlantic of the "the words, economy history We can nowbreak thestatedownintoitscomponents on between correlation globalization positive ment and Williamson (2001) thatthe much unambiguous by Lindert This book on the other. handand convergence the one income differences rich countries that existed greater among is correlation that the and [O'Rourke Williamson's] argues
16 thatbothGini and Theil decline even if Australiais Note, however, excluded.The declinesare smaller, though:1. 1 insteadof 2.7 pointsfor theGini,0.6 insteadof 1.4 pointsfortheTheil. 17 For the resultsusing two alternative GDP-per-capitadata sources (Bairoch's and Prados de la Escosura's), see appendixC.

This content downloaded from 41.141.98.200 on Tue, 18 Jun 2013 12:52:49 PM All use subject to JSTOR Terms and Conditions

ECONOMIC INTEGRATION AND INCOME CONVERGENCE


Year Annualized Growth Rates and Initial GDP per Capita (in Logs): Figure 4.- Relationship between FiveLeft, before World War I (1870-1910); Right, Interwar Period (1920-1938)

665

one country and period. Note:Growth ratesshownin ratios (0.05 is 5 percent p.a.). Each dotrepresents

TariffRate Table 2.- Convergenceover the Period 1870-1939UsingWorld Average Weighted Pooled Regression WENAO \nyin-i + \) ln(fil7+8 Tariff rate(weighted) Tariff rateX In y/,-, Constant No. of observations R2 F-value F-test of excludedinstruments -0.0106 (0.290) 0.014 (0.075) -0.0035 (0.630) 0.0005 (0.556) 0.129 (0.155) 261 0.037 2.6 Restricted Sample -0.0004 (0.971) 0.0119 (0.126) 0.0052 (0.555) -0.0005 (0.638) 0.0367 (0.745) 238 0.083 4.80 IV Regression WENAO 0.0038 (0.746) 0.0126 (0.121) 0.005 (0.530) -0.0005 (0.583) 0.008 (0.936) 257 0.03 1.6 569 Restricted Sample 0.0211 (0.243) 0.0104 (0.197) 0.0185 (0.123) -0.002 (0.144) -0.145 (0.347) 235 0.069 4.0 351 FixedEffects WENAO -0.015 (0.168) 0.0145 (0.180) -0.004 (0.523) 0.0006 (0.449) 0.166 (0.080) 261 0.0425 2.6 Restricted Sample 0.0015 (0.896) 0.014 (0.147) 0.006 (0.414) -0.0006 (0.500) 0.026 (0.797) 238 0.0790 4.7 -

overthe5-year interval. variable: annualized GDP percapitagrowth Dependent at 1% levelandless indicated as that ofGDP percapita, Coefficients is calculated inexactly thesamefashion, andoverthesameperiod, rate of population Note:Growth significant p valuesbetween parentheses. 7 and and Williamson of thesamples, see footnotes worldtariff ratecalculated from Coatsworth one asterisk. Tariff rate is two at 5% (2002). For thecomposition level, asterisks; 5-year average weighted by by is R2within. R2forfixed-effects error to correct forheteroskedasticity. runwithrobust 8 above.All regressions (White)standard regressions

Table 3.- The Two Crises and Two Wars: Changes in Intercountry Inequality Index Percentage Changein Inequality Gini Coefficient Theil Coefficient of Coefficient Variation

The Crises 1890-1895 1929-1935 -13 -10 The Wars 1914-1918 1939-1945 +20 +78 +40 +195 +20 +84 -21 -19 -13 -10

causal [my emphasis]." Second, that the deglobalization period was associated with increasinggaps between the nations.Regardingthe first point,the evidence presented the of convergence heregives a less sanguine during picture of we find almost no As a matter fact, period. globalization theevidenceallows us evidenceof it. On thesecondpoint, to reject the view thattherewas divergenceof incomes in the interwar period,and thus among the rich countries must have been the cause of it. Our thatdeglobalization conclusion is that the type of world trade preliminary to thepresenceor was, at leastup to 1950,unrelated regime absence of incomeconvergence amongrichcountries.

This content downloaded from 41.141.98.200 on Tue, 18 Jun 2013 12:52:49 PM All use subject to JSTOR Terms and Conditions

666

THE REVIEW OF ECONOMICS AND STATISTICS

Figure 5.- Ratio between End- and Initial-Year Incomes Plotted against Initial Income Level: (A) Crisis 1890-1895; (B) Crisis 1929-1935; (C) World War I, 1914-1919; (D) World War II, 1939-1945

Note:Lines showsimplelinear regressions.

andWelfare: What Baumol, William, Growth, "Productivity Convergence, theLong-Run Data Show," American Economic Review 76 (Deand Abramowitz, Moses,"Catching Behind," Ahead, Falling Up,Forging cember 1986),1072-1116. Journal (June 1986),385-406. ofEconomic History and EdwardWolff, ConverBaumol, William, Growth, "Productivity AnAppliforStationarity Amara, Jomana, UsingCovariates: "Testing andWelfare: American Economic Review 78 (Degence, Reply," cation to Purchasing Power Parity," of Houston University cember 1155-1159. 1988), Available at (2003). manuscript http://www.uh.edu/~jhamara/ S. A. B. Blackman, andEdward N. Wolff, Baumol, William, Productivity webverson_ll_18.pdf. and American MA: Leadership:The Long View(Cambridge, ofEconomic Economic Lawsand Paul,"TheParadoxes Bairoch, History: MIT Press, 1989). Economic Review 33 225-249. (1989), European History," BenDavid,Dan,"Equalizing Trade Liberalization andIncome Exchange: Economics and World and Paradoxes History: Myths (Chicago: Journal 108:3(1993). Quarterly ofEconomics Convergence," ofChicago Press,1993). University Galton s FallacyandEconomic OxConvergence, Victoires et Deboires, HistoireEconomiqueet Sociale du Bliss,Christopher, Economic ford Papers51 (1999),4-14. Monde duXVIe FolioHistoire Sieclea NosJours Gallimard, (Paris: Cannon,EdmundS., and Nigel W. Duck, "Galton's Fallacy and 1997). Economic Economic and Philippe "TwoWavesof Globalisation: Oxford Papers 52 (2000), Convergence," Richard, Martin, Baldwin, 415-419. Fundamental National BuSimilarities, Differences," Superficial John G. Williamson, "The Rootsof Latin reauofEconomic Research no.6904 ( 1999). Coatsworth, H., and Jeffrey working paper American Protectionism: before theGreat and XavierSala-i-Martin, Journal Barro,Robert, Looking Depression," of "Convergence," National Bureau of Economic Research Political 100:(1992),223-251. working Economy paperno. 8999 Economic Growth McGraw-Hill, (2002). (NewYork: 1995).

REFERENCES

This content downloaded from 41.141.98.200 on Tue, 18 Jun 2013 12:52:49 PM All use subject to JSTOR Terms and Conditions

ECONOMIC

INTEGRATION AND INCOME CONVERGENCE

667

"On Cross-Country Growth and Convergence: Crafts, Nicholas,"Globalizationand Growthin the Twentieth Charalambos, Century," Tsangarides, Evidence from African IMF working and OECD Countries," JournalofAfrican paperno. 2000/44(2000). Economies 10:4 (2001), 355-389. de La Fuente, Angel(1998), "Convergence Equationand IncomeDynam"The Evolutionof Global Labor Markets in theFirst ics: The Sources of OECD Convergence,1970-95," Centrefor Williamson, Jeffrey, and Second WorldSince 1830: Background Evidence and LiteraEconomic Policy Research discussion paper series no. 1794 Harvard Institute of EconomicResearchdiscussion ture," paperno. (1998). 1571 (1991). Easterly,William, and Ross Levine, "It's Not Factor Accumulation: and History," Journalof Economic "Globalization, Convergence StylizedFacts and GrowthModels," The WorldBank Economic 56:2 (1996), 277-306. History Review 15:2 177-220. Labor Marketsand Policy Backlash in the Past," "Globalization, HistoricalFoundationsto GlobalForeman-Peck, James,"Introduction: Journalof EconomicPerspectives12:4 (1998), 51-72. ization,"in JamesForeman-Peck (Ed.), HistoricalFoundationsto and Shlomo Yitzhaki,"Growthand Convergence: An MA: EdwardEl- Wodon, Quentin, Globalization(Cheltenham, U.K., Northampton, AlternativeEmpirical Framework,"World Bank mimeograph Collection,1998). gar Reference (2002). of California AndreGunder, Re-Orient Frank, Press, 1998). (University WorldBank, Globalization, Growth and Poverty: Buildingan Inclusive FranciscoJ.,and MatildeMas, "Inequalityin Spain, 1973-91: Goerlich, WorldEconomy," WorldBank policyresearch (2002). report Contribution to a Regional Database," Review of Income and Wealth 47:3 (2001). Back on theAgenda?" Kanbur, Ravi,and Nora Lustig,"WhyIs Inequality APPENDIX A at the Annual Bank Conference on DevelopmentEcopresented World DC available at nomics, Bank, Washington, (1999), http:// UnivariateAnalysis of the Three Inequality Indexes, 1870-1950 l_e.htm (accessed www.iadb.org/sds/publication/publication_67 10, 2005). February We can make a further check of theeffect of globalization or degloband Trade (London: MIT Press, 1991). Paul, Geography Krugman, alization on incomeconvergence statisby lookingat whether inequality Level Indices, Geneva: League of Nations League of Nations, Tariff ticsbehaveddifferently in thetwo periods(forexample,beingstationary (1927). in one case and notin theother). We do twotests:first, fortheunit testing EconomicSurvey1935/1 936 (Geneva: League of Nations, World root (so thatthe rejectionof the hypothesis implies stationarity) using 1936). modified GLS test, and thentesting forstationarity Dickey-Fuller directly World EconomicSurvey1938/1939(Geneva: League of Nations, It has been arguedthat theunivariate unitroottests usingtheKPSS test.18 1939). in small samples tend to be fairlyconservative, thatis, to lead to the and ForeignTrade (Geneva: League of Nations, Industrialization roothypothesis too often (Amara,2003). To balance acceptanceoftheunit Economic,Financialand Transit 1945). Department, test.In thelatter this,we also use a direct case, we testforlevel stationary EconomicSurvey, 1919-1939 (London: GeorgeAllen and stationarity Lewis, Arthur, The resultsare shown in (as opposed to trendstationarity). Unwin,1949). tableAl. of International Li, Qmg, and David Papell, Convergence The KPSS test implies Gini and Theil stationarity in both periods. Output:Time Series Evidencefor16 OECD Countries," International Reviewof Dickey-Fuller cannotrejectthehypothesis of a unitrootin tests, however, Economicsand Finance 8:3 (1999), 267-280. either thetwotestsdo notagreewitheach other, thekey period.Although Lindert, Peter,and Jeffrey Williamson,"Does GlobalizationMake the point is thatthe two periods cannotbe shown to have been different, testwe use. WorldMore Unequal?" National Bureau of Economic Research whatever Another series way to look at theissue is to check whether inequality working paperno. 8228 (2001). be trend-stationary the WorldEconomy,1820-1992 (Paris: might (in bothperiods)witha level breakoccasioned Maddison,Angus, Monitoring between by the First World War. We can also allow trendsto differ OECD, 1995). and deglobalization a further check globalization The WorldEconomy:A Millennial Perspective,(Paris: OECD periods,thusproviding on thespeed of convergence or divergence. We testforthepotential level CentreStudies,2002). Development and trend breaksusing Perron's(1989) approach,wherelevel dummies The World (Paris: OECD DevelopEconomy:HistoricalStatistics and trend variablesare introduced at some significant dates and thethus mentCentreStudies,2003). transformed series is testedforstationarity. We estimate Jose Manuel and Lorenzo "ConverMaudos, Joaquin, Pastor, Serrano, and gence in OECD Countries:Technical Change, Efficiency + 32:6 (2000). A/,= fl0+ \/,_i + axDwar + a2DtTl + a7tDtT2 (Al) Productivity," AppliedEconometrics Worlds and Global International Milanovic,Branko, Apart: Inequality, 1950-2000 (Princeton: Princeton Press,2005). University where/,is theinequality statistic of variation), (Gini,Theil,or coefficient G. Williamson,Globalization and Dwar is the O'Rourke, Kevin H., and Jeffrey dummyforthe war years (1914-1918), DtTx and DtT2 are The 19th Evolution of the AtlanticEconomy time trendvariables for the History: Century globalizationand deglobalizationperiods, MA: MIT Press, 1999). (Cambridge, and e, is an error term.Several alternative formulations of respectively, Perron,Pierre,"The Great Crash, the Oil Shock and the Unit Root equation(Al), including ones withan overalltime-trend variableand with Econometrica 57 (1989), 1361-1401. twice-or thrice-lagged weretried, buttheresults were Hypothesis," statistics, inequality Prados de la bscosura, Leandro, International Comparison or Real essentiallythe same. The resultsof the equations (Al) for the three Data Set," Explorationsin inequality coefficients Product,1820-1990: An Alternative (see table A2) show thatthe negativetimetrend in the interwar EconomicHistory37 (2000), 1-41. thanduringthe (pro-convergence) period was stronger Pritchett, Lant, "Divergence,Big Time," Journalof Economic Perspec- globalization era, and that it always remains statistically significant. There is also some indicationthatWorldWar I mighthave occasioned tives 11:3(1997), 3-17. Quah, Danny,"Galton's Fallacy and Tests of theConvergenceHypoth- a downward level shiftin the series, althoughthe variable is never In conclusion,the interwar statistically significant. periodshows stronesis," Scandinavian Journal of Economics 94:4 (1993), 427ger evidence of income convergencethanthe 1870-1914 globalization 443. Available at http://econ.lse.ac.uk/staff/dquah7p/dp-93galperiod. fall.pdf. "Empiricsfor Economic Growthand Convergence,"European EconomicReview40 (1996), 1353-1375. Rodriguez,Francisco,and Dani Rodrik,"Trade Policy and Economic A Skeptic'sGuide to theCross-National Evidence" (VerGrowth: 18 sion May 2000). Downloadable from For unit-root tests,I also used Perronand augmented http://ksghome.harvard.edu/ Dickey Fuller ~ .drodrik.academic.ksg/skepti 1299.pdf. tests.They all yield the same results.

This content downloaded from 41.141.98.200 on Tue, 18 Jun 2013 12:52:49 PM All use subject to JSTOR Terms and Conditions

668

THE REVIEW OF ECONOMICS

AND STATISTICS

Table Al. - TestingStationarityof Inequality Indexes,1870-1939 GLS Test Dickey-Fuller Measure Testvalue at Max. Lag Critical Value (5%) Conclusion Period1870 to 1914 Gini Theil -0.91 -1.19 -2.12 -2.12 Unitrootaccepted Unitrootaccepted Period1918 to 1939 Gini Theil -0.41 -0.67 -5.14 -5.14 Unitrootaccepted Unitrootaccepted 0.33 0.30 0.46 0.46 Stationarity accepted Stationarity accepted 0.38 0.35 0.46 0.46 Stationarity accepted Stationarity accepted Test value at Max. Lag KPSS TestforLevel Stationarity Critical Value (5%) Conclusion

Note:Maximum rule. lag selected usingSchwert's

Table A2.- Testingfor Level and Trend Breaks in Inequality Statistics Statistic coefficient Inequality (lagged) war(1914-1919) Dummy Timetrend 1870-1913) (globalization, Timetrend (interwar, 1919-1939) Number of observations Adjusted/?2 F-value DurbinWatson Gini Coefficient -0.369** (0) -0.005 (0.25) -0.0002* (0.04) -0.0007** (0.01) 69 0.14 4 1.91 Theil Coefficient -0.364** (0) -0.002 (0.50) -0.0008 (0.13) -0.0003* (0.025) 69 0.13 4 1.91 of Coefficient Variation -0.380** (0) -0.009 (0.30) -0.0004* (0.04) -0.002** (0.01) 69 0.13 4 1.95

Note:p valuesbetween at 1% leveland less indicated Coefficients at 5% level,by one asterisk. parentheses. significant by twoasterisks;

APPENDIX B
TariffRate Table Bl. - Convergenceover the Period 1870-1939UsingWorld Average Unweighted PooledRegression WENAO lny,,-, ln(/i,,+8+\) Tariff rate(unweighted) Tariff rateX In yitt-X Constant No. of observations R2 F-value F-test of excluded instruments -0.009 (0.269) 0.013 (0.108) -0.003 (0.570) 0.0004 (0.504) 0.123 (0.132) 261 0.0285 2.06 (0.087) Restricted Sample -0.0049 (0.699) 0.011 (0.176) 0.001 (0.889) -0.00003 (0.970) 0.075 (0.506) 238 0.0553 3.47 (0.008) IV Regression WENAO 0.007 (0.538) 0.012 (0.169) 0.005 (0.381) -0.0006 (0.425) -0.0184 (0.853) 257 0.0188 1.45 (0.217) 518 Restricted Sample 0.022 (0.243) 0.009 (0.279) 0.015 (0.147) -0.0017 (0.166) -0.156 (0.342) 235 0.0349 2.74 (0.029) 309 FixedEffects WENAO -0.0143 (0.213) 0.014 (0.188) -0.004 (0.481) 0.0005 (0.425) 0.161 (0.104) 261 0.0316 1.93 (0.105) Restricted Sample -0.0019 (0.880) 0.014 (0.148) 0.0016 (0.808) -0.0001 (0.894) 0.0593 (0.595) 238 0.0510 2.91 (0.022) -

variable: annualized GDP percapitagrowth overthe5-year interval. Dependent at 1% levelandless indicated Coefficients Note:Growth rate ofpopulation is calculated inexactly as that ofGDP percapita, between thesamefashion, andoverthesameperiod, significant p-values parentheses. 7 and see footnotes of thesamples, at 5 % level,byone asterisk. andWilliamson Tariff rateis 5-year world tariff rate calculated from Coatsworth (2002). Forthecomposition bytwoasterisks; average unweighted 8 above.All regressions runwith R2forfixed is R2within. robust error to correct forheteroskedasticity. effects (White)standard regressions

This content downloaded from 41.141.98.200 on Tue, 18 Jun 2013 12:52:49 PM All use subject to JSTOR Terms and Conditions

ECONOMIC INTEGRATION AND INCOME CONVERGENCE


Figure Cl. - Gini Coefficients,1800-1938 (Calculated Using Bairoch and Prados de la Escosura's Data)

669

income between de la Esco1929and 1939.UsingPrados convergence between 1860 and 1913, and then sura'sdata,we findconvergence theinterwar Convergence usingBairochand Pradosde la Escosura'sData divergence during years.21 As a glanceat figures Cl andC2 reveals, theoriginal income diverthe we have In addition toMaddison's which are most data, complete, is to to Bairoch much and seems have lasted according sharper gence that covertheperiod 1870-1939. twoother series They GDP-per-capita than the oneobtained from Maddison's data. toBairoch, According de la longer areBairoch's those (1997) data,and* by Prados recently produced without around 1800andgoeson,almost starts 19 interrupcoefficientsthedivergence Cl andC2 show the GiniandTheil Escosura (2000). Figures 1 After is until Great until 890. stable the tion, that, inequality Depression, the same set of countries. For these alternative andcovering sources, using theWorld thelastdecadebefore WarII is there converduring inthethree databases theperiod thecountry 1870-1938, (Mad- andonly coverage in If we lookat Maddison's thedivergence however, data, thesame(see gence. andPrados de la Escosura) is practically begins Bairoch, dison, andreaches itspeakaround 1880.After hisseries is 1820(when before Prados de la Escosura's table originate) 1870, Cl). Forthe coverage period until World WarI anda somewhat there is atbestmild versus Bairoch andMaddison). that, limited 19 for more convergence (13 or 15 countries onein theinterwar Bairoch'sand Pradosde la Escosura'sdata are availableonlyfor faster period.22 we find that both GiniandTheil selected Bairoch's series, years.20 Using a very indexes arestable between 1890and1929, andthen strong display

APPENDIX C

and Maddison's data showincome maybe explained convergence by 19 PPPs by Pradosde la Escosura.The implication is dollars. theuse of current Bairoch's GDPs percapitaare givenin 1960 international have increased morein richthanin poor incurrent dollars ofequalpurchas- that de la Escosura's areexpressed Prados pricesof nontradables for a given are countries. so that between-country comparisons year ingpower parity, 22 Theincrease in inequality theIndustrial Revolution is also between theyears. The database is scaled butnotcomparisons following possible, ifoneusesBairoch's rather than Maddison's data. in sucha waythat theU.S. GDP percapita is equalto 1. much eachyear) (for According greater 20 to runtheusualconvergence the intercountry Gini in 1820 was 17.3. According Thisis thereason we cannot why regres- to Maddison, it was (forthesamesetof countries) sionson these data. Bairoch, only6 in 1800and9 in 21 Bairoch's data showpoorWENAO shortcut 1830.This is due to thefactthat dataare obtained Pradosde la Escosura's by theso-called in than Maddison's. Forexample, with GDPs percapita a regression between thepricelevel (purchasing countries that method, is, from higher the richest andpoorest WENAOcountry the ratio between ratedivided rate)on theleft-hand 1800and1830, exchange exchange by market power versus theNetherlands and theUnited other (respectively rateandseveral Finland) side,andtheGDP percapitaat current Kingdom exchange to Bairoch, 1.5 and 1.9.In Maddison's on theright-hand side.The was,according current account controls data, balance) respectively (openness, theratioin 1820 (United versus was 2.4. for which thedataare however, is run, ofcourse, for thecountries Finland) Kingdom only regression Bairoch's estimates of(relative) income with Consistently with Theestimated from suchan equation available. this, percapita together parameters countries at the timeof the Industrial usedto predict theprice of the future less developed variable arethen valuesforeachindependent Maddison's. Thusthestarting-point arehigher than level(that is, PPP) forthemissing (see Pradosde Revolution inequality yearsandcountries after the data is generally less in Bairoch's la Escosura, that Prados de la Escosura's case,andtheresulting 2000,pp. 8-11). Thefact divergence Revolution in theinterwar showincomedivergence greater. periodwhilebothBairoch's Industrial

This content downloaded from 41.141.98.200 on Tue, 18 Jun 2013 12:52:49 PM All use subject to JSTOR Terms and Conditions

670

THE REVIEW OF ECONOMICS AND STATISTICS


Figure C2.- Theil Coefficients,1800-1938 (Calculated Using Bairoch and Prados de la Escosura's Data)

Table Cl. - List of WENAO CountriesIncluded in Maddison's, Bairoch's, and Prados de la Escosura's Data sets Year 1870 Country Australia Austria Austria-Hungary Belgium Canada Switzerland Germany Denmark Spain Finland France UnitedKingdom Greece Ireland Italy Netherlands Norway New Zealand Portugal Sweden United States x x x x x x x x x x x x x x x x x x x x xxx xxx xxx xxx x xxx xxx xxx xxx xxx xxx xxx xxx xxx xxx xxx xxx xxx xxx Year 1890 xxx xxx xxx xxx xxx xxx xxx xxx xxx xxx xxx xxx xxx xxx xxx xxx xxx xxx Year 1900 xx'x Year 1913 xxx xxx xxx xxx xxx xxx xxx xxx xxx xxx xxx xxx xxx xxx xxx xxx xxx Year 1929 xxx xxx xxx xxx xxx xxx xxx xxx xxx xxx xxx xxx xxx xxx xxx xxx xxx xxx Year 1938 xx xx xx xx xx xx xx xx xx xx xx xx xx xx xx xx xx xx MBPMBPMBPMBPMBPMBP xx xx xxxxxxxxxx

xxx xxx xxx^xxx xxx xxx xxx xxx xxx xxx xxx xxx xxx xxx xxx xxx xxx xxx

xxxx

B = Bairoch, P = Pradosde la Escosura. Note:M = Maddison,

This content downloaded from 41.141.98.200 on Tue, 18 Jun 2013 12:52:49 PM All use subject to JSTOR Terms and Conditions

You might also like