3. 4 acemoglu the colonial

8/13/2019 3. 4 Acemoglu the Colonial

1/34

American Economic Association

The Colonial Origins of Comparative Development: An Empirical InvestigationAuthor(s): Daron Acemoglu, Simon Johnson, James A. RobinsonSource: The American Economic Review, Vol. 91, No. 5 (Dec., 2001), pp. 1369-1401Published by: American Economic AssociationStable URL: http://www.jstor.org/stable/2677930

Accessed: 21/05/2009 16:59

Your use of the JSTOR archive indicates your acceptance of JSTOR's Terms and Conditions of Use, available at

http://www.jstor.org/page/info/about/policies/terms.jsp. JSTOR's Terms and Conditions of Use provides, in part, that unless

you have obtained prior permission, you may not download an entire issue of a journal or multiple copies of articles, and you

may use content in the JSTOR archive only for your personal, non-commercial use.

Please contact the publisher regarding any further use of this work. Publisher contact information may be obtained at

http://www.jstor.org/action/showPublisher?publisherCode=aea.

Each copy of any part of a JSTOR transmission must contain the same copyright notice that appears on the screen or printed

page of such transmission.

JSTOR is a not-for-profit organization founded in 1995 to build trusted digital archives for scholarship. We work with the

scholarly community to preserve their work and the materials they rely upon, and to build a common research platform that

promotes the discovery and use of these resources. For more information about JSTOR, please contact [email protected].

American Economic Associationis collaborating with JSTOR to digitize, preserve and extend access to The

American Economic Review.

http://www.jstor.org
http://www.jstor.org/stable/2677930?origin=JSTOR-pdfhttp://www.jstor.org/page/info/about/policies/terms.jsphttp://www.jstor.org/action/showPublisher?publisherCode=aeahttp://www.jstor.org/action/showPublisher?publisherCode=aeahttp://www.jstor.org/page/info/about/policies/terms.jsphttp://www.jstor.org/stable/2677930?origin=JSTOR-pdf


2/34

The Colonial Originsof ComparativeDevelopment:An Empirical nvestigationBy DARON ACEMOGLU,SIMON JOHNSON,AND JAMESA. ROBINSON*

Weexploit differencesin European mortalityrates to estimate the effect of institu-tions on economic performance. Europeans adopted very different colonizationpolicies in differentcolonies, withdifferentassociated institutions. nplaces whereEuropeans aced high mortalityrates, they could not settle and were more likely toset up extractive institutions.These institutionspersisted to thepresent. Exploitingdifferences n Europeanmortalityrates as an instrumentor current nstitutions,weestimate large effects of institutions on income per capita. Once the effect ofinstitutions s controlledfor, countries n Africa or those closer to the equatordo nothave lower incomes. (JEL 011, P16, P51)

What are the fundamental causes of thelarge differences in income per capita acrosscountries? Although there is still little con-sensus on the answer to this question, differ-ences in institutions and property rights havereceived considerable attention in recentyears. Countries with better institutions,more secure property rights, and less distor-

tionary policies will invest more in physicaland human capital, and will use these factorsmore efficiently to achieve a greater level ofincome (e.g., Douglass C. North and RobertP. Thomas, 1973; Eric L. Jones, 1981; North,1981). This view receives some supportfromcross-country correlations between measuresof property rights and economic development(e.g., Stephen KnackandPhilip Keefer, 1995;Paulo Mauro, 1995; Robert E. Hall andCharles I. Jones, 1999; Dani Rodrik, 1999),and from a few micro studies that investigatethe relationship between property rights andinvestment or output (e.g., Timothy Besley,1995; ChristopherMazingo, 1999; Johnson etal., 1999).At some level it is obvious that institutionsmatter. Witness, for example, the divergentpaths of North and South Korea, or East andWest Germany,where one partof the countrystagnated under central planning and collec-tive ownership, while the other prosperedwith private property and a market economy.Nevertheless, we lack reliable estimates ofthe effect of institutions on economic perfor-mance. It is quite likely that rich economieschoose or can afford better institutions. Per-haps more important,economies that are dif-ferent for a variety of reasons will differ both

* Acemoglu: Department of Economics, E52-380b,Massachusetts Institute of Technology, Cambridge, MA02319, and Canadian Institute for Advanced Research(e-mail: [email protected]); ohnson:Sloan School of Man-agement, Massachusetts Institute of Technology, Cam-bridge,MA 02319 (e-mail: [email protected]);Robinson:Department of Political Science and Department of Eco-nomics, 210 BarrowsHall, Universityof California,Berke-ley, CA 94720 (e-mail: [email protected]).We thank Joshua Angrist, Abhijit Banerjee, Esther Duflo,Stan Engerman, John Gallup, Claudia Goldin, RobertHall, Chad Jones, Larry Katz, Richard Locke, AndreiShleifer, Ken Sokoloff, JudithTendler, three anonymousreferees, and seminar participants at the University ofCalifornia-Berkeley, Brown University, Canadian Insti-tute for Advanced Research, Columbia University, Har-vard University, Massachusetts Institute of Technology,National Bureau of Economic Research, NorthwesternUniversity, New York University, Princeton University,University of Rochester, Stanford University, ToulouseUniversity, University of California-Los Angeles, and theWorld Bank for useful comments. We also thank RobertMcCaa for guiding us to the data on bishops' mortality.

1369


3/34

1370 THE AMERICANECONOMICREVIEW DECEMBER2001in their institutions and in their income percapita.To estimate the impact of institutionson eco-nomic performance,we need a source of exog-enous variation n institutions. n this paper,wepropose a theory of institutional differencesamong countries colonized by Europeans,' andexploit this theoryto derive a possible source ofexogenous variation.Our theoryrests on threepremises:1. There were different types of colonizationpolicies which createddifferentsets of insti-tutions.At one extreme,Europeanpowers setup extractivetates, xemplifiedby the Bel-

gian colonizationof the Congo. These institu-tions did not introducemuch protection orprivateproperty,nor did they provide checksand balances against governmentexpropria-tion.Infact, the mainpurposeof the extractivestatewas to transferas muchof the resourcesof the colonyto the colonizer.At the other extreme, many Europeansmi-gratedand settled in a numberof colonies,creating what the historian Alfred Crosby(1986) calls Neo-Europes. he settlers riedto replicateEuropean nstitutions,with strongemphasis on private property and checksagainstgovernment ower. Primary xamplesof this includeAustralia,New Zealand,Can-ada,and the UnitedStates.2. The colonizationstrategywas influencedbythefeasibilityof settlements. nplaceswherethe disease environmentwas not favorable oEuropeansettlement, he cardswere stackedagainstthe creationof Neo-Europes,and theformation of the extractive state was morelikely.3. The colonial state and institutions persistedeven afterindependence.

Based on these three premises, we use themortalityrates expected by the first Europeansettlers in the colonies as an instrumentfor

current institutions in these countries.2 Morespecifically, our theory can be schematicallysummarizedas(potential) settler > settlementsmortality

early currentinstitutions institutionscurrentperformance.

We use dataon the mortality ates of soldiers,bishops, and sailors stationed n the colonies be-tween the seventeenthand nineteenthcenturies,largely based on the work of the historianPhilipD. Curtin.These give a good indicationof themortality ates faced by settlers.Europeanswerewell informedabout these mortalityratesat thetime, even though they did not know how tocontrol the diseasesthatcaused these high mor-talityrates.Figure 1 plots the logarithm of GDP percapita today againstthe logarithmof the settlermortalityrates per thousandfor a sampleof 75countries(see below fordata details). It shows astrong negative relationship. Colonies whereEuropeans faced higher mortality rates are to-day substantiallypoorer than colonies that werehealthy for Europeans.Our theory is that thisrelationshipreflects the effect of settler mortal-ity working throughthe institutionsbroughtbyEuropeans.To substantiate his, we regresscur-rent performanceon currentinstitutions, andinstrument he latter by settler mortality rates.Since our focus is on propertyrightsand checksagainst government power, we use the protec-tion against risk of expropriation ndex fromPolitical Risk Services as a proxy for institu-tions. This variable measures differences in in-stitutions originating from different types ofstates and state policies.3 There is a strong

1 By colonial experience we do not only mean thedirect control of the colonies by Europeanpowers, butmoregenerally, European nfluence on the rest of the world. Soaccording to this definition, Sub-Saharan Africa wasstrongly affected by colonialism between the sixteenthand nineteenthcenturiesbecause of the Atlanticslave trade.

2 Note that althoughonly some countrieswere colonized,thereis no selection biashere. This is because the questionwe are interested in is the effect of colonization policyconditional on being colonized.3Government expropriations not the only institutional

feature that matters. Ourview is that there is a clusterof


4/34

VOL.91 NO. 5 ACEMOGLUETAL.: THECOLONIALORIGINSOF DEVELOPMENT 1371

10 'Ivp

LO) < PANGA

tl FJ GUY AGOXi PAKIND SDN GMB0aco BGD NERMD NGAtl 6 ETH TA SIn- 6

0

2 4 6 8Logof SettlerMortalityFIGURE 1. REDUCED-FORM RELATIONSHIP BETWEEN INCOME AND SETTLER MORTALITY

(first-stage)relationshipbetween settler mortal-ity rates and current nstitutions,which is inter-esting in its own right. The regression showsthat mortalityrates faced by the settlers morethan 100 years ago explains over 25 percentof the variation n current nstitutions.4We alsodocumentthat this relationshipworks throughthe channels we hypothesize: (potential) settlermortality rates were a major determinant ofsettlements;settlements were a majordetermi-nant of early institutions (in practice, institu-tions in 1900); and thereis a strongcorrelationbetween early institutions and institutions to-day. Our two-stage least-squares stimateof theeffect of institutions on performanceis rela-tively precisely estimated and large. For ex-ample, it implies that improving Nigeria's

institutions to the level of Chile could, in thelong run, lead to as much as a 7-fold increase inNigeria's income (in practice Chile is over 11times as rich as Nigeria).The exclusion restriction mplied by our in-strumental variable regression is that, condi-tional on the controls ncluded n the regression,the mortality rates of European settlers morethan 100 years ago have no effect on GDP percapita today, other than their effect throughinstitutional development. The major concernwith this exclusion restriction s that the mor-tality rates of settlerscould be correlatedwiththe current disease environment, which mayhave a direct effect on economic performance.In this case, our instrumental-variables sti-mates may be assigning the effect of diseases onincome to institutions.We believe that this isunlikely to be the case and that our exclusionrestriction is plausible. The great majority ofEuropeandeaths in the colonies were causedbymalariaand yellow fever. Althoughthese dis-eases were fatal to Europeanswho had no im-munity, they had limited effect on indigenousadultswho had developed various types of im-munities. These diseases are thereforeunlikelyto be the reason why many countriesin Africaand Asia arevery poor today (see the discussionin Section III, subsection A). This notion is

institutions, ncluding constraintson governmentexpropri-ation, independent udiciary, property rights enforcement,and institutionsproviding equal access to education andensuring civil liberties, that are important to encourageinvestment and growth. Expropriation isk is related to allthese institutional eatures.In Acemoglu et al. (2000), wereportedsimilar results with otherinstitutionsvariables.4 Differences in mortalityrates are not the only, or eventhe main, cause of variation n institutions.For our empir-ical approach o work, all we need is thatthey are a sourceof exogenous variation.


5/34

1372 THEAMERICANECONOMICREVIEW DECEMBER2001supportedby the mortalityrates of local peoplein these areas.Forexample, Curtin 1968 Table2) reports hat the annualmortalityrates of localtroops serving with the British armyin Bengaland Madras were respectively 11 and 13 in1,000. These numbers are quite comparable o,in fact lower than, the annual mortalityrates ofBritish troops serving in Britain, which wereapproximately15 in 1,000. In contrast, he mor-tality rates of British troops serving in thesecolonies were much higherbecauseof their ackof immunity. For example, mortalityrates inBengal and Madrasfor Britishtroops were be-tween 70 and 170 in 1,000. The view that thedisease burden for indigenous adults was notunusual in places like Africa or India is alsosupportedby the relativelyhigh populationden-sities in these places before Europeansarrived(Colin McEvedy and RichardJones, 1975).We document hat our estimatesof the effectof institutionson performanceare not drivenbyoutliers.Forexample,excluding Australia,NewZealand,Canada,and the United States does notchange the results, nor does excluding Africa.Interestingly,we show that once the effect ofinstitutions on economic performance is con-trolledfor,neitherdistancefromtheequatornorthe dummy for Africa is significant.These re-sults suggest thatAfrica is poorerthan the restof the world not because of pure geographicor cultural factors, but because of worseinstitutions.The validityof ourapproach-i.e., our exclu-sion restriction-is threatened f other factorscorrelatedwiththe estimatesof settlermortalityaffect income per capita.We adopttwo strate-gies to substantiate that our results are notdrivenby omitted factors.First,we investigatewhether institutions have a comparableeffecton income once we control for a number ofvariablespotentiallycorrelatedwith settlermor-tality and economic outcomes. We find thatnone of these overturnourresults;the estimateschange remarkablyittle when we include con-trolsfor the identityof themaincolonizer, legalorigin, climate, religion, geography,naturalre-sources, soil quality,and measuresof ethnolin-guistic fragmentation.Furthermore,he resultsare also robust to the inclusion of controls forthe currentdisease environment e.g., the prev-alence of malaria,life expectancy, and infant

mortality)and the current raction of the popu-lation of Europeandescent.Naturally, it is impossible to control for allpossible variablesthat might be correlatedwithsettler mortality and economic outcomes. Fur-thermore,our empiricalapproachmight capturethe effect of settler mortality on economic per-formance, but working through otherchannels.We deal with these problems by using a simpleoveridentification est using measuresof Euro-pean migration to the colonies and early insti-tutions as additional instruments.We then useoveridentification ests to detect whether settlermortality has a direct effect on currentperfor-mance. The results are encouraging for ourapproach; hey generateno evidence for a directeffect of settler mortality on economicoutcomes.We are not aware of others who have pointedout the link between settler mortalityand insti-tutions, though scholars such as William H.McNeill (1976), Crosby (1986), and JaredM.Diamond(1997) havediscussedthe influence ofdiseases on humanhistory.Diamond(1997), inparticular, emphasizes comparative develop-ment, but his theoryis based on the geograph-ical determinants of the incidence of theneolithic revolution.He ignores boththe impor-tance of institutionsand the potentialcauses ofdivergencein more recentdevelopment,whichare the main focus of our paper.Workby Ro-nald E. Robinson and John Gallagher(1961),Lewis H. Gann and Peter Duignan (1962),DonaldDenoon (1983), andPhilip J. Cain andAnthony G. Hopkins (1993) emphasizes thatsettler colonies such as the United States andNew Zealandare differentfrom other colonies,andpointout that these differences wereimpor-tant for their economic success. Nevertheless,this literaturedoes not developthe link betweenmortality, settlements,and institutions.Ourarguments most closely relatedto workon the influence of colonial experienceon insti-tutions. FrederichA. von Hayek (1960) arguedthat the British common law traditionwas su-periorto the Frenchcivil law, which was devel-oped during the Napoleonic era to restrainjudges' interferencewith statepolicies (see alsoSeymour M. Lipset, 1994). More recently,Rafael La Porta et al. (1998, 1999) emphasizethe importanceof colonial origin (the identityof


6/34

VOL 91 NO. 5 ACEMOGLUETAL: THECOLONIALORIGINSOF DEVELOPMENT 1373the colonizer) and legal origin on current nsti-tutions, and show that the common-law coun-tries and former British colonies have betterproperty rights and more developed financialmarkets. Similarly, David Landes (1998 Chap-ters 19 and 20) and North et al. (1998) arguethat former British colonies prosperedrelativeto formerFrench, Spanish, and Portuguesecol-onies because of the good economic andpolit-ical institutionsand culture they inheritedfromBritain. In contrast to this approach whichfocuses on the identity of the colonizer, weemphasize the conditions in the colonies. Spe-cifically, in our theory-and in the data-it isnot the identity of the colonizer or legal originthatmatters,but whetherEuropeancolonialistscould safely settle in a particular location:where they could not settle, they createdworseinstitutions. In this respect, our argument isclosely related to that of Stanley L. Engermanand Kenneth L. Sokoloff (1997) who also em-phasize institutions,but link them to factor en-dowmentsand inequality.Empirically,our work is related to a numberof other attempts to uncover the link betweeninstitutions and development, as well as toGraziella Bertocchi and Fabio Canova (1996)andRobinM. Grier(1999), who investigatetheeffect of being a colony on postwar growth.Two papers deal with the endogeneity of in-stitutions by using an instrumental variablesapproachas we do here. Mauro (1995) instru-ments for corruptionusing ethnolinguistic rag-mentation.Hall and Jones (1999), in turn,usedistancefrom the equatoras an instrument orsocial infrastructurebecause, they argue, lati-tude is correlated with Western influence,which leads to good institutions.The theoreticalreasoningfor these instruments s not entirelyconvincing. It is not easy to argue that theBelgian influence in the Congo, or Westerninfluence in the Gold Coast duringthe era ofslavery promoted good institutions. Ethnolin-guistic fragmentation, n the otherhand,seemsendogenous, especially since such fragmenta-tion almost completely disappeared n Europeduring the era of growth when a centralizedstate and market emerged (see, e.g., EugenJ. Weber, 1976; Benedict Anderson, 1983).Econometrically, he problemwith both studiesis that their instruments can plausibly have a

direct effect on performance.For example, Wil-liiam Easterly and Ross Levine (1997) arguethat ethnolinguistic fragmentation can affectperformance by creating political instability,while Charles de Montesquieu [1748] (1989)and more recently David E. Bloom and JeffreyD. Sachs (1998) and John Gallup et al. (1998)argue for a direct effect of climate on perfor-mance. If, indeed, these variables have a directeffect, they are invalid instrumentsand do notestablish that it is institutions that matter. Theadvantageof ourapproachs that conditionalonthe variableswe alreadycontrolfor, settlermor-tality more than 100 years ago shouldhave noeffect on output today, other than through itseffect on institutions. Interestingly,our resultsshow that distance from the equator does nothave an independenteffect on economic perfor-mance, validatingthe use of this variable as aninstrument in the work by Hall and Jones(1999).The next section outlines our hypothesis andprovides supportinghistoricalevidence. SectionII presents OLS regressions of GDP per capitaon our index of institutions. Section III de-scribes our key instrument or institutions, themortalityrates faced by potential settlersat thetime of colonization. Section IV presents ourmainresults.Section V investigatesthe robust-ness of our results, and Section VI concludes.

I. The Hypothesis and Historical BackgroundWe hypothesizethatsettlermortalityaffectedsettlements;settlements affected early institu-tions; and early institutions persisted andformed the basis of current nstitutions.In thissection, we discussand substantiate hishypoth-

esis. The next subsection discusses the link be-tween mortalityrates of settlers and settlementdecisions, then we discuss differences in colo-nization policies, and finally, we turn to thecauses of institutionalpersistence.A. Mortalityand Settlements

There is little doubtthatmortalityrates werea key determinantof European settlements.Curtin(1964, 1998) documents how both theBritish and Frenchpressinformedthe publicofmortalityrates in the colonies. Curtin (1964)


7/34

1374 THEAMERICANECONOMICREVIEW DECEMBER2001also documents how early British expectationsfor settlement in West Africa were dashed byvery high mortality among early settlers, abouthalf of whom could be expected to die in thefirst year. In the Provinceof Freedom SierraLeone), Europeanmortality n the first year was46 percent,in Bulama(April 1792-April 1793)there was 61-percent mortality among Europe-ans. In the firstyear of the Sierra Leone Com-pany (1792-1793), 72 percent of the Europeansettlers died. On Mungo Park's Second Expedi-tion (May-November 1805), 87 percentof Eu-ropeans died during the overland trip fromGambia o theNiger,andall the Europeansdiedbefore completingthe expedition.

An interesting example of the awarenessofthe disease environment comes from the Pil-grim fathers. They decided to migrate to theUnited Statesrather hanGuyanabecauseof thehigh mortality rates in Guyana (see Crosby,1986 pp. 143-44). Another example comesfrom the BeauchampCommittee n 1795, set upto decide where to send British convicts whohad previously been sent to the United States.One of the leading proposalswas the island ofLemane, up the Gambia River. The committeerejected this possibility because they decidedmortalityrates would be too high even for theconvicts. SouthwestAfrica was alsorejected orhealth reasons. The final decision was to sendconvicts to Australia.The eventual expansionof many of the col-onies was also related to the living conditionsthere. In places where the early settlers facedhigh mortalityrates,there would be less incen-tive for new settlers to come.5

B. Types of Colonizationand SettlementsThe historical evidence supportsboth the no-tion that there was a wide range of differenttypes of colonization and that the presence orabsence of Europeansettlers was a key deter-minantof theformcolonialism took.Historians,

includingRobinson and Gallagher 1961), Gannand Duignan (1962), Denoon (1983), and Cainand Hopkins (1993), have documented the de-velopment of settlercolonies, where Europe-ans settled in large numbers, and life wasmodeled afterthe home country.Denoon (1983)emphasizesthat settlercolonies hadrepresenta-tive institutions which promotedwhat the set-tlers wanted and that what they wanted wasfreedom and the ability to get rich by engagingin trade.He arguesthat therewas undeniablysomething capitalist in the structure of thesecolonies. Private ownership of land and live-stock was well established very early .. (p.35).Whenthe establishmentof European-liken-stitutions did not arise naturally, the settlerswere readyto fight for them against the wishesof the home country.Australia s an interestingexamplehere. Most of the early settlers n Aus-traliawere ex-convicts, but the land was ownedlargely by ex-jailors, and there was no legalprotection against the arbitrarypower of land-owners. The settlers wantedinstitutionsandpo-litical rightslike those prevailingin Englandatthe time. They demandedjury trials, freedomfrom arbitraryarrest,and electoral representa-

tion. Althoughthe British governmentresistedat first,the settlersargued hatthey wereBritishand deserved the same rights as in the homecountry (see Robert Hughes, 1987). Cain andHopkins write (1993 p. 237) from the late1840s the British bowed to local pressuresand,in line with observed constitutional changestakingplace in Britainherself, acceptedthe ideathat, in mature colonies, governors should infuture form ministries from the majorityele-ments in elected legislatures. They also sug-gest that the enormous boom in publicinvestmentafter 1870 [in New Zealand] .. wasan attempt to build up an infrastructure.. tomaintain high living standardsin a countrywhere voters expected politicians actively topromotetheireconomic welfare. (p. 225).65 Naturally,otherfactors also influencedsettlements.Forexample, despite the relatively high mortality rates, manyEuropeans migrated to the Caribbeanbecause of the veryhigh incomes thereat the time (see, e.g., Richard S. Dunn,1972; David W. Galenson, 1996; Engermanand Sokoloff,

1997; David Eltis, 2000).

6 RobertH. Bates (1983 Chapter3) gives a nice exampleof the influenceof settlers on policy in Africa. The Britishcolonial governmentpursued many policies that depressedthe price of cocoa, the main produce of the farmers inGhana. In contrast,the British government supportedtheprices faced by the commercial cereal farmers in Kenya.


8/34

VOL.91 NO. 5 ACEMOGLUETAL.: THECOLONIALORIGINSOF DEVELOPMENT 1375This is in sharpcontrast o the colonial expe-rience in Latin America during the seventeenthand eighteenthcenturies,and in Asia andAfrica

during the nineteenth and early twentieth cen-turies. The main objective of the Spanish andthe Portuguese colonization was to obtaingoldand other valuables from America. Soon afterthe conquest, the Spanish crown grantedrightsto land andlabor(the encomienda)andset up acomplex mercantilist ystem of monopolies andtrade regulations to extract resources from thecolonies.7Europeans developed the slave trade in Af-rica for similar reasons. Before the mid-nine-teenth century, colonial powers were mostlyrestricted o the African coast and concentratedon monopolizingtrade n slaves, gold, and othervaluable commodities-witness the names usedto describe West Aflican countries: the GoldCoast,theIvoryCoast.Thereafter, olonial pol-icy was driven in part by an element of super-power rivalry,butmostly by economic motives.Michael Crowder (1968 p. 50), for example,notes it is significant hatBritain's largestcol-ony on the West Coast [Nigeria] should havebeen the one where hertraderswere most activeand bears out the contention that, for Britain... flag followed trade. 8Lance E. Davis andRobert A. Huttenback(1987 p. 307) concludethat the colonial Empire provides strong evi-dence for the belief that governmentwas at-

tuned to the interests of business and willing todivert resources to ends that the business com-munitywouldhave found profitable. They findthat before 1885 investment in the British em-pire had a return25 percent higher than that ondomestic investment, hough afterwards he twoconverged. Andrew Roberts (1976 p. 193)writes: [from] .. 1930 to 1940 Britainhadkeptfor itself 2,400,000 pounds in taxes from theCopperbelt,while NorthernRhodesia receivedfrom Britain only 136,000 pounds in grantsfor development. Similarly, Patrick Manning(1982) estimates that between 1905 and 1914,50 percentof GDP in Dahomey was extractedby the French, and Crawford Young (1994 p.125) notes that tax rates in Tunisia were fourtimes as high as in France.Probably he most extreme case of extractionwas that of King Leopold of Belgium in theCongo. Gann and Duignan (1979 p. 30) arguethat following the example of the Dutch inIndonesia, Leopold's philosophy was that thecolonies should be exploited, not by the opera-tion of a marketeconomy,butby state interven-tion andcompulsorycultivationof cashcrops tobe sold to and distributedby the state at con-trolled prices. Jean-Philippe Peemans (1975)calculates that tax rates on Africans in theCongo approached60 percent of their incomeduring the 1920's and 1930's. Bogumil Jew-siewicki (1983) writes that during the periodwhen Leopold was directly in charge, policywas basedon theviolent exploitationof naturaland humanresources, with a consequent de-structionof economic and social life ... [and]... dismembermentof political structures.Overall, there were few constraintson statepower in the nonsettlercolonies. The colonialpowersset up authoritarianndabsolutiststateswith thepurposeof solidifyingtheircontrol andfacilitatingthe extraction of resources.Young(1994 p. 101) quotesa Frenchofficial in Africa:the Europeancommandant is not posted toobservenature, .. He has a mission ... to imposeregulations, to limit individual liberties..., tocollect taxes. Manning (1988 p. 84) summa-rizes this as: InEuropethe theories of repre-sentativedemocracywon out over the theoristsof absolutism.... But in Africa, the Europeanconquerors et upabsolutistgovernments,basedon reasoningsimilar to thatof Louis XIV.

Bates shows that this was mainly because in Kenya, butnotin Ghana, there were a significant number of Europeansettler farmers, who exerted considerable pressure onpolicy.7 See James Lang (1975) and James Lockhart andStuart B. Schwartz(1983). Migration to Spanish Americawas limited by the Spanish Crown, in part because of adesire to keep control of the colonists and limit theirindependence (see, for example, John H. Coatsworth,1982). This also gives further support to our notion thatsettlers were able to influence the type of institutions setup in the colonies, even against the wishes of the homecountry government.8 Although in almost all cases the main objective ofcolonial policies was to protect economic interests andobtain profits, the recipientsof these profits varied. In thePortuguesecase, it was the state; n the Belgian case, it wasKing Leopold; and in the British case, it was often privateenterpriseswho obtained concessions or monopoly trading

rightsin Africa (Crowder, 1968 Part III).


9/34

1376 THE AMERICANECONOMICREVIEW DECEMBER2001C. InstitutionalPersistence

There s a varietyof historical vidence,as wellas our regressions n Table 3 below, suggestingthat the controlstructures et up in the nonsettlercolonies duringthe colonial era persisted,whilethere s little doubt hatthe institutions f law andorderand privateproperty stablishedduring heearly phases of colonialism n Australia,Canada,New Zealand, he UnitedStates,Hong Kong, andSingaporehave formed the basis of the current-day institutions f these countries.9Young emphasizes thatthe extractive nstitu-tions set up by the colonialists persisted longafter the colonial regime ended. He writes al-though we commonly described the indepen-dent polities as 'new states,'in reality theyweresuccessors to the colonial regime, inheriting tsstructures, ts quotidianroutinesand practices,andits more hiddennormative heories of gov-ernance (1994 p. 283). An example of thepersistence of extractive state institutions intothe independenceera is provided by the persis-tence of the most prominentextractivepolicies.In Latin America, the full panoplyof monopo-lies and regulations,which had been createdbySpain, remained intact after independence,formost of the nineteenth century. Forced laborpolicies persistedand were even intensified orreintroducedwith the expansion of export agri-culturein the latter partof the nineteenthcen-tury. Slavery persisted n Brazil until 1886, andduring the sisal boom in Mexico, forced laborwas reintroduced ndpersistedup to the startofthe revolution in 1910. Forced labor was alsoreintroduced n Guatemala and El Salvador toprovide labor for coffee growing.In the Guate-malancase, forcedlabor asted until the creationof democracy in 1945. Similarly, forced laborwas reinstated in many independent Africancountries,for example, by Mobutuin Zaire.

There are a numberof economic mechanismsthat will lead to institutionalpersistence of thistype. Here, we discuss threepossibilities.(1) Setting up institutions hat place restrictionson governmentpower and enforce propertyrights is costly (see, e.g., Acemoglu andThierryVerdier, 1998). If the costs of cre-ating these institutions have been sunk bythe colonial powers, thenit maynot pay theelites at independenceto switch to extrac-tive institutions.In contrast,when the newelites inherit extractive institutions, theymay not want to incur the costs of introduc-ing better institutions, and may instead

prefer to exploit the existing extractive in-stitutionsfor theirown benefits.(2) The gains to an extractive strategy maydependon the size of the rulingelite. Whenthis elite is small,eachmemberwould havea largershareof the revenues, so the elitemay have a greater ncentive to be extrac-tive. Inmanycases whereEuropeanpowersset up authoritariannstitutions, they dele-gatedthe day-to-dayrunningof the state toa small domestic elite. This narrowgroupoften was the one to control the stateafter independenceand favored extractiveinstitutions. 0(3) If agentsmake irreversible nvestmentsthatare complementaryto a particularset ofinstitutions, they will be more willing tosupport hem, makingthese institutionsper-sist (see, e.g., Acemoglu, 1995). Forexam-ple, agents who have invested in humanandphysical capitalwill be in favor of spending

9The thesis that institutionspersist for a long time goesback at least to KarlA. Wittfogel 1957), who argued hat hecontrolstructures et up by the large hydraulic mpires uchas China,Russia, and the OttomanEmpirepersisted or morethan 500 years to the twentieth century. Engerman andSokoloff (1997), La Porta et al. (1998, 1999), North et al.(1998), and Coatsworth 1999) also argue that colonial insti-tutionspersisted.Engerman t al. (1998) provide urther vi-dence supporting his view.

10 William Reno (1995), for example, argues that thegovernmentsof postindependenceSierraLeone adopted hetactics and institutionsof the British colonizers to cementtheir political power and extract resourcesfrom the rest ofsociety. CatherineBoone (1992) providesa similar analysisof the evolution of the modern state in Senegal. Mostscholars also view the roots of authoritarianismunderMobutu n the colonial state practices n the Belgian Congo(e.g., Thomas M. Callaghy, 1984, or Thomas Turner andYoung, 1985, especially p. 43). The situation in LatinAmerica is similar. Independenceof most Latin Americancountlies came in the early nineteenthcentury as domesticelites took advantageof the invasionof Spain by Napoleonto capturethe control of the state. But, the only thing thatchanged was the identity of the recipientsof the rents (see,for example, Coatsworth, 1978, or JohnLynch, 1986).


10/34

VOL 91 NO. 5 ACEMOGLUETAL.: THECOLONIALORIGINSOF DEVELOPMENT 1377TABLE-DESCRIPTIVETATISTICS

By quartilesof mortalityWhole world Base sample (1) (2) (3) (4)

Log GDP per capita (PPP)in 1995 8.3 8.05 8.9 8.4 7.73 7.2(1.1) (1.1)Log output per worker n 1988 -1.70 -1.93 -1.03 -1.46 -2.20 -3.03(with level of United States (1.1) (1.0)normalized o 1)Average protectionagainst 7 6.5 7.9 6.5 6 5.9expropriation isk, 1985-1995 (1.8) (1.5)Constrainton executive in 1990 3.6 4 5.3 5.1 3.3 2.3(2.3) (2.3)Constrainton executive in 1900 1.9 2.3 3.7 3.4 1.1 1(1.8) (2.1)Constrainton executive in first year 3.6 3.3 4.8 2.4 3.1 3.4of independence (2.4) (2.4)

Democracyin 1900 1.1 1.6 3.9 2.8 0.19 0(2.6) (3.0)European ettlements n 1900 0.31 0.16 0.32 0.26 0.08 0.005(0.4) (0.3)Log European ettler mortality n.a. 4.7 3.0 4.3 4.9 6.3(1.1)Number of observations 163 64 14 18 17 15

Notes: Standarddeviationsare in parentheses.Mortality s potentialsettlermortality,measured n termsof deathsperannumper 1,000 meanstrength raw mortalitynumbersareadjusted o whatthey wouldbe if a force of 1,000 living people werekeptin place for a whole year, e.g., it is possible for this number o exceed 1,000 in episodes of extrememortalityas thosewho die arereplacedwith new arrivals).Sources and methods for mortalityare described n Section III, subsection B, andin theunpublishedAppendix (available romtheauthors;or see Acemogluet al., 2000). Quartilesof mortalityare for ourbasesampleof 64 observations.These are:(1) less than65.4; (2) greater han or equalto 65.4 and less than78.1; (3) greater hanor equal to 78.1 and less than 280; (4) greaterthan or equal to 280. The numberof observationsdiffers by variable;seeAppendixTable Al for details.

money to enforce property rights, whilethose who have less to lose may not be.II. Institutionsand Performance:OLS Estimates

A. Data and Descriptive StatisticsTable 1 provides descriptive statistics for thekey variablesof interest.The first column is forthe whole world,and column(2) is for our basesample, limited to the 64 countries that wereex-colonies and for which we have settler mor-tality, protectionagainst expropriation isk, andGDP data (this is smaller than the sample inFigure 1). The GDP per capita in 1995 is PPPadjusted(a more detailed discussion of all datasources is provided in Appendix Table Al).Income (GDP) per capitawill be our measureofeconomic outcome. There are large differences

in income per capita in both the world sample

and our basic sample, and the standarddevia-tion of log income per capita in both cases is1.1. In row 3, we also give outputper worker n1988 from Hall and Jones (1999) as an alterna-tive measure of income today. Hall and Jones(1999) prefer this measure since it explicitlyrefers to worker productivity. On the otherhand, given the difficultyof measuringthe for-mal laborforce, it may be a more noisy measureof economic performance than income percapita.We use a varietyof variables to capture n-stitutional differences. Our main variable, re-ported in the second row, is an index ofprotectionagainst expropriation.These data arefrom Political Risk Services (see, e.g., WilliamD. Coplin et al., 1991),andwere firstused in theeconomics and political science literaturesbyKnack and Keefer (1995). Political Risk Ser-vices reportsa value between 0 and 10 for eachcountry and year, with 0 corresponding o the


11/34


12/34

VOL. 91 NO. 5 ACEMOGLUETAL.: THECOLONIALORIGINSOF DEVELOPMENT 1379TABLE 2-OLS REGRESSIONS

Whole Base Whole Whole Base Base Whole Baseworld sample world world sample sample world sample(1) (2) (3) (4) (5) (6) (7) (8)

Dependent variableis log outputperDependentvariable is log GDP per capitain 1995 workerin 1988Average protection 0.54 0.52 0.47 0.43 0.47 0.41 0.45 0.46against expropriation (0.04) (0.06) (0.06) (0.05) (0.06) (0.06) (0.04) (0.06)risk, 1985-1995Latitude 0.89 0.37 1.60 0.92(0.49) (0.51) (0.70) (0.63)Asia dummy -0.62 -0.60(0.19) (0.23)Africa dummy -1.00 -0.90

(0.15) (0.17)Other ontinent dummy -0.25 -0.04(0.20) (0.32)R2 0.62 0.54 0.63 0.73 0.56 0.69 0.55 0.49Number of observations 110 64 110 110 64 64 108 61Notes: Dependentvariable:columns(1)-(6), log GDPper capita (PPP basis) in 1995, currentprices (fromthe World Bank'sWorldDevelopmentIndicators1999); columns (7)-(8), log outputper worker n 1988 from Hall and Jones (1999). Averageprotectionagainstexpropriation iskis measuredon a scale from 0 to 10, wherea higherscore means moreprotectionagainstexpropriation, veragedover 1985 to 1995, from Political Risk Services. Standard rrorsare in parentheses.In regressionswith continentdummies,the dummyfor America is omitted. See AppendixTableAl for more detailed variabledefinitionsand sources. Of the countries n our base sample,Hall and Jones do not reportoutputper worker n the Bahamas, Ethiopia,and Vietnam.

Sachs and coauthors, have arguedfor a directeffect of climate on performance,and Gallup etal. (1998) and Hall and Jones (1999) documentthe correlationbetween distance from the equa-tor and economic performance.To control forthis, in columns (3)-(6), we add latitude as aregressor (we follow the literature n using theabsolutevalue measureof latitude, .e., distancefromthe equator,scaled between0 and 1). Thischanges the coefficient of the index of institu-tions little. Latitude tself is also significantandhas the sign found by the previous studies. Incolumns (4) and (6), we also add dummies forAfrica, Asia, and other continents, with Amer-ica as the omitted group. Although protectionagainst expropriationrisk remains significant,the continentdummies arealso statisticallyandquantitatively ignificant.The Africa dummy incolumn(6) indicates thatin oursample Africancountries are 90 log points (approximately145percent) poorereven after taking the effect ofinstitutions nto account.Finally, in columns (7)

and (8), we repeat our basic regressions usingthe log of output per worker from Hall andJones (1999), with very similar results.Overall,the results in Table 2 show a strongcorrelation between institutions and economicperformance.Nevertheless, there are a numberof important reasons for not interpreting hisrelationship as causal. First, rich economiesmay be able to afford, or perhapsprefer, betterinstitutions.Arguably more important han thisreverse causality problem,there are many omit-ted determinants f income differences that willnaturallybe correlatedwith institutions.Finally,the measures of institutions are constructedexpost, and the analysts may have had a naturalbias in seeing better nstitutions n richerplaces.As well as these problems introducingpositivebias in the OLS estimates, the fact that theinstitutionsvariable s measuredwith consider-able errorandcorrespondspoorlyto the clusterof institutions that matter in practice createsattenuation and may bias the OLS estimates


13/34

1380 THEAMERICANECONOMICREVIEW DECEMBER2001

10 HKG S CAN

r) | ~~~~~~~~~~~~MLTBHSGM PER DOMTW)

N- 8 SLV BO[GU IDN. ~~~HTI SDN 'MM TGO

RD 7 A l NEBRGDGA0~~~~~~~~a) EhE TZAm~ 6m0

4'4 6 8 10Average ExpropriationRisk 1985-95

FIGURE 2. OLS RELATIONSHIPBETWEEN EXPROPRIATIONRISK AND INCOME

downwards. All of these problems could besolved if we had an instrument or institutions.Such an instrumentmust be an important actorin accountingfor the institutionalvariation hatwe observe, buthave no direct effect on perfor-mance. Ourdiscussion in Section I suggests thatsettler mortalityduringthe time of colonizationis a plausible instrument.

III. Mortalityof Early SettlersA. Sources of European Mortalityin the Colonies

In this subsection, we give a brief overviewof the sources of mortalityfacing potentialset-tlers. Malaria particularlyPlasmodium alcipo-rum) and yellow fever were the major sourcesof European mortality in the colonies. In thetropics, these two diseases accounted for 80percent of Europeandeaths, while gastrointes-tinal diseases accountedfor another15 percent(Curtin,1989 p. 30). Throughout he nineteenthcentury,areaswithoutmalariaandyellow fever,such as New Zealand,were more healthythanEurope because the major causes of death inEurope-tuberculosis, pneumonia, and small-pox-were rare in these places (Curtin, 1989p. 13).

Both malaria and yellow fever are transmit-ted by mosquitovectors. In the case of malaria,the main transmitter s the Anopheles gambiaecomplex and the mosquito Anopheles unestus,while the main carrierof yellow fever is Aedesaegypti. Both malaria and yellow fever vectorstend to live close to human habitation.In places where the malariavectoris present,such as the West African savanna or forest, anindividual can get as many as several hundredinfectious mosquitobites a year. For a personwithout immunity, malaria (particularlyPlas-modium falciporum)is often fatal, so Europe-ansin Africa, India,or the Caribbean aced veryhigh death rates.In contrast,death ratesfor theadult local populationwere much lower (seeCurtin[1964] and the discussion in our intro-duction above).Curtin 1998 pp.7-8) describesthis as follows:

Childrenin West Africa ... would be in-fected with malariaparasitesshortly afterbirthand were frequentlyreinfectedafter-wards; if they lived beyond the age ofaboutfive, they acquiredan apparentm-munity. The parasiteremainedwith them,normallyin the liver, but clinical symp-toms were rare so long as they continuedto be infected with the same species of P.falciporum.


14/34

VOL.91 NO. 5 ACEMOGLUETAL.: THE COLONIALORIGINSOF DEVELOPMENT 1381The more recent books on malariology confirmthis conclusion. For example, In stable en-demic areas a heavy toll of morbidityand mor-tality falls on young children but malaria is arelatively mild conditionin adults HerbertM.Gilles and David A. Warrell, 1993 p. 64; seealso the classic referenceon this topic, LeonardJ. Bruce-Chwatt, 1980 Chapter4; Roy Porter,1996).12Similarly,the World Health Organiza-tion (WHO) points out thatin endemic malariaareas of Africa and the Western Pacific today... the risk of malaria severity and death isalmostexclusively limited to non-immunes,be-ing most serious for young children over sixmonthsof age... survivingchildrendevelop theirown immunitybetween the age of 3-5 years(JoseA. Najeraand JoahimHempel, 1996).People in areas where malaria s endemic arealso more likely to have genetic immunityagainstmalaria.Forexample, they tend to havethe sickle-cell trait,which discouragesthe mul-tiplicationof parasites n the blood, or deficien-cies in glucose-6-phosphatedehydrogenaseandthalassaemiatraits,which also protect againstmalaria.Porter(1996 p. 34) writes: Insuch aprocess,..., close to 100 percent of Africansacquired a genetic trait that protects themagainstvivax malariaandprobably againstfal-ciporummalariaas well. Overall, he WHO es-timatesthat malariakills about1 millionpeopleper year,mostof them children. t does not,how-ever, generallykill adultswhogrew upinmalaria-endemic areas(see NajeraandHempel, 1996).Although yellow fever's epidemiology isquite different from malaria, t was also muchmore fatal to Europeans han to non-Europeanswho grew up in areaswhereyellow fever com-monly occurred.13 Yellow fever leaves its sur-viving victims with a lifelong immunity,whichalso explainsits epidemic pattern,relying on aconcentrated nonimmune population. Curtin

(1998 p. 10) writes: Because most Africanshad passed through a light case early in life,yellow fever in West Africa was a strangers'disease, attacking those who grew up else-where. Similarly, Michael B. A. Oldstone(1998 p. 49) writes:

Most Black Africans and their descen-dants respond to yellow fever infectionwith mild to moderate symptoms such asheadache, fever, nausea, and vomiting,and then recoverin a few days. This out-come reflects the long relationship be-tween the virus and its indigenous hosts,who through generationsof exposure tothe virus have evolved resistance.In contrast, fatality rates among nonimmuneadults, such as Europeans,could be as high as90 percent.Advances in medical science have reducedthe danger posed by malaria and yellow fever.Yellow fever is mostly eradicated (Oldstone,1998 Chapter 5), and malaria has been eradi-cated in many areas. Europeans developedmethods of dealing with these diseases thatgraduallybecame more effective in the secondhalf of the nineteenth century. For example,they came to understandthat high doses ofquinine,derived from the cinchonabark,actedas a prophylactic and prevented infection orreduced the severity of malaria. They alsostartedto undertakeserious mosquito eradica-tion efforts andprotectthemselvesagainstmos-quitobites. Further,Europeansalso learned hatan often effective method of reducing mortalityfrom yellow fever is flightfrom the area,sincethe transmittermosquito,Aedes aegypti,has onlya shortrange. Nevertheless,duringmuch of thenineteenth entury, here was almost a completemisunderstandingf the natureof both diseases.For example,the leading theoryfor malariawasthat t wascausedby miasma romswamps,andquininewas not used widely. The role of smallcollections of water to breed mosquitoes andtransmit hese diseaseswasnotunderstood.twasonly in the late nineteenth entury hatEuropeansstarted o control hese diseases.4

12 Because malariaspecies arequite local, a person mayhave immunity o the local versionof malaria,but be highlyvulnerableto malariaa short distance away. This is proba-bly the explanation or why Africans had suchhigh mortal-ity when they were forced to move by colonial powers.(Curtinet al., 1995 p. 463).13 Because yellow fever struck Europeans as an epi-demic, many of the very high death rates we reportbelowfor European roops are from yellow fever.

14 Even during the early twentieth century, there wasmuch confusion about the causes of malariaand yellow


15/34

1382 THEAMERICANECONOMICREVIEW DECEMBER2001These considerations, ogether with the datawe have on the mortality of local people andpopulationdensities before the arrivalof Euro-

peans, make us believe that settler mortality s aplausible instrument for institutional develop-ment: these diseases affected European settle-ment patterns and the type of institutions theyset up, but had little effect on the health andeconomy of indigenouspeople.15A final noteworthy feature, helpful in in-terpreting our results below, is that malariaprevalence depends as much on the microcli-mate of an area as on its temperature andhumidity, or on whether it is in the tropics;high altitudes reduce the risk of infection, soin areas of high altitude, where hill stationscould be set up, such as Bogota in Colombia,mortality rates were typically lower than inwet coastal areas. However, malaria couldsometimes be more serious in high-altitudeareas.For example, Curtin(1989 p. 47) pointsout that in Ceylon mortality was lower in thecoast than the highlands because rains in thecoast washed away the larvae of the transmit-ter mosquitoes. Similarly, in Madras manycoastal regions were free of malaria, whilenorthern India had high rates of infection.Curtin (1998 Chapter 7) also illustrates howthere were marked differences in the preva-lence of malaria within small regions ofMadagascar. This suggests that mortality

rates faced by Europeans are unlikely to be aproxy for some simple geographic or climac-tic feature of the country.B. Data on Potential Settler Mortality

Our data on the mortality of European set-tlers come largely from the work of PhilipCurtin. Systematic military medical recordkeeping began only after 1815, as an attemptto understand why so many soldiers weredying in some places. The first detailed stud-ies were retrospective and dealt with Britishforces between 1817 and 1836. The UnitedStates and French governments quicklyadopted similar methods (Curtin, 1989 pp. 3,5). Some early data are also available for theDutch East Indies. By the 1870's, most Euro-pean countries published regular reports onthe health of their soldiers.The standardmeasure is annualized deathsper thousand mean strength. This measurereports the death rate among 1,000 soldierswhereeach death s replacedwith a new soldier.Curtin(1989, 1998) reviews in detail the con-structionof these estimatesfor particularplacesandcampaigns,and assesses which datashouldbe considered reliable.Curtin (1989), Death by Migration, dealsprimarily with the mortality of Europeantroops from 1817 to 1848. At this time mod-ern medicine was still in its infancy, and theEuropean militaries did not yet understandhow to control malaria and yellow fever.These mortality rates can therefore be inter-preted as reasonable estimates of settler mor-tality. They are consistent with substantialevidence from other sources (see, for exam-ple, Curtin[1964, 1968]). Curtin(1998), Dis-ease and Empire, adds similar data on themortalityof soldiers in the second half of thenineteenth century.16 In all cases, we use the

fever. The WashingtonPost on Nov. 2, 1900 wrote: Of allthe silly and nonsensicalrigmaroleof yellow fever that hasyet found its way into print .. the silliestbeyond compare sto be found in the argumentsand theories generatedby amosquito hypothesis (quoted in Oldstone, 1998 pp.64-65).Many campaignsduringthe nineteenthcenturyhad veryhigh mortalityrates.For example, the French campaigninMadagascarduringthe 1890's and Frenchattempts o buildthe Panama Canal duringthe 1880's were mortality disas-ters, the first due to malaria, he second due to yellow fever(see Curtin,1998, and David McCullogh,1977). In Panama,to stop ants the Frenchused water pots under the legs ofbeds in barracksandhospitals.Thesepotsprovidedan idealmilieu for the breedingof Aedes aegypti, causing very highratesof mortality(Oldstone, 1998 p. 66).15 In Acemoglu et al. (2001), we document hat many ofthese areas in the tropical zone were richer and moredenselysettled in 1500 than the temperateareas ater settledby the Europeans.This also supportsthe notion that thedisease environmentdidnot create an absolute disadvantage

for these countries.

16 These numbershave to be used with more care be-cause there was a growing awareness of how to avoidepidemics of the worst tropical diseases, at least duringshort military campaigns. For example, the campaign inEthiopia at the end of the nineteenthcenturyhad very lowmortalityrates because it was short and well managed (seeFigure 1). Althoughthe mortalityrates from this successfulcampaign certainly underestimate he mortalityrates faced


16/34

VOL. 91 NO. 5 ACEMOGLUETAL: THECOLONIALORIGINSOF DEVELOPMENT 1383earliest available number for each country,reasoning that this is the best estimate of themortality rates that settlers would have faced,at least until the twentieth century.The main gap in the Curtin data is for SouthAmerica since the Spanish and Portuguesemilitaries did not keep good records of mor-tality. Hector Gutierrez (1986) used Vaticanrecords to construct estimates for the mortal-ity rates of bishops in Latin America from1604 to 1876. Because these data overlapwith the Curtin estimates for several coun-tries, we are able to construct a data series forSouth America.17 Curtin(1964) also providesestimates of mortality in naval squadronsfordifferent regions which we can use to gener-ate alternative estimates of mortalityin SouthAmerica. Appendix B in Acemoglu et al.(2000), which is available from the authors,gives a detailed discussion of how these dataare constructed, and Appendix Table A5(available from the authors), shows that thesealternative methods produce remarkablysim-ilar results. Appendix Table A2 lists our mainestimates, and Table Al gives informationabout sources.

IV. Institutions and Performance: IV ResultsA. Determinantsof CurrentInstitutions

Equation (1) describes the relationshipbe-tween current nstitutions and log GDP. In ad-ditionwe have(2) Ri = AR + OR Ci + X>iYR+ VRi,(3) Ci = AC+ ,Bcsi + Xj')/c+ 1Ci,(4) Si = As + (351ogMi + X1,yS+ vSi,

where R is the measureof current nstitutions(protectionagainst expropriationbetween 1985and 1995), C is our measure of early (circa1900) institutions,S is the measureof Europeansettlements n the colony (fraction of the popu-lation withEuropeandescent in 1900), and M ismortalityrates faced by settlers.X is a vectorofcovariates that affect all variables.The simplest identificationstrategy might beto use Si (or Ci) as an instrument for Ri inequation (1), and we report some of these re-gressions in Table8. However,to the extent thatsettlersaremorelikely to migrate o richerareasand early institutions reflect other characteris-tics that are importantfor income today, thisidentificationstrategywould be invalid (i.e., CiandSi could be correlatedwith sk). Instead,weuse the mortalityrates faced by the settlers, logMi, as an instrument or Ri. This identificationstrategy will be valid as long as log Mi isuncorrelatedwith si-that is, if mortality ratesof settlers between the seventeenth and nine-teenth centurieshaveno effect on income todayother than through their influence on institu-tional development.We argued above that thisexclusion restriction s plausible.Figure3 illustratesherelationship etween he(potential) ettlermortality ates and the index ofinstitutions.We use the logarithmof the settlermortality ates,sincethereareno theoretical ea-sons to prefer he level as a determinant f insti-tutions ratherthan the log, and using the logensures hat heextremeAfricanmortalityatesdonot play a disproportionateole. As it happens,there s an almost inearrelationship etweenthelog settlermortalityand our measureof institu-tions. This relationshipshows that ex-colonieswhere Europeansfaced higher mortalityrateshave substantiallyworse institutionsoday.In Table3, we document hat thisrelationshipworksthrough hechannelshypothesizedn Sec-tion I. Inparticular, e presentOLSregressions fequations 2), (3), and (4). In the top panel,weregressthe protectionagainstexpropriationari-ableon the othervariables.Column(1) uses con-straintsfaced by the executive in 1900 as theregressor,andshows a close associationbetweenearly nstitutions nd nstitutionsoday.Forexam-ple, past institutionsaloneexplain20 percentofthe variationn the index of current nstitutions.The second column adds the latitudevariable,

by potential settlers in Ethiopia, we did not exclude thiscountry because excluding it would have helped our hy-pothesis.17 Combining data from a variety of sources will in-troduce measurement error in our estimates of settlermortality. Nevertheless, since we are using settler mor-tality as an instrument, this measurement errordoes notlead to inconsistent estimates of the effect of institutions

on performance.


17/34

1384 THEAMERICANECONOMICREVIEW DECEMBER2001

10 USANZL CANAUS SGP

C)co) IND GMB

0) Kv 8 ' MYS ,fflLANGAB

i?r) Z A ^~~~~~LAM -F,6._ T ~~~~~~~~CMRINGH

NABGA MDG4 SDN MLI4 ~~~~~~~~~~~HTIZAR

2 4 6 8Logof SettlerMortalityFIGURE 3. FIRST-STAGE RELATIONSHIP BETWEEN SETTLER MORTALITY AND EXPROPRIATIONRISK

with little effecton theestimate.Columns 3) and(4) use the democracyindex, and confirmtheresults n columns(1) and (2).Both constraintson theexecutive and democ-racy indices assign low scores to countries thatwere colonies in 1900, and do not use the ear-liest postindependence information for LatinAmerican countries and the Neo-Europes. Incolumns (5) and (6), we adopt an alternativeapproachanduse the constraintson the execu-tive in the first year of independence and alsocontrol separately or time since independence.The results are similar, and indicate that earlyinstitutionstend to persist.Columns(7) and (8) show the associationbe-tweenprotection gainstexpropriation ndEuro-pean settlements.The fraction of Europeans n1900 alone explains approximately 0 percentofthe variation n our institutionsvariabletoday.Columns(9) and (10) show the relationshipbe-tween the protectionagainstexpropriation ari-ableand themortality ates acedby settlers.Thisspecificationwill be the firststage for our maintwo-stage east-squares stimates 2SLS).It showsthatsettlermortalityalone explains27 percentofthe differences n institutionswe observetoday.Panel B of Table 3 provides evidence in

supportof the hypothesis that early institutionswere shaped, at least in part,by settlements,andthat settlementswere affected by mortality.Col-umns (1)-(2) and (5)-(6) relate our measure ofconstrainton the executive and democracy in1900 to the measureof European ettlements n1900 (fraction of the population of Europeandecent). Columns (3)-(4) and (7)-(8) relate thesame variables o settlermortality.These regres-sionsshow thatsettlement atterns xplainaround50 percentof the variation n early institutions.Finally, columns (9) and (10) show the relation-ship between settlementsand mortality ates.B. Institutionsand EconomicPerformanceTwo-stage least-squares estimates of equa-tion (1) are presented in Table 4. Protectionagainst expropriationvariable,Ri, is treated asendogenous, and modeled as

(5) Ri = + log Mi + X'8 + vi,where Mi is the settlermortalityrate in 1,000mean strength.The exclusion restriction s thatthis variable does not appear n (1).


18/34

VOL.91 NO. 5 ACEMOGLUETAL.: THE COLONIALORIGINSOF DEVELOPMENT 1385TABLE 3-DETERMINANTS OF INSTITUTIONS

(1) (2) (3) (4) (5) (6) (7) (8) (9) (10)Panel A DependentVariable Is Average ProtectionAgainst ExpropriationRisk in 1985-1995Constrainton executive in 0.32 0.261900 (0.08) (0.09)Democracyin 1900 0.24 0.21(0.06) (0.07)Constrainton executive in first 0.25 0.22year of independence (0.08) (0.08)European ettlements n 1900 3.20 3.00(0.61) (0.78)Log Europeansettler mortality -0.61 -0.51(0.13) (0.14)Latitude 2.20 1.60 2.70 0.58 2.00(1.40) (1.50) (1.40) (1.51) (1.34)R2 0.2 0.23 0.24 0.25 0.19 0.24 0.3 0.3 0.27 0.3Numberof observations 63 63 62 62 63 63 66 66 64 64

DependentVariable IsEuropeanDependentVariable Is Constraint DependentVariableIs Settlements nPanelB on Executive in 1900 Democracyin 1900 1900European ettlements n 1900 5.50 5.40 8.60 8.10(0.73) (0.93) (0.90) (1.20)Log European ettler mortality -0.82 -0.65 -1.22 -0.88 -0.11 -0.07(0.17) (0.18) (0.24) (0.25) (0.02) (0.02)Latitude 0.33 3.60 1.60 7.60 0.87

(1.80) (1.70) (2.30) (2.40) (0.19)R2 0.46 0.46 0.25 0.29 0.57 0.57 0.28 0.37 0.31 0.47Number of observations 70 70 75 75 67 67 68 68 73 73Notes: All regressions are OLS. Standard rrorsare in parentheses.Regressionswith constrainton executive in firstyear ofindependencealso includeyearssince independenceas a regressor.Average protectionagainst expropriation isk is on a scalefrom 0 to 10, where a higher score means more protection against expropriationof private investmentby government,averagedover 1985 to 1995. Constrainton executive in 1900 is on a scale from 1 to 7, with a higherscore indicatingmoreconstraints.Democracy in 1900 is on a scale from 0 to 10, with a higher score indicatingmore democracy. Europeansettlements is percentof populationthat was Europeanor of Europeandescent in 1900. See AppendixTable Al for moredetailed variabledefinitions and sources.

Panel A of Table 4 reports 2SLS estimatesof the coefficient of interest, a from equation(1) and Panel B gives the corresponding firststages.18 Column (1) displays the strong first-stage relationshipbetween (log) settler mortal-ity and current nstitutions n our base sample,also shown in Table 3. The corresponding2SLS

estimate of the impactof institutionson incomeper capita is 0.94. This estimate is highly sig-nificantwith a standard rrorof 0.16, and in factlarger than the OLS estimates reported inTable 2. This suggests that measurementerrorin the institutions variables that creates attenu-ation bias is likely to be more important hanreverse causality and omitted variables biases.Here we are referring to measurement rrorbroadlyconstrued.In reality the set of institu-tions that matterfor economic performance svery complex, andany single measure s boundto capture only part of the trueinstitutions,

18 We have also run these regressions with standarderrorscorrected orpossible clusteringof the mortalityratesassignedto countries n the samedisease environment.Thisclustering has little effect on the standarderrors, and doesnot change our results.


19/34

1386 THEAMERICANECONOMICREVIEW DECEMBER2001TABLE4-IV REGRESSIONSF LOGGDP PERCAPITA

BaseBase Base sample,Base Base sample sample dependentBase sample Base sample sample sample with with variable sBase Base without without without without continent continent log outputsample sample Neo-Europes Neo-Europes Africa Africa dummies dummies per worker(1) (2) (3) (4) (5) (6) (7) (8) (9)Panel A: Two-Stage Least Squares

Average protection against 0.94 1.00 1.28 1.21 0.58 0.58 0.98 1.10 0.98expropriation isk 1985-1995 (0.16) (0.22) (0.36) (0.35) (0.10) (0.12) (0.30) (0.46) (0.17)Latitude -0.65 0.94 0.04 -1.20(1.34) (1.46) (0.84) (1.8)Asia dummy -0.92 -1.10(0.40) (0.52)Africa dummy -0.46 -0.44(0.36) (0.42)Other ontinentdummy -0.94 -0.99(0.85) (1.0)

Panel B: FirstStage for AverageProtectionAgainst ExpropriationRisk in 1985-1995Log European ettler mortality -0.61 -0.51 -0.39 -0.39 -1.20 -1.10 -0.43 -0.34 -0.63(0.13) (0.14) (0.13) (0.14) (0.22) (0.24) (0.17) (0.18) (0.13)Latitude 2.00 -0.11 0.99 2.00(1.34) (1.50) (1.43) (1.40)Asia dummy 0.33 0.47(0.49) (0.50)Africadummy -0.27 -0.26(0.41) (0.41)Other ontinentdummy 1.24 1.1(0.84) (0.84)R2 0.27 0.30 0.13 0.13 0.47 0.47 0.30 0.33 0.28

Panel C: OrdinaryLeast SquaresAverage protection against 0.52 0.47 0.49 0.47 0.48 0.47 0.42 0.40 0.46expropriation isk 1985-1995 (0.06) (0.06) (0.08) (0.07) (0.07) (0.07) (0.06) (0.06) (0.06)Number of observations 64 64 60 60 37 37 64 64 61Notes:Thedependentvariable n columns(1)-(8) is log GDPper capita n 1995,PPP basis.The dependentvariable n column(9) is log outputper worker, rom Hall andJones (1999). Averageprotectionagainst expropriation isk 1985-1995 is measuredon a scale from0 to 10,wherea higher score means more protection againstrisk of expropriation f investmentby the government,from Political Risk Services. Panel Areports he two-stage least-squares stimates, nstrumentingorprotectionagainstexpropriation iskusing log settlermortality;PanelB reportsthecorrespondingirststage.PanelC reports hecoefficient fromanOLSregressionof thedependentvariableagainstaverageprotectionagainstexpropriation isk. Standard rrorsare n parentheses. n regressionswith continentdummies,thedummyfor America s omitted. See AppendixTable Al for more detailed variabledescriptionsand sources.

creatinga typical measurementerror problem.Moreover, what mattersfor current income ispresumablynot only institutions oday, but alsoinstitutions n the past. Our measure of institu-tions which refers to 1985-1995 will not beperfectly correlatedwith these.19

Does the 2SLS estimate make quantitativesense? Does it implythat nstitutional ifferencescan explain a significant ractionof income dif-

19We can ascertain, o some degree, whether the differ-ence between OLS and 2SLS estimates could be due tomeasurementerror in the institutions variable by makinguse of an alternativemeasure of institutions,for example,the constraintson the executive measure. Using this mea-

sure as an instrument or the protectionagainst expropria-tion index would solve the measurementerror,but not theendogeneity problem.This exercise leads to an estimateofthe effect of protection againstexpropriation qual to 0.87(with standard rror0.16). This suggests that measurementerror n the institutionsvariables(or the signal-to-noiseratio in the institutions variable) is of the right order ofmagnitudeto explain the differencebetween the OLS and2SLS estimates.


20/34

VOL.91 NO. 5 ACEMOGLUET AL.: THECOLONIALORIGINSOF DEVELOPMENT 1387ferencesacrosscountries?Let us once againcom-pare two typical ountrieswith high and lowexpropriationisk,Nigeriaand Chile (these coun-tries are typical or the IV regression n the sensethat they are practicallyon the regression ine).Our 2SLS estimate, 0.94, implies that the 2.24differences in expropriation isk between thesetwo countries hould translate nto 206 log point(approximately-fold)difference. npractice, hepresence of measurement rror complicates hisinterpretation,ecause some of the differencebe-tween Nigeria and Chile's expropriation ndexmay reflect measurementerror.Therefore,the7-folddifference s an upperbound.In any case,the estimates n Table4 imply a substantial, utnot implausiblyarge,effect of institutional iffer-ences on income per capita.Colunm (2) shows that adding latitude doesnot change the relationship;the institutionscoefficient is now 1.00 with a standard rrorof0.22.20 Remarkably,the latitude variablenowhas the wrong sign and is insignificant.Thisresult suggests thatmany previous studiesmayhave found latitude to be a significant determi-nant of economic performance because it iscorrelatedwith institutions(or with the exoge-nous componentof institutionscausedby earlycolonial experience).Columns(3) and (4) document hat our resultsare not driven by the Neo-Europes.When weexclude theUnited States,Canada,Australia, ndNew Zealand, he estimatesremainhighly signif-icant,and n fact increasea little.Forexample, hecoefficient for institutions s now 1.28 (s.e. =0.36) without he latitude ontrol,and 1.21(s.e. =0.35) when we controlfor latitude.Columns(5)and (6) show thatour resultsare also robusttodroppingall the Africancountries rom our sam-ple. The estimates without Africa are somewhatsmaller,but also moreprecise.Forexample,thecoefficientfor institutions s 0.58 (s.e. = 0.1)without he latitudecontrol,and still 0.58 (s.e. =0.12) when we control or latitude.21

In columns(7) and (8), we addcontinentdum-mies to the regressions (for Africa, Asia, andother, with America as the omittedgroup).Theadditionof these dummies does not changetheestimatedeffect of institutions, nd the dummiesare jointly insignificantat the 5-percentlevel,though hedummy or Asia is significantly iffer-ent from hatof America.The fact that he Africandummy is insignificant uggests thatthe reasonwhy African countriesare poorer is not due tocultural or geographic factors, but mostly ac-counted or by the existence of worse institutionsin Africa. Finally, in column (9) we repeat ourbasic regressionusing log of outputper workerascalculatedby Hall and Jones(1999).The result sveiy close to our baselineresult.The 2SLS coef-ficient s 0.98 insteadof 0.94 as in column(1).22This shows thatwhetherwe use incomeper capitaoroutputper workerhaslittleeffecton ourresults.Overall, he results n Table 4 show a largeeffectof institutionson economic performance.n therest of thepaper,we investigate herobustness fthese results.

20 In 2SLS estimation, all covariatesthat are included inthe second stage, such as latitude, are also includedin thefirst stage. When these first-stage effects are of no majorsignificance for our argument,we do not report hem in thetables to save space.21 We should note at this point that f we limit the sample

to African countries only, the first-stagerelationshipusing

the protection against expropriationvariablebecomes con-siderably weaker, and the 2SLS effect of institutions s nolonger significant.The 2SLS effect of institutionscontinueto be significantwhen we use some (but not all) measuresofinstitutions. Therefore, we conclude that the relationshipbetween settler mortality and institutions s weaker withinAfrica.22 The results with other covariatesare also very similar.We repeated the same regressions using a variety of alter-native measures of institutions, ncluding constraintson theexecutive from the Polity III data set, an index of law andorder traditionfrom Political Risk Services, a measure ofpropertyrights from the Heritage Foundation,a measureofrule of law from the Fraser Institute, and the efficiencyof the judiciary from Business International.The resultsand the magnitudesare very similar to those reportedinTable 4. We also obtainedvery similarresultswith the 1970values for the constraintson the executive and income percapita in 1970, which show that the relationshipbetweeninstitutional measures and income per capita holds acrosstimeperiods. These results are reported n the Appendixofthe working paper version, and are also available from theauthors.23 In the working paperversion, we also investigated herobustnessof ourresultsin different subsampleswith vary-ing degrees of data quality and different methodsof con-structing the mortalityestimates. The results change verylittle, for example, when we use data only from Curtin(1989), Death by Migration,when we do not assign mor-

tality rates from neighboring disease environments, when


21/34

1388 THEAMERICANECONOMICREVIEW DECEMBER2001V. Robustness

A. AdditionalControlsThe validity of our 2SLS results in Table 4dependson the assumptionhatsettlermortalitynthe past has no directeffect on current conomicperformance.Although his presumption ppearsreasonable at least to us), here we substantiatetfurtherby directly controlling or many of thevariables hat couldplausiblybe correlatedwithbothsettlermortality ndeconomicoutcomes,andchecking whether he additionof these variablesaffectsourestimates.24Overall,we find thatourresultschangeremarkablyittle with the inclusion

of thesevariables, ndmanyvariables mphasizedin previous work become insignificantonce theeffect of institutions s controlled or.La Porta et al. (1999) argue for the impor-tance of colonial origin (identity of the maincolonizing country)as a determinant f currentinstitutions.The identity of the colonial powercould also matter because it might have aneffect through culture, as argued by DavidS. Landes (1998). In columns (1) and (2) ofTable5, we add dummiesfor British and Frenchcolonies (colonies of other nations are the omit-ted group).This has little affect on our results.Moreover, the French dummy in the firststage is estimated to be zero, while the Britishdummy is positive, and marginally significant.Therefore, as suggested by La Porta et al.(1998), British colonies appearto have betterinstitutions,but this effect is much smaller andweaker than in a specification that does notcontrol for the effect of settler mortality oninstitutional development.25Therefore, it ap-

pearsthatBritish colonies are found to performsubstantiallybetter n other studiesin large partbecause Britain colonized places where settle-ments were possible, and this made British col-onies inherit better institutions. To furtherinvestigate this issue, columns (3) and (4) esti-mate our basic regressionfor British coloniesonly. They show that both the relationshipbe-tween settler mortalityandinstitutionsand thatbetween institutions and income in this sampleof 25 British colonies are very similar to thosein our base sample. For example, the 2SLSestimate of the effect of institutions on incomeis now 1.07 (s.e. = 0.24) withoutcontrolling orlatitude and 1.00 (s.e. = 0.22) with latitude.These results suggest that the identity of thecolonizer is not an important determinantofcolonization patterns and subsequent institu-tional development.von Hayek (1960) and La Portaet al. (1999)also emphasize he importance f legal origin.Incolumns 5) and(6), we control orlegal origin. nour sample,all countrieshave eitherFrenchorBritish egal origins,so we simplyadd a dummyfor French egal origin (many countries hat arenot French olonies nonethelesshave Frenchegalorigin).Our estimateof the effect of institutionson incomeper capita s unaffected.26An argument dating back to Max Weberviews religionas a key determinant f economicperformance.To control for this, in columns(7)and (8), we add the fraction of the populationsthat are Catholic, Muslim, and of other reli-gions, with Protestantsas the omittedgroup.Inthe table we reportthe joint significance level(p-value) of the correspondingF-statistic forthese dummiesas well as the 2SLS estimate of

the use data or LatinAmericafrom naval stations nsteadofbishops, and when we do not use data from small Africansamples. These results are availablein Appendix Table A5availablefrom the authors,or in Acemoglu et al. (2000).24 Joseph N. Altonji et al. (2000) develop aneconometricmethodology to assess the importanceof omitted variablebias. The basic idea is thatif the estimate of the coefficientof interest does not change as additional covariates areincluded in the regression,it is less likely to change if wewere able to add someof themissing omittedvariables.Ourmethodologyhere is an informal version of this approach.25 Moreover, he Britishcolonial dummy is negative andsignificantn thesecondstage.The net effect of beinga British

colonyon income per capita s in fact negative.More specif-

ically, British olonieshave, on average,an ndex of institutionthatis 0.63 points lower. Given the 2SLS estimate of 1.10,this translatesinto 69 log points higher income per capitafor British colonies (1.10 X 63 69). The second-stageeffect of being a British colony is -78 log points, im-plying -9 log point (approximately 10 percent) negativenet effect of being a British colony. A possible explana-tion for this pattern is that (Anglo-Saxon?) researchersare overestimating how bad French institutions are, andthe second-stage regression is correcting for this.26 The first stage shows that French egal origin is asso-ciated with worse institutions,but similarly, he net effect ofhaving French legal origin is actually positive: -67 X

1.1 + 89 = 15 log points (approximately15 percent).


22/34

VOL. 91 NO. 5 ACEMOGLUETAL.: THE COLONIALORIGINSOF DEVELOPMENT 1389TABLE 5-IV REGRESSIONS OF LOGGDP PER CAPITA WITH ADDITIONAL CONTROLS

British BritishBase Base colonies colonies Base Base Base Base Basesample sample only only sample sample sample sample sample(1) (2) (3) (4) (5) (6) (7) (8) (9)

Panel A: Two-Stage Least SquaresAverage protectionagainst 1.10 1.16 1.07 1.00 1.10 1.20 0.92 1.00 1.10expropriation isk, 1985-1995 (0.22) (0.34) (0.24) (0.22) (0.19) (0.29) (0.15) (0.25) (0.29)Latitude -0.75 -1.10 -0.94 -1.70(1.70) (1.56) (1.50) (1.6)British colonial dummy -0.78 -0.80(0.35) (0.39)Frenchcolonial dummy -0.12 -0.06 0.02(0.35) (0.42) (0.69)French egal origin dummy 0.89 0.96 0.51

(0.32) (0.39) (0.69)p-value for religion variables [0.001] [0.004] [0.42]Panel B: First Stage for Average Protection Against ExpropriationRisk in 1985-1995

Log Europeansettler mortality -0.53 -0.43 -0.59 -0.51 -0.54 -0.44 -0.58 -0.44 -0.48(0.14) (0.16) (0.19) (0.14) (0.13) (0.14) (0.13) (0.15) (0.18)Latitude 1.97 2.10 2.50 2.30(1.40) (1.30) (1.50) (1.60)British colonial dummy 0.63 0.55(0.37) (0.37)French colonial dummy 0.05 -0.12 -0.25(0.43) (0.44) (0.89)French egal origin -0.67 -0.7 -0.05(0.33) (0.32) (0.91)R2 0.31 0.33 0.30 0.30 0.32 0.35 0.32 0.35 0.45Panel C: OrdinaryLeast SquaresAverage protection against 0.53 0.47 0.61 0.47 0.56 0.56 0.53 0.47 0.47expropriation isk, 1985-1995 (0.19) (0.07) (0.09) (0.06) (0.06) (0.06) (0.06) (0.06) (0.06)Number of observations 64 64 25 25 64 64 64 64 64

Notes: PanelA reports he two-stage east-squaresstimateswithlog GDPper capita PPP basis)in 1995 as dependent ariable,andPanelB reportshecorrespondingirststage.The base case in columns(1) and(2) is all colonies thatwere neitherFrenchnorBritish.Thereligionvariablesare included n the firststageof columns(7) and(8) but notreported ere(to save space).PanelCreports he OLS coefficient rom regressing og GDP per capitaon averageprotectionagainstexpropriation isk,with the othercontrolvariablesndicated n thatcolumn(fullresultsnotreportedo save space).Standard rrorsare n parentheses ndp-valuesforjoint significance ests are in brackets.The religionvariablesarepercentageof population hat areCatholics,Muslims,andother eligions;Protestants the base case. Oursample s all eitherFrenchor British egal origin (asdefinedby La Portaet al.,1999).

the effect of institutions.27Finally, column (9)adds all the variables in this table simulta-neously. Again, these controls have very littleeffect on our main estimate.Another concern is that settler mortality is27 The religion dummies are significant n the first stage,but once again they are estimated to have offsetting effects

in the second stage, implying little net effect of religiononincome.

correlated with climate and other geographiccharacteristics.Our instrument may thereforebe picking up the direct effect of these vari-ables. We investigate this issue in Table 6. Incolumns (1) and (2), we add a set of temper-ature and humidity variables (all data fromPhilip M. Parker, 1997). In the table wereport joint significance levels for these vari-ables. Again, they have little effect on ourestimates.


23/34

1390 THEAMERICANECONOMICREVIEW DECEMBER2001TABLE6-ROBUSTNESS CHECKS ORIV REGRESSIONSF LOGGDP PERCAPITA

Base Base Base Base Base Base Base Base Basesample sample sample sample sample sample sample sample sample(1) (2) (3) (4) (5) (6) (7) (8) (9)

Panel A: Two-Stage Least SquaresAverage protection against 0.84 0.83 0.96 0.99 1.10 1.30 0.74 0.79 0.71expropriation isk, 1985-1995 (0.19) (0.21) (0.28) (0.30) (0.33) (0.51) (0.13) (0.17) (0.20)Latitude 0.07 -0.67 -1.30 -0.89 -2.5(1.60) (1.30) (2.30) (1.00) (1.60)p-value for temperature ariables [0.96] [0.97] [0.77]p-value for humidity variables [0.54] [0.54] [0.62]Percent of Europeandescent in 1975 -0.08 0.03 0.3(0.82) (0.84) (0.7)p-value for soil quality [0.79] [0.85] [0.46]p-value for naturalresources [0.82] [0.87] [0.82]Dummy for being landlocked 0.64 0.79 0.75

(0.63) (0.83) (0.47)Ethnolinguistic ragmentation -1.00 -1.10 -1.60(0.32) (0.34) (0.47)Panel B: First Stage for Average ProtectionAgainst ExpropriationRisk in 1985-1995

Log European ettler mortality -0.64 -0.59 -0.41 -0.4 -0.44 -0.34 -0.64 -0.56 -0.59(0.17) (0.17) (0.14) (0.15) (0.16) (0.17) (0.15) (0.15) (0.21)Latitude 2.70 0.48 2.20 2.30 4.20(2.00) (1.50) (1.50) (1.40) (2.60)R2 0.39 0.41 0.34 0.34 0.41 0.43 0.27 0.30 0.59Panel C: OrdinaryLeast Squares

Average protection against 0.41 0.38 0.39 0.38 0.46 0.42 0.46 0.45 0.38expropriation isk, 1985-1995 (0.06) (0.06) (0.06) (0.06) (0.07) (0.07) (0.05) (0.06) (0.06)Notes: Panel A reports he two-stage least-squares stimates withlog GDPper capita (PPP basis) in 1995, and PanelB reportsthe corresponding irst stages. Panel C reports he OLS coefficient from regressing og GDP per capita on average protectionagainst expropriation isk, with the other controlvariables ndicated n that column (full results not reported o save space).Standarderrors are in parentheses and p-values for joint significance tests are in brackets. All regressions have 64observations,except those includingnaturalresources,which have 63 observations.The temperature ndhumidityvariablesare:average, minimum,and maximummonthly high temperatures, nd minimum and maximummonthlylow temperatures,and morning minimum and maximum humidity, and afternoon minimum and maximum humidity (from Parker, 1997).Measuresof natural esourcesare:percentof worldgold reservestoday, percentof world ron reservestoday, percentof worldzinc reservestoday, numberof mineralspresent n country,andoil resources(thousandsof barrelsper capita).Measures ofsoil quality/climatearesteppe (low latitude),desert(low latitude),steppe (middle latitude),desert(middle latitude),dry steppewasteland,desertdry winter,and highland.See AppendixTable Al for more detailedvariabledefinitions and sources.

A related concern is that in colonies whereEuropeans settled, the currentpopulation con-sists of a higher fraction of Europeans. Onemight be worried that we are capturing thedirect effect of having more Europeans (whoperhaps broughta Europeanculture or spe-cial relations with Europe). To control forthis, we add the fraction of the population ofEuropean descent in columns (3) and (4) ofTable 6. This variable is insignificant, whilethe effect of institutions remains highly sig-

nificant, with a coefficient of 0.96 (s.e. =0.28). In columns (5) and (6), we control formeasures of natural resources, soil quality (inpractice soil types), andfor whether the coun-try is landlocked. All these controls are insig-nificant, and have little effect on our 2SLSestimate of the effect of institutions on in-come per capita.In columns (7) and (8), we include ethno-linguistic fragmentation as another controland treat it as exogenous. Now the coefficient


24/34

VOL.91 NO. S ACEMOGLUET AL: THE COLONIALORIGINSOF DEVELOPMENT 1391of protection against expropriation is 0.74(s.e. = 0.13), which is only slightly smallerthan our baseline estimate. In AppendixA, weshow that the inclusion of an endogenousvariable positively correlatedwith income orinstitutions will bias the coefficient on insti-tutions downwards. Since ethnolinguisticfragmentation s likely to be endogenous withrespect to development (i.e., ethnolinguisticfragmentation ends to disappear after the for-mation of centralized markets; see Weber[1976] or Andersen [1983]) and is correlatedwith settler mortality, the estimate of 0.74likely understatesthe effect of institutions onincome. In column (9) of Table 6, we includeall these variables together. Despite the largenumber of controls, protection against expro-priation on income per capita is still highlysignificant, with a somewhat smaller coeffi-cient of 0.71 (s.e. = 0.20), which is againlikely to understate the effect of institutionson income because ethnolinguistic fragmen-tation is treated as exogenous.Finally, in Table 7, we investigate whetherour instrumentcould be capturingthe generaleffect of disease on development. Sachs and aseries of coauthorshave argued for the impor-tance of malariaand other diseases in explain-ing Africanpoverty (see, for example, Bloomand Sachs, 1998; Gallup and Sachs, 1998;Gallup et al., 1998). Since malariawas one ofthe main causes of settler mortality,our esti-mate may be capturing he direct effect of ma-laria on economic performance. We areskepticalof this argumentsince malariapreva-lence is highly endogenous; it is the poorercountrieswithworse institutions hathave beenunable to eradicatemalaria.28While Sachs andcoauthors argue that malaria reduces outputthroughpoorhealth, high mortality,and absen-teeism, most people who live in high malaria

areas have developed some immunity to thedisease (see the discussion in Section III, sub-section A). Malaria shouldthereforehave littledirect effect on economic performance though,obviously, it will have very high social costs).In contrast,for Europeans,or anyone else whohas not been exposed to malariaas a youngchild, malaria is usually fatal, making malariaprevalencea key determinantof Europeanset-tlements and institutionaldevelopment.In any case, controllingfor malariadoes notchange our results. We do this in columns (1)and (2) by controlling for the fraction of thepopulationwho live in an area where falcipo-rum malaria s endemicin 1994 (as constructedand used by Gallup et al., 1998). Since malariaprevalence in 1994 is highly endogenous, theargument n Appendix A implies that control-ling for it directly will underestimate he effectof institutions on performance.In fact, the co-efficient on protectionagainst expropriation snow estimated to be somewhat smaller, 0.69instead of 0.94 as in Table 4. Nevertheless,theeffect remainshighly significantwith a stan-dard error of 0.25, while malaria itself isinsignificant.In a comment on the workingpaper versionof our study, John W. McArthurand Sachs(2001) discuss the role of geography and insti-tutions in determiningeconomic performance.They accept our case for the importanceofinstitutions,but arguethatmore general speci-fications show that the disease environmentandhealth characteristicsof countries(their geog-raphy )matter for economic performance.Inparticular, hey extend our work by controllingfor life expectancy and infant mortality, andthey also instrument or these health variablesusing geographicvariablessuch as latitude andmean temperature.Table 7 also expands uponthe specifications hat McArthurandSachs sug-gest. Columns (3)-(6) include life expectancyand infantmortalityas exogenous controls. Theestimates show a significant effect of institu-tions on income, similar to, but smaller than,our baseline estimates. Infantmortalityis alsomarginally significant. Since health is highlyendogenous, the coefficient on these variableswill be biasedup, while the coefficient of insti-tutions will be biased down (see AppendixA).These estimates are therefore consistent with

28 For example, heUnited States eliminatedmalariaromthe PanamaCanal Zone, and Australiaeliminated t fromQueensland see Crosby, 1986 pp. 141-42). Even in Africa,there have been very successfulcampaignsagainstmalaria,including hose nAlgeriaand thatconductedby the Rio-TintoZinc mining company n Zambia (thenNorthernRhodesia).TheWHO'sRoll BackMalariaprogram ontainsa number feffective recommendations or controlling malaria that arerelatively straightforward o implement if families haveenoughmoney (e.g., insecticide-treateded nets).

8/13/2019 3. 4 Ace

3. 4 acemoglu the colonial

Documents