Munich Personal RePEc Archive Does land abundance explain African institutions? Fenske, James University of Oxford November 2011 Online at https://mpra.ub.uni-muenchen.de/34862/ MPRA Paper No. 34862, posted 20 Nov 2011 12:35 UTC DOES LAND ABUNDANCE EXPLAIN AFRICAN INSTITUTIONS? JAMESFENSKE† ABSTRACT. IshowthatabundantlandandscarcelaborshapedAfricaninstitutionsbefore colonialrule.Itestamodelinwhichexogenouslandqualityandendogenouspopulation determinetheexistenceoflandrightsandslavery. Iusecross-sectionaldataonaglobal sampleofsocietiestodemonstratethat,asinthemodel,landrightsoccurredwhereland qualitywashighandwherepopulationdensitywasgreatest. Slaveryexistedwhereland wasgoodandpopulationdensitywasintermediate. Themodelcanexplaininstitutional differencesacrossregions,butnotwithinregions.Ipresentsuggestiveevidencethatthis isduetoinstitutionalspillovers. 1. INTRODUCTION The“landabundance”viewofAfricanhistoryisaninfluentialexplanationoftheeco- nomic institutions that existed on the continent before colonial rule (Austin, 2008a; Hopkins,1973;Iliffe,1995). Thistheoryholdsthat,sinceunclearedlandwasfreelyavail- able, land had no price, and rights to land were ill-defined. Because cultivators would not become free workers, coerced and household labor substituted for wage employ- ment. Lagerlo¨f’s(2010)modelof“slaveryandotherpropertyrights”mirrorstheseargu- mentsinaformalmodel. Inthispaper,Iusecross-sectionaldataonasampleofglobal †DEPARTMENTOFECONOMICS,UNIVERSITYOFOXFORD E-mail address:[email protected]. Date:November19,2011. I would like to thank my advisors Timothy Guinnane, Benjamin Polak, and Christopher Udry for their guidance. I would also like to thank Tayo Adesina, Achyuta Adhvaryu, Gareth Austin, Reena Badiani, BenjaminChabot,AdriandelaGarza,RahulDeb,ShataksheeDhongde,OdedGalor,Nils-PetterLagerlo¨f, Naomi Lamoreaux, Giuseppe Moscarini, Nathan Nunn, Sheilagh Ogilvie, Florian Ploeckl, Mark Rosen- zweig,MirSalim,VeronicaSantarosa,EdVytlacil,WarrenWhatley,AdemolaYakubu,andtheparticipants oftheEconomicHistoryAssociationAnnualMeeting,theCanadianNetworkforEconomicHistory,the NEUDC,theLSEAfricanEconomicHistorySeminar,andseminarsattheUniversityofWarwick,Prince- tonUniversity,WashingtonUniversityinSt. Louis,NorthwesternUniversity,theUniversityofRochester, theUniversityofToronto,BostonUniversityandtheUniversityofOxfordfortheircommentsandadvice. ExtrathanksareduetoChristianDippelandNathanNunnfortheirgeneroussharingofmaps. 1 2 JAMESFENSKE societiestotestthisview. Ishowthatlandrightsandslaveryexistedinthoseregionspre- dictedbythemodel,butthatinstitutionalspilloverspreventthemodelfrompredicting differenceswithinbroadgeographicregions.1 Thepre-colonialinstitutionsexplainedbythe“landabundance”viewconstrainedthe actionsofcolonialpowers(e.g. Austin(2008b)). Asaresult,pre-colonialinstitutionsand the forces that shaped them affect current performance in Africa (e.g. Gennaioli and Rainer(2007)orTertilt(2005)). Itiswellestablishedineconomicsthatinstitutionsmat- ter(Acemogluetal.,2001;Dell,2011). Landrightsandslavery,inparticular,continueto affect outcomes in Africa and in the rest of the world. Land tenure shapes investment incentives (Goldstein and Udry, 2008), labor-supply (Field, 2007), and violence (Andre andPlatteau,1998). Nunn(2008a)showsthatthoseAfricancountriesthatexportedthe most slaves are comparatively poor today. The pre-colonial prevalence of indigenous slaveryisnegativelycorrelatedwithcurrentincomewithinAfrica(Bezemeretal.,2009). WithintheAmericas, legaciesofslaveryexplaindifferencesinincomeacrosscountries and U.S. counties (Engerman and Sokoloff, 1997; Nunn, 2008b), as well as long term racial gaps in education and other measures (Miller, 2008; Sacerdote, 2005). Explain- ing pre-colonial land tenure and slavery is, then, important in understanding African poverty. The “land abundance” view of African history argues that the continent’s geography has given it an abundance of land relative to labor, which explains the general features ofitsdevelopment. Inothercontexts,geographicfeatures,suchascontinentalorienta- tion,ruggedness,settlermortality,suitabilityforspecificcrops,andotherbiogeographic endowments predict contemporary institutional differences across countries (Easterly and Levine, 2003; Engerman and Sokoloff, 1997; Nunn and Puga, 2011). The model I test, from Lagerlo¨f (2010), similarly allows geography to shape institutions. There are twoimportantvariablesdetermininglandrightsandslavery. Thefirstisexogenousland quality. Thisincreasesthereturnstolandownership,compensatesfortheinefficiencies ofslavery,andsustainsgreaterpopulationsintheMalthusiansteadystate. Thesecond 1In earlier versions of this paper, I extended the model to include raiding of neighboring societies for slavesandanexplanationforpolygyny.Theseextensionshavebeenremoved.Earlier,IusedonlyAfrican data.NowIusedatafromtheentireworld,andsotheresultsdiffersubstantiallyfromthosepresentedin earlierversions. LANDABUNDANCE 3 is population, which responds to both the geographic and institutional environments. It shapes the relative values of land and labor and the relative costs of free and forced workers. Itestthisthesis. Iusedataonacross-sectionofglobalsocietiesfromMurdock’s(1967) Ethnographic Atlas to support a model of land rights and slavery in which the land- laborratiodeterminestheinstitutionsthatexist. Ifindthatthemodelcorrectlypredicts that land rights and slavery were found in those societies that occupied the best land, andthatgreaterpopulationdensitieswerecorrelatedwithrightsoverland. Slaverywas presentwhenpopulationdensitieswereintermediate,asinthemodel. Whilethemodel predictsdifferencesacrossregions,thereareformsofslaverythatitcannotpredict,and itisnotcapableofpredictingdifferenceswithinregions. In Section 2, I outline the literature in African history on how land abundance has shaped economic institutions. Here, I present the basic features of the model and its testable implications. In Section 3, I describe the data used and lay out the economet- ric specifications. In Section 4, I report the results of these tests. In Section 5, I show that these results are robust to different measures of the institutional outcomes, alter- native proxies for land quality and historical population density, and the possible en- dogeneity of land quality. I also argue that this theory of land rights and slavery better explainsthedatathansomeprominentalternatives,includingecologicalriskasacause of common property, and suitability for specific crops as a predictor of slavery. While themodelisgenerallyrobusttoremovinginfluentialobservations,thereareexceptions thatpointtowardsthemechanismsbywhichinstitutionaltransitionsoccur. InSection 6,Idemonstratethatthemodelisunabletopredictdifferenceswithinregions,andthat thereissubstantialspatialcorrelationininstitutionaloutcomes. Thisisconsistentwith whatanthropologistscall“Galton’sproblem”–thelackofindependenceinculturalob- servations. InSection7Iconclude. 2. THE LAND ABUNDANCE VIEW OF AFRICAN HISTORY 2.1. The literature. A first-order task in African history is explaining the continent’s long-run differences in economic organization from the rest of the world. The start- ing point of the land abundance view is the difference in settlement patterns between 4 JAMESFENSKE Africa and the rest of the world on the eve of colonial rule. Herbst (2000, p. 16) es- timates the population density of Sub-Saharan Africa in 1900 at 4.4 persons per Sq. Km, contrasted with 38.2 for South Asia, 45.6 for China, and 62.9 for Europe.2 Expla- nations of low African population densities stress geographic factors, the disease en- vironment, and historical factors such as the slave trades (Mahadi and Inikori, 1987, p. 63-64). This sparse settlement, Hopkins (1973, p. 23-27) argues, shaped institutions, because Africans “measured wealth and power in men rather than in acres.”3 Here, I focusontheimplicationsforlandrightsandslavery. Before the Atlantic slave trade, African settlement was characterized by small clear- ingssurroundedbyvastwastelandsintheEquatorialregion,circlesofincreasinglywild vegetation in the West African forest, and clusters with oscillating frontiers in the West African Savanna (Iliffe, 1995, p. 36,64-67). Austin (2009, p. 33) argues that, as a con- sequence, land was “easily and cheaply accessible in institutional terms”; pre-colonial authoritieswereeagertoattract“morepeoplewithwhomtosubduenatureand,ifnec- essary, their neighbors,” so that strangers could generally acquire land indefinitely for token payments. These payments were made solely to acknowledge the sovereignty of thelocalauthorities. Citizensweregivenlandvirtuallyfreely. Austin(2008a,p.591-594) notes that ‘islands’ of intensive agriculture have existed in Africa where insecurity has createdartificiallandscarcityandinspecificlocationsofexceptionalvalue. Thesehad minerals,trees,marketaccess,orsuitabilityforparticularcrops. Against these views, Spear (1997, p. 154-157) argues that population density cannot explain individual cases. While on Mount Meru both the Arusha and the Meru inten- sified their agriculture as population rose, the less densely settled Meru did so more readily. Berry(1988),similarly,hasnotedthatinheritancerules,tenancycontracts,and laborarrangementsoftenpreventtreecropsfromleadingtoindividualizedlandtenure in West Africa. Thornton (1992, p. 75-76) suggests that ownership of land results from legalclaims,notpopulationpressure. 2HisestimateforNorthAfricais9.4personsperSq.Km. 3Austin(2008a,p.589)arguesthatHopkinswasthefirsttomakethisanalysissystematic;earlierwriters onAfricadidaccountfortheexistenceofslavery, forexample, bynotingAfrica’slandabundance–see Dowd(1917). LANDABUNDANCE 5 For Austin (2008a, p. 606-610), scarcity of labor explains African use of forced labor. He builds on the arguments of Nieboer (1900) and Domar (1970), who argue that co- ercion is cheaper than paying a wage when labor is scarce and wages high. With some notable exceptions (Rodney, 1966), slavery was prevalent in much of Africa even prior to the Atlantic slave trade (Fage, 1969). Watson (1980, p. 10) suggests that the ability of slavesandtheirdescendantstoassimilateintotheirowners’lineageswasa“logicalex- tensionoftheinstitutionalizedneedformorepeople.” Landabundancehasbeenused to explain differences across societies. Northrup (1979) contrasts the densely-settled Igbo of the palm belt with the relatively sparsely populated northeastern Igbo during thepalmoiltrade. Slaverydidnotexpandinthepalmbelt,whilethenortheasternIgbo usedslavestocolonizenewland. The use of underpopulation to explain African slavery is controversial. Writers such asKopytoffandMiers(1977, p.68-69), Lovejoy(1978, p.349), orMiersandKlein(1998, p. 4-5) have stressed that they were employed in non-economic uses, distributed by non-market means, and that colonial rulers turned a blind eye to slavery for political reasons. Kopytoff(1987,p.46)andGoody(1980,p.26-31)suggestthatdependentsmust be “seduced” rather than coerced, so slavery can only exist in complex societies and stateswith“well-developedsystemsofcompulsion.” I clarify this literature and test its claims. With the Lagerlo¨f (2009) model as a rea- sonable formalization of the “land abundance” view, it becomes clear that some of the critiques of the land abundance view are not in fact inconsistent with it. While high wagesresultingfrompopulationdensityexplainthepreferenceforslaveryoverfreela- bor under certain conditions in the model, there are also under conditions in which populationistoosparseforslaverytobeworthwhile,correspondingwiththelesscom- plex societies in Africa that have poorly developed systems of compulsion. More im- portantly, I test the land abundance view in a global sample of societies, defending it against many of these critiques. I show that the institutional effects of population and agriculturalproductivityfollowregularpatternseveniftheycannotexplaineverycase. I show that the presence of slavery is systematically related to the economic value of slavesandtopopulation. 6 JAMESFENSKE 2.2. Model. In this paper, I test the model of “slavery and other property rights” from Lagerlo¨f (2009). This is for two reasons. First, his model echoes the arguments made by historians of Africa, making explicit the testable implications of their views. Greater population lowers average product, which is shared equally in an egalitarian regime. This creates incentives to create rights over land. Similarly, the relative costs of land rightsandslaveryaredeterminedbythecompetitivewage, whichisitselfafunctionof populationsize. Ifpopulationpressureincreaseslaborsupplyanddepressesthewage, freelaborbecomesprofitablerelativetokeepingslaves. Second, his model extends the “land abundance” literature. If population is suffi- ciently low, slavery will not exist, since population pressure has not adequately de- pressed the returns to an egalitarian sharing of output while the opportunity costs of wasting labor on coercion remain high. This reconciles the land abundance view with the critiques of Kopytoff and Goody. In addition, the quality of land determines both the relative profitability of institutional regimes for a given population and the level of population that can be supported. This variable has been generally neglected by the Africanistliterature. Lagerlo¨f(2009)makestheconceptof“landabundance”morepre- cise; it is the availability of cultivable land relative to both population and productivity thatmatters. Here,Ibrieflysketchthebasicelementsofthemodelandstateitstestable implications. The model takes a society in period t with a population P of non-elite agents and a t comparatively small elite that does not work. The elite chooses institutions. Output Y t dependsonlandM,land-augmentingproductivityA˜ ,andthelaborusedL : t t (1) Y = (MA˜ )αL1−α ≡ AαL1−α. t t t t t At the beginning of each period, the elite chooses between three regimes based on whichoneyieldsthemthegreatestprofitsπi,whereidenotesoneofthreeinstitutional t regimes. Thefirstisegalitarianism. Underthisarrangement,therearenolandrightsor slavery. Theeliteandthenon-eliteeachreceiveaverageproduct,andso: LANDABUNDANCE 7 A α (2) πE = t . t (cid:16)P (cid:17) t Thesecondpossibleoutcomeisslavery. Here,theeliteenclosetheentireland,creat- ingrightsoverit. TheyenslaveS slavesfromthepopulation,payingthemonlysubsis- t tenceincomec¯. Eachslaverequiresγ guards,whoarealsopaidc¯,andsotheelitepayoff is: (3) πS = max {AαS1−α −(1+γ)c¯S }. t t t t St≤Pt/(1+γ) Thethirdpossibleoutcomeisfreelabor. Again,theeliteenclosetheentireland. Now, however,theyhiremembersofthepopulationatacompetitivewagew ,whichdepends t onP . Theelite’spayoffis: t (4) πF = max{AαL1−α −w L }. t t t t t Lt≥0 Lagerlo¨f(2009)showsthatthestatespaceinA andP canbedividedintothreesets: t t SE, in which the elite prefer egalitarianism; SS, in which they prefer slavery, and; SS, in which they prefer free labor. The boundaries of these regions are defined by three functionsofP : Ψ(P ),Ω(P ),andΦ(P ).4 ThesearedepictedinFigure1. t t t t The slavery region, SS, is where A ≥ max{Ψ(P ),Ω(P )}, and P > (1 + γ)1−α. A ≥ t t t t t Ψ(P )impliesthatpopulationgivenA isstillsufficientlylowthatw ishighrelativetothe t t t costofkeeping slaves. A ≥ Ω(P ) impliesthatpopulationis sufficientlydensethatthe t t average product under egalitarianism has fallen, while high productivity also ensures theeliteiswillingtowastesomelaboronguardingslavesinordertotakeagreatershare 4 c¯(1+γ)1−α α1 Ψ(P )= P t (cid:16)1−α(1+γ)1−α(cid:17) t Ω(P )= c¯(1+γ)1−αPt1+α α1 t (cid:16) P −(1+γ)1−α (cid:17) t Φ(P )= 1 1−1α c¯ α1P−1−αα t (cid:16)1−α(cid:17) (cid:16)1−α(cid:17) t 8 JAMESFENSKE FIGURE 1. Institutionalregionsandd𝒮ynamics ( ) 𝐴𝑡 Ω 𝑃𝑡 LS( ) S ( ) 𝑃𝑡 Ψ 𝑃𝑡 LE/F( ) 𝒮 𝑃𝑡 LA( ) E 𝑃𝑡 F 𝒮 𝒮 𝐴 ( ) Φ 𝑃𝑡 1/ ∝ 𝑡 𝑃 ofoutputforthemselves. Theopportunitycostoftheseguardsisparticularlyhighwhen population is very low, which explains both the slope of Ω(P ) and the condition that t P > (1+γ)1−α. t SF is the free labor region, in which Φ(P ) ≤ A ≤ Ψ(P ) and P > 1/α. P > 1/α t t t t t ensures that population is great enough that the average product has fallen, making enclosureworthwhile. A ≤ Ψ(P )occurswhenpopulationgrowthpushesdownwages t t sufficiently relative to the costs of keeping slaves. The condition that Φ(P ) ≤ A is of t t less interest, driven by an assumption that the wage is bounded below by c¯. SE occurs intheremainderofthestatespace,whereaverageproductandthecounterfactualwage arebothrelativelyhigh. The dynamics of the model are Malthusian and Boserupian. They are Malthusian in thatfertilityisincreasinginincome. Twoupward-slopingzeropopulationgrowthlines LANDABUNDANCE 9 exist – one under slavery and one under both egalitarianism and free labor. To the left of these, income is high and population is growing. To the right, income is low and populationisfalling. TheseareshowninFigure1asLE/F(P )andLS(P ).5 t t The dynamics are Boserupian in that agricultural technology in period t + 1 has an intercept of A¯ and depends positively on both A and P . Lagerlo¨f (2009) takes A¯ as t t the “minimum level of agricultural technology,” and I interpret it as exogenous land quality. Theresultisanupward-slopingzero-technological-growthlineLA(P ).6 Above t this,productivitydegrades,whilebelowthisitimproves. ThisisalsoshowninFigure1. A steady state exists where either LE/F(P ) or LS(P ) intersects LA(P ). Figure 1 depicts t t t asteadystateinthefreelaborregion. 2.3. Tests. Whatarethetestableimplicationsofthismodeland,byextension,theland abundance view? First, land quality A¯ should positively predict the existence of land rights and slavery. Land rights do not exist under egalitarianism, and if A¯ is too low, it isimpossibletosupportasteadystateundereitherregime. Similarly,A¯mustbehighin order for a steady state to exist with slavery. However, since larger values of A¯can sup- portsteadystatesinboththeslaveryandfreelaborregions,therelationshipbetweenA¯ and slavery is expected to be weaker than for land rights. Second, population density, which I take as corresponding to P in the model, will predict land rights and slavery. t Whilethisisanendogenousvariable,thisisstillacorrelationimpliedbythemodel. For landrightstoexist,P mustbegreaterthanthecutoffsimpliedbyΩ(P ),1/α,andΦ(P ). t t t For slavery to exist, P must be great enough that enclosure of land is worthwhile and t the opportunity costs of coercion are not too high, but also sparse enough that wages arenottoolow. Itmust,then,bebetweenthecutoffsimpliedbyΩ(P )andΨ(P ). Itisthe t t impliedrelationshipsbetweenlandquality,populationdensity,landrights,andslavery that I test in assessing the ability of the “land abundance view” to explain pre-colonial institutionsinAfrica. 5 q 1 LE/F(P )= αP t (cid:16)β(cid:17) t LS(Pt)=(1+γ)1−αα(cid:16)βq +c¯(cid:17)α1Pt 6LA(Pt)=A¯+Dθ1Pt.