Collier and Gunning: Explaining African Economic ... of Economic Literature Vol. XXXVII (March 1999), pp. 64–111 Collier and Gunning: Explaining African Economic Performance Explaining

Journal of Economic LiteratureVol. XXXVII (March 1999), pp. 64–111

Collier and Gunning: Explaining African Economic Performance

Explaining African EconomicPerformance

PAUL COLLIERand

JAN WILLEM GUNNING1

1. Introduction

AFRICAN ECONOMIC performancehas been markedly worse than that

of other regions. During the 1980s, percapita GDP declined by 1.3 percent p.a.,a full 5 percentage points below the av-erage for all low-income developingcountries. During 1990–94 the declineaccelerated to 1.8 percent p.a. and thegap widened to 6.2 percentage points. Aliterature has rapidly developed whichattempts to explain this performance bygrowth regressions. This approach ishighly aggregative and reduced form,and is as yet unrelated to both case stud-ies and microeconomic research. We re-view these literatures and compare themto see whether they identify the samecauses of slow growth.

Section 2 reviews the explanations

offered by the endogenous growth lit-erature, grouping them into six factors.Slow growth can in part be explained interms of a few variables measuring theenvironment and institutions. However,initially studies found that a significantAfrica dummy remained: Africa wasgrowing inexplicably slowly. Subsequentresearch has focused on trying to elimi-nate this dummy. In the next two sec-tions we shift from this aggregate per-spective to the evidence on agents andmarkets. Section 3 focuses on twoagents: rural households and manufac-turing firms. Both have had to contendwith high risks. Farm households haveresponded by devising institutions thathave mitigated risk, by sacrificing in-come for security, and by retreatinginto untaxable activities, notably subsis-tence. Manufacturing firms have beenless able to adapt and so have fared par-ticularly badly. Drawing on the new lit-erature on “social capital,” we arguethat neither households nor firms haveas yet sufficiently created the social in-stitutions that promote growth. Section4 focuses on factor and product mar-kets. Both regulation and the high-riskenvironment have constrained thedevelopment of financial markets.The early literature on African labormarkets emphasized dysfunctional

64

1 Paul Collier is at the World Bank, on leavefrom the Centre for the Study of African Econo-mies, University of Oxford. Jan Willem Gunning isat the Centre for the Study of African Economies,University of Oxford, on leave from the Depart-ment of Economics, Free University, Amsterdam.The authors thank Anke Höffler for research assis-tance in Section 2, and Chris Adam, Janine Aron,Kees Burger, Bill Kinsey, Remco Oostendorp,Ritva Reinikka, Francis Teal, Steve Younger, andthree referees for comments. The findings, inter-pretations, and conclusions expressed in this paperare entirely those of the authors. They do not nec-essarily represent the views of the World Bank,its executive directors, or the countries theyrepresent.

government intervention through wageregulation. We suggest that the processthat accentuated financial repression—inflation—liberalized labor markets.Product markets have been charac-terized by extensive government inter-ventions through taxation, price setting,and public trading monopolies. Re-cently, many of these interventions havebeen reversed. However, such liberali-zations have not been fully credible,and so the effect of controls has provedpersistent. In Section 5 we bring to-gether the argument. Both a hostile en-vironment, particularly high risks, andinadequate social capital, particularlydysfunctional government, have low-ered the returns on investment. Thelow returns on investment have causedcapital flight on a massive scale.

Social capital is now improving. Al-though many African governments con-tinue to be dysfunctional, some havenow substantially liberalized controlsand improved service provision. De-mocratization is both underpinningthese policy changes politically and per-mitting formerly repressed civic institu-tions to emerge, thus building socialcapital. While high natural risks arepersistent, those risks arising from gov-ernment behavior are starting to dimin-ish. However, since these improve-ments are recent, their effect on growthremains largely prospective.

2. Aggregate Performance

In Table 1 we present four regres-sions that purport to explain Africangrowth performance. In each, the de-pendent variable is the average growthrate of per capita GDP over two dec-ades or longer. These regressions use anear-global sample of countries, andimpose the same specification for all re-gions save for the inclusion of regionaldummies as level or interaction effects.

Africa’s slow growth is “explained” if itis fully accounted for by differences be-tween Africa and other regions in thestandard explanatory variables. If suc-cessful, this implies that the Africadummy will be insignificant. Some re-gressions have found the Africa dummyto be both large and significant (e.g.Robert J. Barro and Jong-Wha Lee1993; William Easterly and Ross Levine1997); others eliminate it (Jeffrey D.Sachs and Andrew M. Warner 1997;Paul Collier and Jan Willem Gunning1997; Jonathan Temple 1998), thoughto an extent by transferring the puzzleelsewhere. Sachs and Warner find a sig-nificant “tropics” dummy. We find theAfrica dummy significant when inter-acted with some of the explanatoryvariables (as given in Table 1); interac-tions with the other variables in theregression are all insignificant.

Africa’s slow growth is thus partly ex-plicable in terms of particular variablesthat are globally important for thegrowth process, but are low in Africa.This shifts the question to why they arelow. Partly, slow growth is explicable interms of a distinctive effect of variablesin Africa, which shifts the question toexplaining this different response. Wenow organize the growth regression evi-dence on Africa, which includes theabove regressions but ranges muchwider, into six sets of variables.

A Lack of Social Capital. Social capi-tal can be generated both by the com-munity and by the government. Civicsocial capital is the economic benefitsthat accrue from social interaction.These economic benefits can arise fromthe building of trust, which lowerstransaction costs, from the knowledgeexternalities of social networks, andfrom an enhanced capacity for collec-tive action. Public social capital consistsof the institutions of government thatfacilitate private activity, such as the

Collier and Gunning: Explaining African Economic Performance 65

courts. We will argue that African gov-ernments have behaved in ways damag-ing to the long-term interests of the ma-jority of their populations because theyserved narrow constituencies. They

have been damaging partly through“sins of commission” such as agricul-tural taxation, and partly through “sinsof omission,” such as failure to provideadequate infrastructure.

TABLE 1THE AFRICA DUMMY IN FOUR GROWTH REGRESSIONS

Variable Barro-Lee Easterly-Levine Sachs-Warner Collier-Gunning

Period covered 1965–85 1960–89 1965–90 1960–89

Policies

Investment/GDP 0.078(0.028) 0.0013[4.72]Africa*Inv/GDP 0.00043[2.48]Openness 0.033(0.0087) 0.020[4.63] 8.48[3.41] 0.0184[5.21]Log GDP*Openness –0.77[2.52]Africa*Openness 0.0111[1.88]Institutions 0.28[3.49]Financial depth 0.015[2.54]Fiscal stance –0.131(0.037) –0.088[2.88] 0.12[5.34]

Initial Conditions

Initial Income –0.026(0.0038) 0.066[2.69] –1.63[7.89]Idem squared –0.005[3.10] –0.0009[6.97]Labor force-pop. 1.20[3.41]Landlocked –0.58[2.63] –0.00789[2.40]ELF –0.016[2.54] –0.0148[3.76]Male schooling 0.0090(0.0044)Female schooling –0.0052(0.0047)Schooling 0.009[2.28] 0.0148[1.64]Life expectancy 0.0712(0.0148) 45.53[2.58]Idem squared –5.40[2.39]Natural res. exports –3.28[3.30]Social disturbance –0.0163(0.0087) –14.874[1.56] –0.000004[0.10]Tropics –0.85[3.54]Latin America –0.0087(0.0037) –0.017[4.74] –0.0132[4.27]East Asia 0.0040(0.0057)Africa –0.0116(0.0051) –0.012[2.46] 0.02[0.05] –0.0052[0.98]

Notes: The dependent variable is the growth rate of per capita GDP, measured over the full period in Barro andLee, and Sachs and Warner, and as decade averages (with decade dummy variables) in Easterly and Levine, andCollier and Gunning. Openness: proxied by the parallel market premium, except for Sachs and Warner who devisetheir own classification. Fiscal stance is measured by Barro and Lee as G/Y; by Easterly and Levine, and Sachs andWarner, as fiscal surplus/Y. Schooling is enrollment in secondary schooling in 1960; Easterly and Levine, and Collierand Gunning enter this as the log of enrollment. Social disturbance is proxied by revolutions in Barro and Lee, byassassinations in Easterly and Levine, and by months of civil war in Collier and Gunning. For test statistics on thefirst three regressions see the original sources. For our regression: n = 236, formed from an unbalanced panel ofdecade average growth rates for 84 countries, of which 21 are in Sub-Saharan Africa: adjr2 = 0.56; F [13,222] =23.91; Breusch-Paqgan chi-squared = 58.8; results are corrected for heteroskedasticity. Standard errors are shown(-), t-scores are shown [-].

66 Journal of Economic Literature, Vol. XXXVII (March 1999)

On various measures Africa is rela-tively lacking in both types of socialcapital (Table 2). These measures havebeen found to be significant in growthregressions, but there is no evidencethat their effect in Africa differs fromthat elsewhere. We now consider whyAfrica is short of social capital.

Possible barriers to social interactionare ethno-linguistic fractionalizationand inequality. Africa has a strikinglyhigh level of fractionalization. An in-dex of ethno-linguistic fractionaliza-tion has been constructed, based on ameasure created by Soviet anthropolo-

gists.2 On this measure the averageAfrican country is more than twice asfractionalized as other developing re-gions. This is likely to be the result ofthe historical low population density ofAfrica. Inequality measures are not veryreliable for Africa. We use the incomeshare of the third and fourth quintiles,which is sometimes referred to as theincome share of the middle class. Onthis measure Africa is identical to otherdeveloping countries. Hence, regardlessof the effect of inequality on growth, itcannot account for any part of the dif-ference in growth performance. Thegrowth regression in Table 1 shows thatfractionalization has a significant nega-tive effect. According to Easterly andLevine, it directly accounts for 35 per-cent of Africa’s growth shortfall, and,because it is also correlated with poorpolicies, overall it accounts for 45 per-cent of the growth shortfall. However,the negative growth effect of ethnicdiversity only applies in societies lack-ing political rights. In countries withdictatorships, ethnic diversity reducesthe growth rate by 3 percent per an-num. Conversely, in countries withfull democratic rights diversity has nodetrimental effect (Collier 1998).

Ethnic diversity has indeed beencostly in Africa, but this is because ofthe low level of political rights. Thegovernments that came to power in in-dependent Africa had two distinctive in-heritances. They were drawn from atiny elite of educated young men andso were separated from the mass ofthe population. This lack of repre-sentativeness rapidly became institu-tionalized. By 1975 most African stateslacked a legislature, and most chief ex-ecutives were unelected: over 60 per-cent of the population lived under such

TABLE 2SOCIO-POLITICAL INDICATORS: DIFFERENCESBETWEEN SUB-SAHARAN AFRICA AND OTHER

LDCs

SSA Other LDCs

Corruption 4.97 6.03Bureaucracy 1.38 1.72Enforceability 1.95 2.09Civil war 1.27 0.72Fractionalization 67.6 32.7 Social development 1.10 –0.43Inequality 31.0 31.0

Notes: Corruption: data from International CountryRisk Guide for 1982; low scores indicate high cor-ruption. Quality of bureaucracy: source as corrup-tion, high scores indicate better quality; range is 0–6.Enforceability of contracts: data from BusinessEnvironmental Risk Intelligence for 1972; lowscores indicate weak enforceability; range is 0–4. Theindex of fractionalization is on the range 0–100 withcompletely homogeneous societies scored as zero.Adelman-Morris Index of “social development” as ofthe early 1960s is constructed on the effective range1.86 (least) to –1.91 (most) over 74 countrieswhich they classified as developing at that time.Inequality: the income share of the third and fourthquintiles. Sources: Corruption and fractionalizationfrom Mauro (1995); bureaucracy and enforceabilityfrom Knack and Keefer (1995); civil war (months peryear) from Singer and Small (1994); A-M Index andinequality, from Temple (forthcoming).

2 The underlying measure is of the probabilityof two randomly drawn citizens being from differ-ent ethno-linguistic groups.


regimes (Ferree, Sigh, and Bates 1996).Secondly, the modern part of the econ-omy was owned by ethnic minorities.3

The combination of the divorce of thegovernment from the population andthe lack of African ownership of the“commanding heights” induced a policyof taxing export agriculture to financethe expansion of industry. Interventionsalso tended to be grandiose: in Tanzaniathe majority of the rural population wasrelocated in a single year, 1974.

As the government expanded its ownemployment, it and its industrial alliescame to have an interest in cheap urbanfood. Subsidizing industry and urbanfood was financed by explicit and im-plicit taxation of exports: mineralswhere these were available, otherwise,agricultural exports. This was supportedby controls on the banking system thatboth lowered interest rates and directedcredit to the favored sectors, thus be-ginning financial repression. This biasagainst export agriculture in favor ofimport-substituting industry was com-mon both to socialist regimes (Ghana,Tanzania, Zambia, Uganda, Ethiopia,Mozambique, and Angola), and to moremarket-oriented ones (Nigeria, Zaire,Zimbabwe). The exceptions were wherethe political elite itself had a stronginterest in export agriculture (Côted’Ivoire, Kenya, Malawi). These anti-export policies were widely adoptedbetween the mid-1960s and the mid-1970s. In some countries their effectswere temporarily disguised by the com-modity booms of 1975–79, but by theearly 1980s most African economieswere declining. Although there wassome increase in accountability be-

tween 1975 and 1980, this was largelyreversed by 1985. After the opening ofEastern Europe, there was a wave ofpolitical liberalization in Africa, buteven by 1991 only 13 percent of thepopulation was living in states in whichlegislators had been chosen in con-tested multiparty elections, and only 10percent in states in which the chiefexecutive had been so chosen.4

Being insulated from the mass of therural population and having only limitedlegitimacy, governments were acutelyexposed to pressures from their ownnarrow base of supporters. As is com-mon with control regimes, although theintensity of controls raised the costsborne by the majority of the population,they also raised the benefits to the mi-nority. Much of the industrial sectorwas entirely reliant upon the controls.For example, in Ghana during thecontrol regime, a quarter of manufac-turing output was estimated to be pro-duced at negative value added at worldprices (Douglas Rimmer 1990). In Ni-geria, manufacturing output declined bya third in the decade after the end ofthe oil boom, as state subsidies andprotection were reduced.

In summary, as Robert H. Bates(1981) argued:

They [‘the mass of rural producers’] sufferfrom government attempts to implement anadverse structure of farm prices, and they failto benefit from the compensatory paymentsconferred through the subsidy programs. Be-cause of official repression they lack politicalorganisations with which to defend theirinterests. (p.121)

One corollary of poor public social capi-tal was a high incidence of corruption. Inmany cases, because the state was weakas well as autocratic, corruption was3 Ironically, the first independent African gov-

ernment to face these concerns was the AfrikanerNationalist government of South Africa, which oncoming to power found that 99 of the top 100companies were English rather than Afrikanerowned.

4 We wish to thank Ferree, Singh and Bates forproviding us with recalculations of their data toshow the population percentages reported hereand above.


uncoordinated and hence competitive.Such corruption is much more costly tosociety than the centralized and there-fore monopolistic corruption of Asia:bribes exceed the revenue-maximizinglevel and can even eliminate transactions(Andrei Shleifer and Robert WardVishny 1993). Another corollary of poorpublic social capital has been that gov-ernments have adopted damaging eco-nomic control regimes with high tradebarriers, while attaching low priority tothe delivery of public services. Weconsider these in turn.

A Lack of Openness to Trade. Themain manifestation of African controlregimes was the restriction of interna-tional trade, whether directly throughquotas, tariffs and export taxes, or indi-rectly through foreign exchange con-trols and marketing boards. By the1980s, Africa had become less openthan other regions. On one measure,not only was Africa the area with thehighest trade restrictions, but the gap

between it and the next most restrictivearea, the Middle East, was wider thanthat between the Middle East andthe most liberalized region, the FarEast (David Dollar 1992). On a second,binary measure, almost all Africaneconomies were closed, whereas 37 per-cent of other developing countrieswere open (Sachs and Warner 1995,1997).5 Table 3(a) summarizes variousmeasures of trade policy.

The effects on growth can be sub-stantial: Sachs and Warner (1997) makethe largest claim, finding that restric-tive trade policy, poor access to the sea,and Dutch disease between them ac-count for a 1.2 percent per annumgrowth shortfall. A more typical esti-mate would be the 0.4 percent found byEasterly and Levine (1997).

Not only have trade restrictions beenmore severe in Africa, but this has beenin the context of very much smallereconomies. There is therefore a reason-able presumption that, on average, agiven level of restriction would be moredamaging. In Table 1 we investigatethis by adding to a standard growth re-gression a term that interacts an Africadummy with a trade policy proxy, theparallel premium.6 The term is negativeand significant: a given level of trade re-strictions is around half as damagingagain in Africa as in other developingareas.

Thus, openness explains why Africahas grown more slowly than other re-gions both because openness is impor-tant for growth, while Africa has beenmuch less open than other regions,and because a given level of trade

TABLE 3OPENNESS INDICATORS:

DIFFERENCES BETWEEN SSA AND OTHER LDCs

SSA Other LDCs

(a) Policy:

Sachs and Warner index (1 = open)

0.04 0.37

parallel premium (%) 0.40 0.26RER misalignment (%) 0.13 0.07

(b) Endowment:

landlocked (1 = yes) 0.45 0.06market access 0.23 0.04natural resources 0.14 0.11

Sources: Sachs and Warner index: Sachs and Warner(1995); parallel premium: Easterly and Levine (1997);RER misalignment: Elbadawi and Ndulu (1996);landlocked: Sachs and Warner (1995a); market access:Sachs and Warner (1997); natural resources: Sachs andWarner (1995).

5 Rodrik (1998) criticises both of these mea-sures. He finds that Africa’s participation in inter-national trade is normal for its non-policy charac-teristics. Coe and Hoffmaister (1998), using agravity model, support this result.

6 The Sachs and Warner index gives similarresults.


restrictions has been more damaging inAfrica than elsewhere.

Deficient Public Services. The restric-tive trade policies adversely affectedpublic service delivery, lowering the re-turn on public sector projects (JonathanIsham and Daniel Kaufmann 1999).Public services have also worsened dueto the lack of civic social capital. Be-cause African governments have permit-ted only a low level of civil liberties, or-dinary people are denied the channel ofpopular protest, and this worsens pro-ject performance (Isham, Kaufmann,and Pritchett 1995). A minority of thepopulation feels sufficiently confidentto protest when its interests are threat-ened, whereas the majority has been ex-cluded and is quiescent. Since protest iseffective, this bifurcation of the popula-tion has entrenched the control regimewith its highly selective beneficiaries,while permitting more generalized ser-vice delivery to be inefficient. A dra-matic instance of these selective priori-ties is that expenditure per student intertiary education is 44 times that on aprimary school pupil, compared with arange elsewhere in the world of 3–14

times (Sanjay Kumar Pradhan 1996, Ta-ble 4.11). However, the return on Afri-can projects remains significantly lowerthan elsewhere, even controlling for thepolicy environment and civil liberties.The public sector has been used to cre-ate employment rather than to deliverservices, and this reduces productivity.To finance extra employees, non-wageexpenditures are squeezed and wagesare reduced, being compensated by de-clining effort. Because Africa had themost rapid educational expansion in theworld, pressure for job creation was un-usually severe (Alan Gelb, John Knight,and Richard Sabot 1989). There is alsoevidence that wages in the public sectorreward kin group connections ratherthan skills (Collier and Ashish Garg1999). Hence, there are several chan-nels causing public service delivery tobe deficient.

The share of public expenditure inGDP is generally higher than in otherdeveloping regions, and expenditure onthe most evidently nonfunctional items,namely defence and interest payments,is lower. Whereas other developing re-gions allocate 16–17 percent of GDP to

TABLE 4GOVERNMENT EXPENDITURES AS A PERCENTAGE OF GDP, AVERAGE OVER 1985–89

Purpose SSA S. Asia E. Asia Latin America

Total Expenditure 26.0 23.5 22.9 20.4

(Defense Expenditure) 1.8 2.6 2.1 1.6(Interest Payments) 2.9 3.5 3.0 3.6

Total Potentially Productive Expenditure 21.3 17.4 17.8 16.2

General Public Services 6.9 6.2 6.3 5.3Education 3.1 1.8 4.8 2.8Health 1.3 0.9 1.7 1.2Social Security and Welfare 0.5 1.4 0.7 0.9Economic Services 5.7 7.2 6.1 3.7

Source: Pradhan (1996), Tables 2.10 and A2.2.


potentially productive public expendi-tures, African governments allocate 21percent (Table 4). Despite this, the ac-tual delivery of public services has beenpoor.

The public service that has receivedmost attention in growth regressionshas been education. Paradoxically, edu-cation is the one public service in whichAfrican performance has not beenmarkedly worse than other regions. Al-though Africa has a lower stock of edu-cation than other regions, it has had afaster rate of growth of the stock (seeVikram Nehru, Eric Swanson, andAshutosh Dubey 1995). If the growth ofoutput is determined by the growth ofinputs, then Africa should have grownrapidly. Conversely, if the level of hu-man capital directly affects the rate ofgrowth, then the lack of educationmight be an important explanation ofslow African growth. At present theseissues are unresolved, partly becauseuntil very recently educational growthrates have been very poorly proxied byenrolment rates (Norman Gemmell1996; Lant Pritchett 1996). However,one unambiguous effect of the low levelof education has been high fertilityrates. African population growth hasexceeded labor force growth by 0.4percent p.a., and this reduces per capitagrowth approximately pro rata, asestablished in the Sachs and Warnerregression.

Despite the growth in enrolments,there is evidence of serious deficienciesin the implementation of educationalpolicies. There is a wide gap betweenplans and schools: less than 10 percentof education policies have been imple-mented (John Craig 1990), and evenwhen money is voted for a specific pur-pose, much of it does not reach its in-tended destination. Less than 40 per-cent of the money allocated by theUgandan Ministry of Finance on the

primary school non-salary vote in 1991–95 actually reached primary schools(Emmanuel Ablo and Ritva Reinikka1998). Generally across the social ser-vices, even when money reaches theservice provider, wages are prioritisedover non-wage recurrent inputs: Com-pared with South Asia, Africa spendsdouble the amount on wages per dollarof non-wage recurrent expenditure. Asa result, services suffer. For example, inGhana the average public clinic had 2.2times the number of staff and a 25–percent lower probability of havingdrugs than private facilities. In Kenya,public clinics had ten times morestaff and twenty times the number ofdays without antibiotics as the privatefacilities. In both, the resulting lowerquality of public facilities reduced theirusage.7

There is less infrastructure than else-where. For example, the density of therural road network is only 55 kilometersper square kilometer, compared to over800 in India, and there are only one-tenth the telephones per capita of Asia.The quality of infrastructure is alsolower. The telephone system has triplethe level of faults of Asia’s, and the pro-portion of diesel trains in use is 40 per-cent lower. Prices of infrastructure useare much higher. Freight rates by railare on average around double those inAsia. Port charges are higher (for exam-ple, a container costs $200 in Abidjan asopposed to $120 in Antwerp). Air trans-portation is four times more costly thanin East Asia. Much of internationaltransport is cartelized, reflecting theregulations of African governments in-tended to promote national shippingcompanies and airlines. As a result ofthese high costs, by 1991 freight and in-surance payments on trade amounted to

7 See Lavy and Germain (1994) and Mwabu,Ainsworth and Nyamete (1993).


15 percent of export earnings, whereasthe average for developing countries isonly 6 percent. Further, the trend hasbeen rising for Africa whereas it hasbeen falling elsewhere: the comparablefigures for 1970 were 11 percent and 8percent (Azita Amjadi and Alexander J.Yeats 1995).

The evidence on the growth effects ofsuch underprovision is as yet scant.There is some evidence that the infra-structure expenditure of African govern-ments is growth-enhancing (Dhanesh-war Ghura and Michael T. Hadjimichael1996; Temitope Oshikoya 1994) andthat deficiencies in the operation of so-cial services and infrastructure havelowered the rate of return on publicprojects to significantly below those inother regions.8 The most specific aspectto be investigated has been telephoneprovision, for which the deleterious ef-fects are found to be large (Easterlyand Levine 1997).

Geography and Risk. Africa is distinc-tive with respect to climate, location,and comparative advantage. Sub-Saha-ran Africa is predominantly tropical.There is some evidence that this has re-duced African growth, an importantchannel being proneness to malaria(Sachs and Warner 1997; John LukeGallup and Sachs 1998). Much of thecontinent is semiarid, and this has madeagricultural production intrinsicallyrisky. One-third of the available land istoo dry for rainfed agriculture andabout one-half of the rest is of marginalquality (Food and Agriculture Organiza-tion 1986; Adrian Wood and Jörg Mayerforthcoming). Soil quality is poor. Sincethe 1960s, there has been a deteriora-

tion trend in African rainfall.9 However,there is too little evidence to infer itseffect on growth. A corollary of thesesemiarid conditions is low populationdensity. Because of this, transport costsare intrinsically high even aside frompolicy-induced deficiencies of infra-structure. This is compounded by thefact that the African population is dis-proportionately landlocked. Table 3(b)summarizes some endowment-relatedreasons why Africa is relatively closedto international trade.

Low population density implies thatthe African population is relatively well-endowed with natural resources. Woodand Mayer show that this gives Africa acomparative advantage in natural re-source exports. However, a conse-quence of export concentration in natu-ral resources is that Africa’s terms oftrade are determined by commodityprices. As a result, Africa’s terms oftrade have been volatile. These shockshave had large short-term effects onoutput: a shock worth one percentagepoint of GDP directly changes constantprice GDP by 0.6 percent (Angus Dea-ton and Ron Miller 1996). Most evi-dence suggests that this volatility hasreduced growth.10 However, this is con-tentious. Exposure to trade shocks is

8 Isham, Kaufman, and Pritchett (1995) find thisfrom a study of the rates of return on all 3,435World Bank projects over the period 1974–93 ascalculated from ex post evaluations, and show thatthis is a reasonable proxy for the rate of return onpublic sector investments in general.

9 Grove (1991) compares mean rainfall in theperiod 1960–89 with that during 1930–59. In thedriest zone, covering the Sahel, the Sudan, andNorthern Ethiopia, rainfall was 20–30 percentlower in the post-1960 period and in the interme-diate zone, Natal-Kalahari, it was 10–20 percentlower. In the wet Equitorial zone, where marginalchanges in rainfall were probably least important,mean rainfall increased by 10 percent.

10 See David Wheeler (1984) and Gerald K.Helleiner (1986). David Bevan et al. (1987a, 1990)and Collier and Gunning (1998) argue that Africantrade shocks have caused economic decline be-cause over-regulation has prevented private agentsfrom responding efficiently, while the govern-ments have lost control of public expenditure.They find that even favorable shocks give rise tolarge contractions in GDP in the post-shockperiod.


potentially beneficial: if the economy isflexible, resources can be deployed inwhichever sector has the highest prices.Deaton and Miller construct exportprice series for 36 African countries,and find no negative effect of priceshocks on growth and investment in theshort term.11

Since the 1980s, the terms of tradehave also tended to deteriorate, al-though there is surprising variation inthe estimates, ranging between 6 per-cent and 36 percent (Collier and Gun-ning 1997, Table 7). This deteriorationin the terms of trade has been esti-mated to account for a reduction in theAfrican growth rate relative to that ofother developing countries of 0.7 per-centage points (Ibrahim A. Elbadawiand Benno J. Ndulu 1996).

This natural volatility is compoundedby policy volatility.12 Although the two

sources of shocks are conceptually dis-tinct, in practice in Africa they are cor-related, since governments have usedtrade policy to equilibrate the balanceof payments. Variation in the real ex-change rate is a useful proxy for theseeffects since it reflects the sum of pol-icy and terms of trade shocks. Africanreal exchange rates have been atypicallyvolatile, and there is some evidencethat this has reduced growth.13 Table5 summarizes a range of variabilitymeasures.

A Lack of Financial Depth. As dis-cussed in Section 4, Africa has muchless financial depth than other develop-ing areas: M2/GDP is 37 percent lower(Table 6). This is usually interpretedas being due to the curtailment of

TABLE 5VOLATILITY INDICATORS: DIFFERENCES BETWEEN SSA AND OTHER LDCs

Inflation

Terms of

Trade

Real Exchange

Rate

Tax Revenue

GDPMacro-

Financial GDP

Sub-Saharan Africa 9.0 16.4 15.1 3.3 1.048 6.14South Asia 7.5 10.4 10.4 1.7 0.733 4.38East Asia 6.0 11.2 6.6 1.7 0.733 4.38Latin America 13.9 15.4 14.9 2.2 1.121 5.17Middle East 7.2 14.3 11.2 1.2 1.094 8.01

Notes: For the Macro-Financial and GDP measures of volatility, Asia is treated as a single aggregate. The samefigure has been used for East and South Asia.Sources: Inflation: (Variance, 1960–89), King and Levine (1993); Terms of Trade shocks: standard deviation ofannual log changes (X100), 1965–92, Collins and Bosworth (1996): Real Exchange Rate: (Standard Deviation ofchanges, 1966–88), Pritchett (1991); Tax Revenue/GDP: (Standard deviation of changes, 1974–89), Bleaney et al.(1995); Macro-Financial: (Standard deviation of equal-weighted average of four underlying measures of volatility:public deficit/GDP; monetary growth; real exchange rate misalignment; current account deficit/GDP, 1961–94,Elbadawi and Schmidt-Hebbel (1997); GDP: (Standard deviation of per capita GDP, 1961–94), Elbadawi andSchmidt- Hebbel (1997).

11 As Deaton and Miller note, their approachrules out long-term effects since reversion totrend is imposed.

12 The volatility of commodity prices may itselfreflect policies rather than natural endowments

since the specialization in commodities is the re-sult of the hostility of the policy environment toother exports.

13 Real exchange rate variability has yet to beused in growth regressions but has been found toreduce African investment (Ghura and Grennes1993; Hadjimichael and Ghura 1995; Bhat-tacharya, Montiel, and Sharma 1996; Fielding1993).


the banking system through financialrepression. However, while M2/M0 is in-deed lower in Africa than elsewhere,the difference is only 16 percent, sothat the main cause is the low level ofM0/GDP. The low holdings of currencyare probably attributable to the largeshare of the subsistence economy, inturn related to the high implicittaxation of agriculture.

Although the lack of financial depthin Africa has reduced growth, the effecthas been modest: a loss of only 0.3 per-centage points (Easterly and Levine1997). There is also some evidence thatit has reduced investment.14 Thus, thelack of financial depth is probably morea by-product of a lack of openness thanof financial policies and has had onlylimited effects on growth. This con-trasts with the considerable emphasisplaced on financial liberalization inreform programmes.

High Aid Dependence. Whereas aid isperipheral to low-income countries as a

group, for Africa it is nearly five timeslarger as a share of GNP: 12.4 percentas opposed to 2.7 percent as of 1994.Potentially, aid dependence has there-fore exerted a powerful effect on Afri-can growth. The effect of aid on growthdepends upon the policy environment(Craig Burnside and Dollar 1997). Ingood policy environments (proxied by acomposite measure for macroeconomicpolicies) aid significantly increasesgrowth, whereas in poor environmentsit actually reduces it. On average, theAfrican policy environment has beenpoor on this measure. The critical pol-icy score below which aid reducedgrowth was 1.5. Of the 21 African coun-tries studied, only Botswana consis-tently had a score above this level, and15 consistently had scores below it, withthe rest having bouts of good policy.The 15 had a mean score of only 0.8, alevel at which the regression would pre-dict growth reduction.15 Hence, surpris-ingly in view of the scale of aid flows,there has been little net effect onAfrican growth.

An Interim Assessment. To what ex-tent do the factors included in thegrowth regressions account for the Afri-can growth shortfall? The evidence dis-cussed above suggests that the lack ofopenness to trade and the low level ofsocial capital have had large, damagingeffects on the growth rate. The effectsof high policy volatility and poor publicservices on growth may also be highlydamaging but are less well established.Geography has probably also played apart in reducing growth, through hightransport costs, poor soils, disease, cli-matic risk, and export concentration incommodities, although only a few ofthese effects have been quantified.

Cumulatively, the above variables

TABLE 6FINANCIAL DEPTH AND GROWTH: DIFFERENCES

BETWEEN SSA AND OTHER LDCs

Means of Explanatory Variables

SSA Other LDCs

M2/GDP 0.22 0.35 M2/M0 2.18 2.61 savings rate 10.45 17.52 fiscal deficit –0.049 –0.042

Sources: M2/GDP calculated from data supplied byKing and Levine (1993). We would like to thank themfor making their data available. M2/M0 from simpleaverage of 35 SSA countries excluding South Africaand 67 non-African developing countries, excludingEurope, IFS Yearbook, 1994, data for 1989, Table 39ab.Savings rate from Sachs and Warner (1997). Fiscaldeficit from Easterly and Levine (1997).

14 See Hadjimichael and Ghura (1995); Had-jimichael et al. (1995); Oshikoya (1994).

15 We would like to thank Burnside and Dollarfor providing us with their underlying scorings ofthe African countries in their sample.


have contributed to a capital-hostileenvironment that has lowered the rateof return on investment. In Table 1,the interaction of the investment andthe Africa dummy indicates that therate of return in Africa has been one-third lower than in other regions. Thisin turn has reduced the rate of privateinvestment. Since the 1980s the pri-vate capital stock per worker has de-clined by 20 percent and is now onlyone third of that in South Asia, thenext most capital-scarce continent.Hence, the most capital-scarce regionhas nevertheless had low returns oninvestment.16

One limitation of the growth regres-sion literature is that to date it has fo-cused upon explaining long-term aver-age African slow growth. This missesthe deterioration in performance sincethe 1960s. The contrast with South Asiaset in only during the 1970s. During the1960s, growth and its sources were verysimilar; post-1973 Africa has experi-enced a collapse in both investment andproductivity (Table 7). Evidently, thiscollapse cannot be attributed to initialconditions such as geography. Althoughthe most obvious inference is that de-

celeration is explained by policy dete-rioration, other factors omitted fromthe regressions, notably the climate andterms of trade, also deteriorated. Arelated limitation is that episodes ofsevere decline are not adequately ana-lysed by the practice of averagingvariables over long periods. Africanperformance has been strongly episodic:only Kenya has been characterized bypersistent slow growth; other countrieshave had episodes of severe decline,the average being a six-year period dur-ing which per capita GDP falls by 25percent. If these episodes were sepa-rately identified, nonlinearities and hys-teresis might be found. A third limita-tion is that no allowance is made forneighborhood effects such as policycontagion. An East Asian governmentwith poor policies, such as that of Viet-nam in the 1980s, could and did copyfrom its neighbors. African govern-ments have not benefited from thesame effect, and until recently, theircontinental information exchange or-ganizations, ECA and ADB, reinforcedpolicy immobility rather than fosteredchange. Easterly and Levine (1998)eliminate the African dummy onceneighborhood effects are included, al-though the result is disputed (Sachs andWarner 1997).

TABLE 7SOURCES OF GROWTH: AFRICA AND SOUTH ASIA COMPARED

Output per worker

Capital per worker

Education per worker

Total factor productivity

1960–73 Africa 1.9 1.3 0.2 0.3 South Asia 1.8 1.4 0.3 0.1

1973–94 Africa –0.6 0.4 0.2 –1.3 South Asia 2.6 0.9 0.3 1.3

Source: Collins and Bosworth (1996).

16 There is now a literature that endogenises theinvestment rate. See, for example, Hadjimichaeland Ghura (1995).


3. Households and Firms: Responses to Risk and Dysfunctional Government

The aggregate evidence reviewedabove depicts an environment charac-terized by intrinsically high risks, hightransport costs and trade barriers, poorinfrastructure, low levels of education,limited financial markets, and highregulation. In considering how agentshave responded to this environment, wefocus upon two key groups: rural house-holds, which have faced many of theseproblems for generations, and manufac-turing firms, which in Africa are largelya creation of the past 50 years. Similarforms of risk-sharing are found bymeans of state-contingent contracts invillage credit markets, in inter-firmgoods markets, and in formal labor mar-kets, but only when social institutionshave enabled high observability in thecontext of long-term relationships. Weargue that rural households have beenmore successful than manufacturingfirms in developing institutions thatreduce the costs of operating in thisenvironment.

3.1 Rural Households

The precolonial African rural econ-omy was in a land-abundant, high-risk,near-subsistence, low-asset equilibrium.Population was limited by wars andpoor nutrition. Social institutions re-flected the needs of this stationaryeconomy, with lineage groups regulat-ing intergenerational transfers and pro-viding risk pooling, but not usually cre-ating marketable property rights orsecuring long-distance trade. Under co-lonial rule the population began to in-crease, partly as a result of peace andbasic public health measures, leading toland scarcity. Peasant agriculture be-came commercialized both by the intro-duction of export crops and by the im-position of taxes and the sale of labor,

often on forced terms. New crops andbreeds of livestock were introduced. Ofthese three processes, populationgrowth gradually accelerated duringthe century, largely due to continuingimprovements in public health, propel-ling the economy from land abundanceto land scarcity in a remarkably shortspace of time. Commercialization andopportunities for innovation slowedand were sometimes even reversedpost-independence, reflecting anti-agri-cultural policies, which we discuss inSection 4.

Social institutions needed to adapt tothese changes, and gradually did so. Animportant issue is whether the pace ofadaptation has been so slow that socialcapital is radically deficient and so a se-rious drag on growth, as suggested bythe aggregate-level regressions.

High Risk and Volatility. Undersemiarid conditions with few invest-ment opportunities and consequent lowpopulation density, agriculture will takethe form of autarky (production for ownconsumption) without a market for per-manent labor (Hans P. Binswanger andJohn McIntire 1987). Farmers facestrikingly greater risks in Africa thanelsewhere. For example, in some areasof Ethiopia, Zimbabwe, and Tanzania,near total crop failure has a probabilityof about 10 percent.17 The high degreeof price volatility noted in Section 2 is afurther source of income risk. As aresult, crop income is highly variable.For example, in Burkina Faso thecoefficient of variation is 67 percentfor the Sahelian zone and 52 percentin the Sudanian zone (Thomas A. Rear-don et al. 1992). However, the combina-tion of moral hazard, covariance of therisks, and geographic isolation pre-cludes insurance. Credit markets cannot

17 See Webb and Reardon (1991) and Kinsey etal. (1998).


be based on collateral: land has littlevalue since it is abundant, and animalsare vulnerable to sickness and theft.Hence credit must be based on highobservability instead of collateral.Where this is not feasible, agents canprotect themselves against shocks boththrough ex ante responses, such as di-versification, and ex post, through con-sumption smoothing, and the choice dif-fers widely both within and betweensocieties.

In diversifying to cope with shocks,the household sacrifices the gains ofspecialization in favor of spreading riskover multiple income-generating activi-ties. A striking stylised fact about Afri-can rural households is that they havehighly diversified economic activities,many of them nonagricultural. Onemeasure of this is that African farmersspend a much smaller fraction of theirworking hours farming than Asian farm-ers, who face less risk because of thewidespread use of irrigation (JohnHogarth Cleave 1974). The first large-scale rural household survey in Africa,conducted in 1974–75 in Kenya, foundthat smallholders derived only half theirincome from farming their own holding,the rest coming from wage employ-ment, nonfarm enterprises and remit-tances (Kenya 1977, p. 54). In WestAfrica, a series of studies in eight coun-tries found an average share of nonfarmincome of 39 percent (Reardon et al.1994, p. 210). In addition, farmingoperations themselves are highlydiversified. Many households chooseto fragment their holdings into manyplots, and nearly all households com-bine food growing with livestock, cashcrops, or wage employment. That diver-sification is partly a response to climaticrisk is shown by differences in the ex-tent of crop diversification betweenecological zones: in the humid forestareas where rainfall is reliable, house-

holds often are highly specialized, grow-ing only one or two crops. Similarly,households living in the Sahelian zonein West Africa are more diversified thanhouseholds in areas with more reliablerainfall.18

However, diversification is of coursecostly. Fragmentation of the holdingto reduce risk increases travel time.Similarly, when multiple cropping isadopted as a response to risk,19 it hasboth static and dynamic costs. In staticterms, the household engages in activi-ties even if they offer only low returns,as long as they are either safe or haverisks uncorrelated with other activities.Dynamically, the household reduces itsscope for learning-by-doing.

Now consider the other risk-copingstrategy, consumption smoothing. Po-tentially, consumption smoothing canbe done either through credit marketsor through asset markets. However, inAfrica the former is fairly rare. For ex-ample, even in the extreme circum-stances of the Zimbabwean drought of1991/92, credit played virtually no role.The accumulation of realizable assetsenables the agent both to cope withshocks and to smooth consumption overthe lifecycle. However, in the tradi-tional African economy there were fewliquid assets other than livestock (andin tse-tse infected areas even this op-tion was largely closed).20 Also, thestrategy of accumulating liquid assetscan be constrained by the capacity todefend them: there can be a trade-offbetween security from economic shocks

18 See Bevan et al. (1989), Carl K. Eicher andDoyle C. Baker (1982) and Reardon et al. (1988,1994).

19 There are, of course, other reasons for multi-ple cropping; e.g. cereals may be underplantedwith legumes to provide nitrogen.

20 Deaton (1992, 1994) demonstrates consump-tion smoothing on data for the Côte d’Ivoire, Udry(1994) does so for Nigeria, and Dercon (1997) forTanzania.


and from violence.21 Nevertheless, con-sumption smoothing by means of assetsis important. Farmers in Côte d’Ivoireand Nigeria accumulated assets in yearspreceding (anticipated) negative shocks.Conversely, during positive shockssmallholders have high savings rates:during the 1975–77 boom, when cof-fee prices temporarily trebled, the mar-ginal savings rate for Kenyan coffee-growing households was approximately65 percent.22

Although consumption smoothing iscommon, the main assets used for it,livestock and food stores, are ill-suitedto the purpose.23 The unreliability offood markets implies that at least partof the asset smoothing must be done di-rectly by holding substantial foodstores, but this is costly due to lossesfrom spoilage and vermin.

In summary, farm households are ex-posed to large risks. Their responsesinclude self-insurance through diversifi-cation, both within agricultural activi-ties and between agricultural and non-agricultural activities. They alsoaccumulate assets for consumptionsmoothing. Both responses are likely toreduce growth, the former by loweringmean income and thereby savings,the latter by the need to keep assets inliquid form.

A Lack of Rural Social Capital.There has been a major shift in econo-mists’ perception of traditional Africansociety. The early literature saw its indi-viduals as economically irrational andits institutions as economic impedi-ments. For example, the absence of pri-vate property was seen as giving rise toan acute “tragedy of the commons.”During the 1960s, private behaviorstarted to be recognized as a rational re-sponse to constraints.24 By the 1980s,traditional social institutions came to beseen as efficient solutions to economicproblems.

Africa’s traditional societies evolvedinstitutions that lowered the costs ofmoral hazard and adverse selection.These institutions were the village andthe kin group. Membership of the kingroup was based on birth and hence wasnonelective, thus solving the problem ofadverse selection. There was little pri-vacy in traditional society. Living inclose proximity reduced informationcosts and so reduced moral hazard.There is evidence of intra-village insur-ance as a result of this virtually com-plete information, repayment of creditbeing state-contingent.25 Clearly, how-ever, intra-village insurance has limita-tions, since people living close togetherare likely to be subject to highly covari-ant risk.26 Lineage rules of inheritanceenforced intergeneration transfer pay-ments. The kin group was able to

21 Most precolonial African societies had notcreated an institution with a monopoly of violenceover a secure territory. In such a social environ-ment income is reduced: each agent has an incen-tive to devote resources to violence (such asstealing cattle) and to defend against the violenceof others. Since asset accumulation increases therisk of violence, security can be sustained throughpoverty. Hence, “peace is purchased at the cost ofpoverty.” (Greif and Bates, 1995)

22 See Kinsey, Burger, and Gunning (1998) onZimbabwe, Deaton (1992) on Cote d’Ivoire, Udry(1994) on Nigeria, and Bevan et al. (1989) onKenya.

23 During a drought animals are liable to die andcattle prices will fall. During the 1984 drought inNiger livestock prices halved (Fafchamps, Udry,and Czukas 1998).

24 The locus classicus is Jones (1960).25 See Posner (1980) and Platteau (1994) for an

economic interpretation of African social institu-tions, and Udry (1993) for evidence of insurancein a Nigerian village.

26 Similar systems have been described for com-munities of traders. Among the Hausa-Fulani inNorthern Nigeria, membership was elective, butconsiderable investment was involved in becominga member, e.g. the requirement to convert toIslam and to learn Hausa. Among the Magribitraders there were draconian exclusion rules forthose caught cheating. See Platteau (1994) andGreif (1993).


enforce adherence to each particularrule through the threat of exclusionfrom the entire package of benefits.

In addition to providing insuranceand intergenerational transfers, tradi-tional social organization largely solvedthe problem of the management ofcommon pool resources. Because theuse of common pool resources occurredin the context of frequent interactionand easy observation, it was relativelystraightforward for societies to regulatetheir use.27 This was, however, not rec-ognized (one of the major mispercep-tions of economic policy towards tradi-tional agriculture); policy makers sawthe solution to the supposed “tragedy ofthe commons” as individualization andregistration of land rights.

The endogeneity of Africa’s social or-ganizations to its material circum-stances is illustrated by the differencebetween lowlands and highlands. Thelowlands are commonly semiarid withhigh climatic risk and low populationdensity, whereas the highlands havemore reliable rainfall. In the lowlands,because of the greater need for insur-ance, people have invested in lineagegroups. These are used for insurance,giving their participants access to geo-graphically dispersed crops and live-stock. The dispersion reduces the co-variance of the returns on the lineagegroup’s assets. By contrast, people liv-ing in the coffee/tea areas with reliablerainfall, such as the Kikuyu in Kenya,had less need for insurance, and insteaddeveloped private property. Theyshared within the household but notwithin the lineage group.28

Thus agents in the traditional econ-omy largely succeeded in devising thesocial mechanisms necessary to cope in

an environment of high risk and fewsuitable liquid assets. These organiza-tions were costly: agents had to foregogains from specialization, had to savefor consumption smoothing, or had toabstain from accumulation in order notto attract violence. These costs can beseen as an implicit insurance premium,their high level reflecting the combina-tion of risk and the factors that pre-cluded the development of an explicit in-surance system. In this sense, stagnationcan be seen as a response to risk.

Governments became concernedabout the potential loss of output andthe stagnation which they saw as theconsequence of traditional social insti-tutions. One approach was titling andsettlement schemes, in which the gov-ernment conferred marketable propertyrights, overriding traditional rights.This was adopted in Kenya, Malawi,Côte d’Ivoire, Botswana, Cameroon,Ghana, Lesotho, Liberia, Mali, Senegal,Sierra Leone, Somalia, Sudan, Swazi-land, Uganda, and Zimbabwe (JeanEnsminger 1995). It proved ineffective.A comparison of land transactions overa 30–year period in a land registrationarea of Kenya and a communalized areaof Tanzania found that the rate of landtransactions was the same in the twocountries and unaltered over the pe-riod. In Kenya, land titling and landmarketability were largely unrelated:only a quarter of the parcels that couldbe sold by the current operator were ti-tled, and only a quarter of the parcelswith land titles could be sold. In Tanza-nia, Mali, Niger, Nigeria, and Ghana,land sales were common even when ille-gal. The land rights conferred by tradi-tional social institutions thus provedmore robust than governments hadanticipated.29

27 Bates (1983) and Baland and Platteau (1996)describe how the “tragedy of the commons” wasavoided in stateless societies.

28 See Bates (1990) and Shipton (1985).

29 See Pinckney and Kimuyu (1994) and Migot-Adholla et al. (1993).


Government belief in scale econo-mies proved unfounded: the spontane-ous subdivision of the large plots as-signed on settlement schemes turnedout to be economically efficient. Tradi-tional land rights also proved to be lesseconomically damaging and more capa-ble of evolution than had been believed.Government concern about insecurityof traditional tenure also proved un-founded. Land rights rapidly evolvedfrom allocation by chiefs to heritabilitybetween generations within the samehousehold. This secured householdproperty rights on land-specific invest-ments, although, as we will show, itactually compounded the problem ofinter-household resource allocation.Land rights also evolved towards mar-ketability. One mechanism by whichthis happens is in response to invest-ment in tree crops.30 Thus, investmentin tree crops and property rights areinterdependent, each promoting the other.

However, land rights have not yetevolved to the point where they provideefficient support for investment. Mostland in Africa is still not readily market-able. Thus, the adaptation of social in-stitutions has been too slow. If the mainmechanism for the evolution of rights istree crop investment, then the hightaxation of tree crops which has beencommon must have delayed marketabil-ity. Thus, the concern that traditionalstructures of land rights would discour-age investment and hence reducegrowth were correct. Growth was re-duced because investments were unnec-essarily illiquid. However, the solutionsattempted by governments were inef-fective, while inadvertently, throughdiscouraging tree crop investment, pub-lic policy slowed what would otherwisehave occurred spontaneously.

We now turn from the effect of therights governing the usage of factors ofproduction on investment, to their ef-fects on inter-household resource allo-cation. With population growth andheritability of land, the land-labor en-dowments of households will rapidlybecome dispersed. This will give rise toan inefficient dispersion of marginalproducts unless offset by market mech-anisms: either specialization in produc-tion according to household compara-tive advantage, or trade in factors.Specialization in production is unlikelyto be sufficient to cope with the verylarge differences in factor proportionsthat have emerged between house-holds.31 Trade in factors can be eitherby land-scarce households trading inland, or labor-scarce households hiringin labor.

Land rights have not yet evolved tothe point at which land-scarce house-holds can purchase land on a scale suffi-cient to offset the effects of differentialinheritance. However, potentially, landrental might suffice to achieve trade inland. Unfortunately, traditional tenurerights are probably less suited to evolu-tion to rental rights than to evolutioninto rights of sale. This is becausethe basis for traditional rights is usage:sale keeps usage and ownershipintact, whereas rental requires theirseparation.

Thus, rural social capital, although ithas adapted considerably in response tothe massive changes in the agriculturaleconomy, has become a constraint upongrowth in two respects. First, as herita-bility creates increasing divergence infactor endowments, the static ineffi-ciency costs mount, thus reducing thegrowth rate, both directly and (bydepressing investment) indirectly.

30 This was originally proposed by Bruce (1988)and subsequently tested econometrically by Besley(1995).

31 In Kenya the mean land-labor ratio is 11times as high in the top quintile as in the bottomquintile (Bevan et al. 1989).


Second, land-specific investmentsare insufficiently liquid and so arediscouraged. 32

We now turn from property rights toa second mechanism whereby socialcapital can affect growth, namely sociallearning. Survey evidence on the adop-tion of innovations suggests that factorendowments are less important thanaccess to information through sociallearning mechanisms.33 In turn, of theinformational mechanisms, social learn-ing is more important than either theextension service or the household’seducational endowment. Although themain rural social network is the kingroup, more modern networks also haveeconomic benefits. In Tanzania, a vari-ety of nontraditional and largely extra-kin social groups such as churches havea large payoff. An increase of one stan-dard deviation in the Putnam index ofsocial capital raises village expenditureby around a quarter; the diffusion of ag-ricultural techniques is more rapid;there is substantially more use of credit;and the quality of local public servicesis enhanced (Deepa Narayan and LantPritchett 1996).

Households in rural Africa have cho-sen to develop small, intense networks,partly because they faced low popula-tion density and high transport costs,and partly because to fulfill an insur-ance function the social networks re-quire near-complete information aboutbehavior. Social networks are not justspatially small, they also encounter bar-

riers within the village. For example,the peer groups in which social learningoccurs are sometimes segmented bygender.34 Thus, the small and intensenetworks of rural Africa may have had ahigh opportunity cost.35

There is indeed reason to believe thatthere are unexploited opportunities forlearning in rural Africa. Yields in Afri-can agriculture compare very unfavor-ably with those achieved elsewhere,even allowing for poorer soils. For ex-ample, cocoa and palm oil yields aretypically only half those recorded inAsia, and yields are also relatively lowfor livestock and for food crops. Had in-novations been well disseminated, theycould have had a large impact. How-ever, the small intense social networkswere relatively inappropriate for suchdissemination. This has been exacer-bated by weak social services. In Zam-bia, new maize varieties were developedin the 1980s that could double small-holder yields, but adoption rates re-mained low because the research wasnot carried far enough: the new tech-niques added to peak labor demand andthe new varieties did not meet farmers’preferences as consumers. Overall, agri-cultural research has not resulted inanything equivalent to the Asian greenrevolution.36

Extension services have usually beenweak. First, they have often given

32 Social networks can also restrain growth byallowing parasitic relatives to tax successful rela-tives. We are indebted to Bill Kinsey for the pointthat many of the resettlement households in hisZimbabwean panel data set gave this as theirmotive for relocating.33 For example, Bevan et al. (1989), Burger(1994), and Burger et al. (1996), show that theadoption of coffee, tea, and improved livestock inKenya is more strongly influenced by informa-tional variables than by endowments.

34 Burger et al. (1996) find that male-headedhouseholds that adopt innovations are copied onlyby other male-headed households, while female-headed households copy only other female-headedhouseholds.

35 There is some evidence that kin groupnetworks are weakening as a result of agriculturalcommercialisation. In Côte d’Ivoire, for example,“under the pressures of export crop production,Dida lineages began to fragment as brothers splitinto separate families and sons kept their wagesfor their own nuclear families rather than turn-ing them over to their fathers or the head of thelineage.” (Ensminger 1995, p. 10)

36 See Oehmke and Crawford (1996), Celis et al.(1991) and Byerlee (1993).


inappropriate advice. Extension servicesin East Africa have long been hostile to-wards the traditional practice of inter-cropping, yet research shows that thereare many advantages: maize-bean mix-tures offer better protection againstpoor rains than pure strands. Secondly,services have been badly organized withonly weak incentives for extensionagents to be productive. Governmentseven demolished extension services: inTanzania after the 1967 Arusha declara-tion, farmer training colleges were con-verted to institutes for political educa-tion, and visits by extension officers toindividual households were ended.37

Credit Constraints and Lack of Fi-nancial Depth. Rural credit markets arevery underdeveloped. In contrast toAsia, there are no specialised money-lenders, and in Eastern and SouthernAfrica inter-household credit is negli-gible. In West Africa there is more in-formal credit, but this is confined toshort-term lending.

The lack of credit is partly due to thelack of collateral. Even land titling doesnot lead to land-secured loans, in sharpcontrast to Asia.38 Potential substitutesfor collateral are either interlinkingcredit with other transactions, as inAsia, or high observability. The formeris constrained by the limited extent ofland and labor transactions, and socredit has depended upon high ob-servability. Informal loans usually occuronly where there are no informationalasymmetries: typically the lender andthe borrower live in the same villageand know each other very well.

This lack of credit has consequencesfor both consumption smoothing and

for activity choice, and hence forgrowth. A consequence of the absenceof credit is that hysteresis is important:households that have had bad luck for anumber of years will have low assetsand therefore little scope for consump-tion smoothing. For example, in West-ern Tanzania only the richer householdswere able to smooth consumption effec-tively. However, in more prosperous en-vironments most households appear notto need credit for consumption smooth-ing. A natural experiment is provided by“resettlement farmers” in Zimbabwewho received their holdings in the early1980s. In a 12–year period, this groupbuilt up a mean herd size of 10 head. Asa result, even during the 1992 drought,livestock sales were modest relative tothe herd size.39

Hysteresis has consequences forgrowth. If households are exposed to aseries of shocks, a two-class society mayemerge in which wealthy farmers haveboth higher crop incomes (since theycan afford to grow riskier crops) and abetter capacity to smooth consumption(since they have accumulated liquid as-sets). The extent to which the gainsfrom specialization in crop productionare exploited is then determined by hys-teresis. In Ethiopia and Tanzania, themore specialized households are morelikely to become nonpoor. This suggeststhat diversification has not only a levelbut also a growth rate effect. Hysteresismay be more pronounced in low-incomethan in high-income areas.40

Poor households might also be con-strained from entering higher return ac-tivities even though they are safe, if theactivities are more capital-intensive and

37 On intercropping see Fisher (1979) and onextension see Leonard (1977).

38 For example, land was used in only 3 percentof the loans in Nigeria (Udry 1993, p. 96). OnAfrican land titling, see Pinckney and Kimuyu(1994) for Kenya and Bruce (1988) and Migot-Adholla et al. (1993) for eight country studies.

39 Deaton (1990) models consumption smooth-ing under a borrowing constraint. For Tanzaniasee Dercon (1997) and for Zimbabwe see Kinseyet al. (1998).

40 See Dercon (1996, 1997) for a two-periodmodel and simulation experiments.


there are indivisibilities in investmentthat cannot be financed, such as mayapply to livestock and tree crops. Forexample, the returns to livestock aretypically high, but the cost of a singlecow may be high relative to householdincome. The combination of risk andlack of credit for investment may thusconfine poor households to low-return,capital-extensive activities, so that de-spite being more risk-averse they areless diversified. Cross-section evidencetypically shows the poorest householdsto be more specialized than middle-in-come households in the growing of foodcrops for own consumption. The im-plied dynamics have been investigatedby asking households how they wouldspend a hypothetical windfall, distin-guishing between investments thatwould simply change the scale of exist-ing activities and those that would in-troduce new activities. In Ethiopia,Tanzania, and Zimbabwe, householdswould invest substantially in livestock,suggesting that they are constrained inreaching their optimal capital stock.41

Conversely, specialization reappearsat high levels of income as entry barri-ers to the high return activities are sur-mounted. Households have less need ofthe risk-reduction that the safe, low-re-turn activities provide, partly becausethey have higher incomes and partly be-cause they have entered activities thatare themselves risk-reducing, such aswage employment and livestock. Thissuggests a non-monotonic relationshipbetween household wealth and diversi-fication. Poor households are undiversi-fied because they have not yet beenable to overcome indivisibilities in in-vestment. At a higher level of wealth,households can enter into activitiesother than food crop growing and main-

tain a diversified portfolio of economicactivities to deal with risk. At stillhigher levels of wealth, householdshave access to activities that are bothrelatively safe and remunerative(such as off-farm wage employment inKenya) or that involve the holding ofliquid assets. In the one case, incomebecomes less volatile (so that special-ization involves insurance); in theother the household relies on con-sumption smoothing rather than incomesmoothing.

The institutional solution to indivisi-bilities in investment in the absence ofa credit market is the rotating savingsand credit association (ROSCA). Theseexist in West Africa and Ethiopia buthave yet to become universal. Just asthe policy-induced low returns on agri-cultural investment have delayed theemergence of marketable propertyrights, so they might have delayed theemergence of ROSCAs.

An Interim Assessment. Three im-pediments to growth that were found atthe aggregate level also appear to beimportant at the level of the ruralhousehold: high risk, lack of social capi-tal, and poor public services. However,the risks facing rural households arelargely different from those identifiedat the aggregate level: disease and cli-mate feature most prominently, andthese are largely omitted in the aggre-gate analysis. Hence, the growth-retard-ing effects of the high risks that farmershave faced are better seen as possibleexplanations for the “Africa dummy” inthe growth regressions than as confir-mation of the negative effects of therisks there identified. Similarly, thepublic services that have failed ruralhouseholds and the lack of social capi-tal, both of which are likely to have re-tarded growth, are not those includedin the growth regressions. Education isincluded in the regressions: the best

41 See Dercon and Krishnan (1996) and Kinsey,Burger, and Gunning (1998).


that can be said is that the effects ofeducation are as ambiguous at the levelof the rural household as in the growthregressions. The lack of credit appearsmore important at the level of the ruralhousehold than suggested by its weakeffect in the aggregate growth regres-sions. However, the extent of informalrural credit is not detectable in themeasures that have been used to proxythe lack of financial depth at theaggregate level.

3.2 Manufacturing Firms

Agents in the modern economy havealso faced a very risky environment.This is partly from natural causes: cli-mate, mortality and morbidity, andcommodity prices.42 A further source ofrisk is the difficulty of enforcing con-tracts. In the now-developed econo-mies, the foundations of the moderneconomy were laid in the emergence ofsocial institutions that reduced the costsof contract enforcement and the othertransactions costs of negotiation and ofweighing and measuring. However,African firms have not been very suc-cessful in this respect: on an index ofcontract enforceability normalized onthe OECD average, Africa scores only0.66, lower than other developingcountries.43

Contract default can reflect eitherunwillingness (opportunistic behavior)or the inability of firms to adhere tocontract terms. The root cause of highdefault is that African firms are exposedto large shocks while being peculiarlyill-equipped to cope with them. Few

risks are insurable because of severeproblems of asymmetric information.For example, the financial accounts offirms are less trustworthy than in mostother regions because of the smallersize of firms and the weaker state of theaccountancy and audit professions. Fur-ther, African firms are on average verymuch smaller than elsewhere and areconsequently much less diversified. Be-ing uninsured and undiversified, theyare more exposed to shocks. In additionto unavoidable financial distress, firmsmay be physically unable to honor con-tracts. Operating in economies withpoor transport networks, firms can onlyreliably meet orders by holding largestocks of inputs and outputs.44 Whenfirms are financially distressed or stocksare insufficient, late payment or late de-livery is passed from one firm to thenext. African firms indeed hold verylarge stocks: they are far removed from“just in time” production systems.45

Until recently, research on Africanfirms was hampered by a lack of com-parable survey data. A major change hasbeen the Regional Program on Enter-prise Development, in which panel datafor a sample of approximately 200manufacturing firms in each of eight Af-rican countries were collected over theperiod 1992–96 under the coordinationof the World Bank. This forms the basisfor much of our evidence on firms. Twostylised facts from the RPED surveysare that by international standards,firms are small and slow growing. Theannual gross investment rate was only 6percent of the value of the capital stockand amounted to only 11 percent ofvalue added, suggesting that net invest-ment was negative. Technical efficiency

42 Drought causes problems for firms because itcauses coordinated declines in expenditure and sodemand shocks. For example, during and immedi-ately after the Southern African drought of1991/92, industrial output declined in Zimbabweby 23 percent. Survey data for industrial workerssuggest a high rate of illness-related absenteeism,and AIDS-related deaths have been rising rapidly.

43 Derived from Knack and Keefer (1995).

44 Fafchamps, Gunning, and Oostendorp (1998)present evidence for Zimbabwe that large stocksare a response to contractual risk.

45 On contract enforcement see Fafchamps(1996b).


as measured by the mean distance fromfrontier production functions was verylow, and labor productivity appears tobe falling.46

We now group this survey evidenceinto five mechanisms by which growthmight have been affected, considering:whether high risks have reduced invest-ment; whether restrictions upon inter-national trade have affected growth; theextent to which contract enforcementrestricts activity; the extent to whichpoor infrastructure raises costs; and theextent to which lack of external financeconstrains investment.

Consequences of a High Risk Envi-ronment. Investors rate Africa as highlyrisky, and we have seen that Africaneconomies are subject to a high degreeof volatility. Firms are faced with aforecasting task for which they have noaccumulated expertise. This translatesinto a high degree of price risk at thelevel of the firm and causes allocativeinefficiency as firms optimize againstdifferent expected prices. For example,even in Zimbabwe, a country with rela-tively sophisticated financial informa-tion, a survey of exchange rate expecta-tions found that while 21 percent offirms expected a devaluation of at least20 percent within a year, 27 percentexpected appreciation or stability.47

Such risks are more important be-cause investment in Africa is more diffi-cult to reverse than is the case else-where. One reason for this is thatequipment once purchased is less read-ily saleable, since markets in second-hand capital are weak. This weakness isrevealed both in the quantity of transac-

tions and in their price. Very few firmsin the RPED surveys use secondhandequipment. Where the market is used,equipment sells at a deep discount. Theweakness of secondhand markets forcapital goods reflects the high degree ofoligopoly characterizing most Africanmanufacturing due to protection andlicensing.48

A second reason for irreversibility isthat the market in firms as going con-cerns is also very limited. This is due toa combination of lack of finance and se-vere problems of asymmetric informa-tion in the absence of reliable audits. Asa result, a large majority of Africanfirms do not survive their owners.

These irreversibilities make Africaninvestment very illiquid, whereas theprice risks associated with recent andinsecure liberalizations create a pre-mium on liquidity. This has discouragedinvestment. There is evidence forGhana that the effect of uncertainty onthe rate of investment is negative andthat this effect is considerably strongerfor firms facing irreversible investmentdecisions (Catherine Pattillo 1998).

Firms also faced unreliability in theirsupplies from other domestic firms. InZimbabwe, firms responded by carryinglarge stocks, on average three months’worth of supplies. Firms also respond torisks by using state-contingent con-tracts. The period for the repayment oftrade credit, while being agreed before-hand, is frequently renegotiated: inKenya more than half of trade creditswere extended, and delaying paymentswas the most common form of dealingwith unexpected liquidity shocks.49

Lack of Openness: Regulation andTaxation. The aggregate level evidence46 An indication of the dearth of large firms is

that the largest of the four size groups with whichthe surveys work starts at only 100 employees. Oninvestment and efficiency see Bigsten et al. (1998,forthcoming).

47 Calculated by the authors from the ZimbabweRPED survey.

48 In Zimbabwe new capital goods sell at a dis-count of around 50 percent (Gunning and Pomp1995).

49 Biggs, Raturi, and Srivastava (1996) and Biggsand Srivastava (1996).


has identified a lack of openness as thesingle most important cause of slowgrowth. At the level of the firm there isevidence of the dramatic effects offoreign exchange controls and licensing.

First, tax rates have often been firmspecific. In Cameroon, by 1993 firmsenjoying one or more special tax re-gimes accounted for 99 percent of totalsales. Second, controls have drasticallyreduced competition. In Zimbabwe—the country with the longest history ofcontinuous restrictions—by the end ofthe control period only one-third of thefirms considered the price of theirproducts as market determined. Com-petition from imports was rated as amajor problem by only 5 percent offirms, and domestic competition byonly 2 percent. Licencing and foreignexchange controls were importantconstraints at the firm level.50

In Section 2 we reviewed aggregate-level evidence on the lack of openness.Firm-level evidence suggests that thishas depressed investment. The expecta-tions of both trade liberalization and ex-change rate depreciation were found tobe significant determinants of Zimbab-wean manufacturing investment.51 Inextreme cases, such as Tanzania, for-eign exchange controls interacting withprice controls changed the very natureof firm activity from manufacturing tohoarding. Firms that anticipated liber-alization had an incentive to importwhatever inputs were permitted, while

selling as few of them (incorporatedinto output) as possible. However, in-evitably, the firm-level evidence under-states the growth-reducing effects ofthe control regime. Firms are unlikelyto perceive constraints on activities inwhich they are not engaged, so thatthose activities most damaged bycontrols will be unrepresented.

At the aggregate level, there is evi-dence that in some countries during the1990s there has been substantial liber-alization of these control regimes. Thefirm-level evidence bears this out. Tak-ing the controls cited above, by 1995 inCameroon, the proportion of firms re-ceiving special tax treatment had fallento 30 percent. In Zimbabwe the propor-tion of firms finding competition aproblem had risen to 37 percent, andinputs and spares could be readilyimported.

Lack of Social Capital: Contract En-forcement and Learning. Social capitalcontributes both to contract enforce-ment and to social learning. There isevidence that both functions are impor-tant for African manufacturing firms.Recall that firms face high risks of con-tract default. Many of these breachesare unavoidable, but often they reflectweak social enforcement mechanisms.One reason for a high rate of opportun-ism is that the African courts generallywork less reliably than those elsewhere.Only about a quarter of African lawyersconsider the judiciary fully independentof the executive. The legal processoften involves long delays, and most ju-dicial officers appear to be only moder-ately knowledgeable about the law. Af-rican courts therefore often fail as aninstrument of reliable contract enforce-ment. As a result, in most African coun-tries the courts are little used for con-flict resolution. In most Africancountries a lawyer is hired in onlyaround 10 percent of disputes, whereas

50 See Navaretti et al. (1996) on Cameroon andFree University, Amsterdam and University ofZimbabwe (1993) and Biggs, Raturi, and Srivas-tava (1996) on Zimbabwe and Kenya.

51 This is shown in Gunning and Oostendorp(1996). Roberts and Tybout (1997) provide possi-ble microeconomic foundations for the result.They explain the decision to enter export marketsin an intertemporal optimization model in whichexporting involves fixed start-up costs. An increasein expected future profitability raises the optionvalue of investing in the ability to export in thecurrent period.


in Zimbabwe, where the courts aresomewhat more reliable, the proportionis around 30 percent.52

The high rate of unavoidable defaultitself facilitates opportunistic behavior,since excuses become more credible un-less victims invest in further informa-tion to distinguish between avoidableand unavoidable defaults. However,were victims to assume that all default-ers were opportunistic, they would rap-idly lose business opportunities. Firmstherefore continue to do business with ahigh proportion of defaulters; in Ghanaover 90 percent of defaulters are for-given. Three-quarters of defaults wereattributed to inability to pay, and inthese cases the supplier was effectivelysharing risk with his customer. Fre-quently this is reflected in incomplete(in fact state-contingent) contracts, nopayment date being specified. Similarly,in Zimbabwe very few of the reportedsupply problems occur with new sup-pliers: where problems occurred thebusiness relation had an average dura-tion of over fifteen years. That supplyproblems are largely unavoidable ratherthan opportunistic is supported by thefact that in 90 percent of the cases thebusiness relation continued.53

Because of the small scale of the for-mal sector, credit-rating agencies havenot developed. Hence firms are forcedinto the more expensive process ofgathering for themselves the informa-

tion needed to distinguish betweenavoidable and opportunistic defaults.The method they use for this is thesocial network. 54

In the absence of adequate state-provided enforcement mechanisms, theappropriate social institution is the kingroup. Manufacturing firms, just asfarm households, have adopted this asthe basis for their social networks.However, in the urban economy thisplaces an ethnic minority kin group at aconsiderable advantage since all itsmembers are confined to urban privatesector activities.55 By contrast, most ofthe members of an African kin group ofthe same size will live in rural areas,and most of those in urban areas will beemployed in the public sector. Sincefew members of the African kin groupwill own manufacturing firms, thegroup is likely to be too small to formthe basis for an entrepreneurial socialnetwork. In Kenya, Asian-owned firmsare at an advantage over African-ownedfirms. To screen new applicants fortrade credit, African-owned firms relyoverwhelmingly on direct observationof the candidate’s premises, whereasAsian-owned firms rely on “askingaround,” suggesting strong informationnetworks. As a consequence, Asian-owned firms have much better access to

52 The underlying data for the judicial reliabilityindex were gathered by Business International.Mauro (1995) utilized them, and we would like tothank him for making his data available to us. Onthe opinions and competence of African lawyerssee Widner (1998). On the use of the courts, seeHEC (1993) for Cameroon; Fafchamps (1996b)for Ghana; University of Leuven and University ofBurundi (1994) for Burundi; University ofGothenburg and University of Nairobi (1994) forKenya; Fehr, Forsund, and Kittelsen (1994) forZambia; and Gunning (1994) for Zimbabwe.

53 See Fafchamps (1996b) and Free University,Amsterdam and University of Zimbabwe (1993).

54 Social networks are potentially useful for twodistinct purposes, learning about the trustworthi-ness of other firms and learning about new tech-niques and market opportunities. Since networksappropriate for the former are likely to be smallerand more homogeneous than those appropriate forthe latter, the size of the network is restricted bythe need for close observation of business partners(Barr 1996). The cost of this is likely to be thatsocial networks are to small to be effective chan-nels for social learning, e.g. about export markets.If a lack of foreign social networks is an importantconstraint, then foreign-owned firms should out-perform domestic firms in export markets, and thisis indeed the case (Bigsten et al. 1998).

55 In many countries the colonial authoritiesmade it illegal for Asians to acquire agriculturalland.


trade credit: small Asian-owned firmshad almost the same access to credit aslarge firms, whereas small African-owned firms were disadvantaged. Simi-larly, 80 percent of Asian-owned firmswere able to share risks through state-contingent repayments, in contrast toonly 33 percent of African-ownedfirms.56

To conclude, firms must survive in ahigh-risk environment. Social networkscan facilitate this both through reducingthe risks of opportunism and by provid-ing partial insurance through state-contingent contracts. However, the kin-group based network, which is the mostsuitable basis for enforcement, is not anoption for most African-owned firms.Further, compared with state-providedenforcement mechanisms, it imposescosts. It restricts business to the smallgroup of firms known to the network.

Poor Public Services. In Section 2 wedescribed various aspects of service pro-vision, and it is evident that many ofthese will disproportionately impingeon manufacturing. This is borne outby comparable firm-level evidence forseven countries in which firms wereasked to score problems among fifteenidentified choices. Two of the fifteenrelated to infrastructure: a general cate-gory “lack of infrastructure,” and “highutility prices.” Aggregating across theseven countries, these ranked as thefourth and fifth problems respectively,after “lack of credit,” “lack of de-mand,”’and “high taxes” (Tyler Biggsand Pradeep Srivastava 1996).

Turning to specific components of in-frastructure, electricity supply is unreli-able, and so firms must rely on privategenerators. In Nigeria 78 percent offirms have standby generators, the highfixed costs of which bear particularly on

small firms, accounting for a quarter ofthe value of their equipment (Kyu SikLee and Alex Anas 1991). Inadequatetelephone services were identified by47 percent of Zimbabwean firms astheir most serious problem. One exportfirm in the survey had to make a30–mile journey in order to make atelephone call.

Recall that an important unresolveddispute in the aggregate growth regres-sions is the effects of the low level of,but rapid increase in, education. A com-parison of the rate of return to humanand physical capital in African manufac-turing across five countries finds thatthe returns to human capital are sys-tematically much lower than to physicalcapital (Arne Bigsten et al. 1998).Hence, Africa’s low endowment of hu-man capital may not have been an im-portant cause of its slow growth in thecontext of other impediments.

Lack of Financial Depth: Investmentand Finance. Formal financial marketsare much more limited in Africa than inmost other developing regions. How-ever, it does not necessarily follow thatthis has been an important constraintupon investment.

There is fragmentary evidence that infirms without internal funds, the lackof credit constrains investment. First,there is econometric evidence forGhana and Zimbabwe that firms report-ing “access to credit” as one of theirmajor problems are less likely to invest.Second, firms make little use of creditto finance their investments. In Zim-babwe and Kenya, firms were asked forthe sources of finance for the last majoracquisition of equipment. Both thesecountries have well-developed financialmarkets by African standards. Never-theless, in Zimbabwe, bank loans ac-counted for less than 2 percent of thetotal value of the investment, and inKenya only 15 percent. Third, there are

56 See Biggs et al. (1996) and Biggs and Srivas-tava (1996).


systematic differences in investmentrates between firms that might plausiblybe related to differential access tocredit: investment is much higher inlarge firms, and pooled panel data indi-cate that this is because large firmstend to invest more rather than thatfirms with high investment rates havegrown large. Fourth, investment is sys-tematically related to changes in profit-ability: there is a significant positiveeffect of innovations in profitabilityon investment rates, suggesting thatchanges in the liquidity of the firmmatter and hence that investment isliquidity constrained.57

However, investment is very likely tobe more constrained by factors otherthan finance. Although profitability issignificant for investment, the effectis quite weak: a doubling of profitswould raise the investment rate from6 percent only to 8 percent.

An Interim Assessment. The con-straints on firm growth are similar tothose that we have identified for ruralhouseholds: high risk, poor public ser-vices, and lack of social capital, and thissupports the aggregate-level evidence.In one respect, the firm-level evidencematches better than the household evi-dence with that at the aggregate level;namely, the lack of credit, though verypronounced, does not appear to be adominant constraint upon investment.

Firms face high risks and have a lim-ited capacity to bear them. Risks are in-flated by the lack of publicly providedcontract enforcement, by poor infra-structure, and by the vagaries of themacroeconomic environment, includingpolicy. Their investments are difficult

to reverse, largely because of trade andlicencing restrictions. Their strategiesfor responding to this environment havebeen partly to reduce the risks andpartly to accommodate them. To reducerisks, firms hold large inventories toguard against unreliable suppliers, de-vote a substantial share of investmentto generators to guard against publicpower interruptions, and restrict busi-ness relationships to those firms theyknow well. To accommodate risk, firmsreduce investment and enter into state-contingent contracts. Both strategiesimpose costs. The recent liberaliza-tions are reducing some of these costs.However, they are also reducing theprotection from which existing firmsbenefited.

4. Factor and Product Markets

We now view the constraints ongrowth from the perspective of mark-ets. We consider financial, labor, andproduct markets.

4.1 Financial Markets

Africa might be expected to havehigh demand for the services of finan-cial markets. First, as a high-riskenvironment there is a need for risk-bearing: portfolio diversification, con-sumption-smoothing, and insurance.Further, much of the African capitalstock is unusually long-lasting, such astree crops and mines, or lacks marketsfor used equipment. Hence, we wouldexpect to find securitization in order toincrease liquidity. Third, because of thelarge relative price changes resultingfrom policy reform and shocks, currentprofitable investment opportunities willonly be weakly correlated with pastprofits, leaving scope for intermedia-tion. These demands have remainedlargely latent. Until recently, this couldbe explained by policies of financial

57 See Gunning and Pomp (1995) and Bigsten etal. (1996) on access to credit; Free University andUniversity of Zimbabwe (1995) and University ofGothenburg and University of Nairobi (1994) onbank loans; and Bigsten et al. (forthcoming) oninvestment and profitability.


repression. During the 1990s there hasbeen widespread financial liberaliza-tion, but the response has as yet beenlimited. We discuss these phases inturn.

Financial Repression and its Costs.While there are large potential gainsfrom financial transactions, in Africathey are costly. There are few assetsand so little collateral. Yet there is alsoa shortage of firms with creditworthyreputations. The costs of financial infor-mation about firms are unusually high:there is often no financial press,audited accounts are unreliable (due tocorruption in the professions), andthere are neither risk-rating agencies,nor agencies that purchase invoices.Hence, a substantial need encountersa high cost of meeting it. MarcelFafchamps (1996) argues that the highrisks intrinsic to enterprise activity inshock-prone economies with low in-comes imply a high level of contract re-negotiation. In financial transactionsthis would manifest itself either as re-scheduling or risk-sharing. We have al-ready discussed the evidence for thisbehavior both between householdsand between firms. However, this de-pends upon the high observation pro-vided by kin group membership, and sois not an option for formal financialinstitutions.

These natural high costs of the finan-cial sector were compounded by policy-generated increases in costs. Implicittaxation through financial repressionranged from 2–2.5 percent of GDPfor Côte d’Ivoire, Zaire, and Kenya,through 4–7 percent for Ghana, Nige-ria, and Zambia, to 19 percent in Zim-babwe. Taxation through unremuner-ated reserve requirements ranges from1.5 percent in Tanzania to 7.5 percentin Ethiopia. These sums are large in re-lationship to the size of the bankingsector, in all cases exceeding the value

added of the banks. Banking is evenmore heavily taxed than exports.58

Costs were also increased due to thepoor performance of the legal system,which has reduced the efficacy of as-sets as collateral. A survey of banksin Uganda (Louis Kasekende and M.Atingi-Ego 1997) found that theirmost desired reform was a fast-trackprocedure for loan recovery cases.

Many banks were nationalized.Where private banks survived they wereoligopolies. Lending practices concen-trated on risk minimization, with banksusing central bank allocations of foreignexchange to firms as the main signal ofcreditworthiness (Mthuli Ncube et al.1998). The public banks provided creditto the government and loss-makingparastatals. They became an off-budgetchannel for government expenditure.For example, by the late 1980s the mo-nopoly government bank in Tanzaniahad lent over 60 percent of its portfolioto a single crop-marketing parastatal tocover operating losses (Collier andGunning 1991a). Additionally, the lend-ing practices of the managers of pub-licly owned banks were inadequately su-pervised, permitting corrupt lending. Asa result of these two types of lending,publicly owned banks have had veryhigh default rates, typically in the 40–95 percent range. Clearly, at these ratesof default the repression of interestrates is a secondary phenomenon. Gov-ernments have been unable or unwillingto control the lending practices of bankmanagers: the Uganda CommercialBank serves as an illustration. By theearly 1990s the majority of the bank’sloans had become nonperforming, andmany of the largest were to its ownmanagers. In response, the governmentestablished a trust with greater legal

58 See Chamley and Honohan (1990), Giovan-nini and de Melo (1993), and Ikhide (1992).


powers of recovery to which, after longdelays on the part of the UCB manag-ers, the largest nonperforming loanswere transferred. The trust offered a 15percent discount for prompt repaymentand threatened asset foreclosure. Themanagers of the UCB repaid theirloans to the trust by means of grantingthemselves new loans from UCB.

The conjunction of naturally high op-erating costs, the diversion of the publicsector banks to other objectives, andthe oligopolistic nature of the remain-ing private banks severely curtailed thedevelopment of the financial sector.Spreads were approximately doublethose in developed economies (DierySeck and Yasim H. el Nil 1993). Thelack of formal financial markets hasposed more of a problem for firmsthan for households, since Africa hasunusually diversified households andunusually undiversified firms.

The high costs of the formal financialsector shifted activity to the informalsector and to trade credit. The informalfinancial sector was not able to providean effective substitute for the formalsector. Formal and informal financialinstitutions serve distinct and largelyunconnected market niches. The formalsector has not been able to channellending through informal intermediar-ies. The RPED surveys show that firmsmake little use of informal finance, inmarked contrast to Asia. The lack of in-formal credit is particularly striking forstart-up finance: new firms are almostentirely financed from the founder’sown savings. Trade credit has been animportant source of finance. However,flows to small firms have been too smallto provide a substitute for direct accessto the credit market.

Short-Term Problems of Financial Re-form. During the 1990s there was wide-spread financial liberalization: interestrate deregulation, the entry of new

banks, and the lifting of directed lend-ing. However, these have yet to pro-duce noticeable financial deepening.Partly, costs are high in the financialsector predominantly for reasons otherthan policy. Additionally, the legacy offinancial repression has been weakbanking organizations that are unable toexploit the opportunities opened up byliberalization. In the short term, finan-cial liberalization in such an environ-ment creates policy problems. The leg-acy of weak financial organizations isexacerbated because, as more generallywith policy reform, the countries withthe strongest financial reforms had hadthe most severe phase of repression.For example, Ghana and Uganda hadboth had prolonged phases of financial“shallowing.” Thus, the strong reform-ing African countries were commonlystarting with banks that had needed nei-ther an information base for the assess-ment of lending opportunities, nor alegal process for the recovery of loans.

In the long run, financial reform canbe expected to improve the allocation ofinvestible funds. In the short term,however, firms with growth potentialmay even have faced increased financ-ing difficulties. Although financial liber-alization enabled banks to charge a riskpremium, other aspects of liberalizationremoved the main signals of credit-worthiness. Foreign exchange liberaliza-tion ended reliance upon foreignexchange allocations, and trade liberal-ization made past profitability an un-reliable guide to future profitability.Hence, in the first years of financial lib-eralization, which is all that Africa hasexperienced, banks lacked the informa-tion and legal security for expandedprivate lending.

While the banks faced difficulties inlending profitably to the private sector,liberalization presented them with moreattractive alternatives. As governments


lost the revenue implicit in financial re-pression, they borrowed on treasury billmarkets, often at very high real interestrates. The composition of bank portfo-lios shifted toward government stocks.Flow of funds analysis reveals that, todate, far from liberalization channellingresources to the private sector, govern-ment borrowing has caused a creditsqueeze. Further, liberalization has notyet altered the lack of intermediationbetween the formal and informal fi-nancial sectors (Machiko Nissanke andErnest Aryeetey 1998).

In the long run, the increased compe-tition provided by the entry of newbanks can be expected to lower bankingcosts. However, in the short run it hascreated two problems. The number ofbanks increased very rapidly in Nigeria,for example, from under ten to 119.The regulatory function of centralbanks was underdeveloped, because thetraditional private commercial bankswere on a sounder footing than the cen-tral bank, while the public commercialbanks were not intended to have asound lending portfolio. Supervision ofnew banks was inadequate, enablingthem to lend to their directors and runPonzi schemes.

In the long run, interbank markets inforeign exchange can be expected toprovide stabilizing speculation and for-ward markets on which remaining riskscan be hedged, but in the short runmost foreign exchange transactions haveremained cash-based in bureaux dechange. The scale of transactions is toosmall to permit normal bank supervisionand control procedures, so that bothbanks and corporate treasury depart-ments are reluctant to take positions.There are no forward markets. Thus,Africa has the institutions for a liberal-ized foreign exchange market, but notyet the specialist organizations withwhich to run one in a way that enables

international traders to shift currencyrisks at low cost.

While in the long run stock marketswill increase the liquidity of the Africancapital stock and enable risk pooling inportfolios, at present the total capitali-zation of Africa’s 16 markets is tiny, andlittle money has been channelled tofirms. This reflects weaknesses in boththe public legal services and the privateprofessions such as auditing needed tosupport a stock market.

While in the long run African govern-ments will be able to borrow cost-effectively on world markets, at presenttheir commercial borrowings are sub-ject to a considerable risk premium.The premium is strongly and signifi-cantly influenced by export price shocksand civil war. Hence, the underlyingcauses of risk in Africa rebound onto fi-nancial markets rather than becomemitigated by them (M.O. Odedokun1996).

In one dimension, however, Africa isalready fully integrated into global fi-nancial markets: Africa has experiencedmore capital flight as a proportion ofprivate wealth than any other region.Table 8 compares the portfolio choicesof private wealth owners across regionsby combining data on capital flightsince 1970 and the domestic privatecapital stock (excluding housing andforeign-owned capital) to construct anestimate of regionally owned wealth andits disposition as of 1990. It is not sur-prising that in a wealth-abundant, capi-tal-hostile environment like the MiddleEast, wealth owners have placed a highproportion of their wealth outside theregion. By contrast, both South andEast Asia have a low proportion of theirwealth outside the region. The choicesof African wealth owners are startling.Despite a lower level of wealth perworker than any other region, Africanwealth owners have chosen to locate 39


percent of their portfolios outside Africa.An Interim Assessment. Clearly, de-

spite the high needs for financial mar-kets, financial markets have been se-verely underdeveloped. This is partlydue to repressive policies toward the fi-nancial sector and partly to high costsreflecting asymmetric information.However, at present it is not possible todetermine the relative importance ofthese two causes. While the lack of re-sponse to the recent financial liberaliza-tions might appear to suggest that un-derdevelopment of the financial sectoris deep-rooted, the phase of poor policymight have left a legacy of financial or-ganizations that are too weak to take upthe opportunities of liberalization.

4.2 Labor Markets

There is some support for rival vi-sions of the formal labor market. Inone, it is the source of many problems:wage levels are too high; there is exces-sive dispersion, with firms facing wagesthat rise as the firm grows; and wagesare insensitive to profit shocks, so thatfirms hold back on hiring. In the othervision, labor markets are flexible andconstitute Africa’s comparative advan-tage. Wages respond to shocks, accom-modating the inadequacies of finan-cial markets through state-contingentcontracts; although wage employmentis limited, this reflects impediments

to firm growth rather than problemsinherent in the labor market.

The Rigid Labor Market View. Thebest-known models to come out ofAfrica are those of John R. Harris andM.P. Todaro (1970) and Joseph E.Stiglitz (1974). Harris and Todaro pos-tulated that wages were institutionallyrigid at above the supply price of labor.As a result, the labor force migrated tocities to participate in the jobs lotteryconstituted by unemployment. Stiglitzrefined this by explaining high urbanwages not as an institutional given butas an equilibrium outcome of efficiencywage setting.

We now turn to the evidence for la-bor market rigidities. In a few coun-tries, labor legislation was until recentlyextremely rigid. For example, until1991 all dismissals required the ap-proval of government in Zimbabwe andof workers’ councils in Ethiopia. How-ever, in all seven of the countries cov-ered by the recent RPED surveys, laborregulations were ranked at or near thebottom of the 15 obstacles considered.The main focus of attention for thosewho regard the labor market as rigidhas therefore been on wages: whetherthe public sector pays above the supplyprice and whether minimum wage legis-lation raises entry wage levels in theprivate sector. In Francophone Africaand Ethiopia, public sector wages are

TABLE 8PORTFOLIO COMPOSITION AND FACTOR PROPORTIONS BY REGION 1990

Capital Flight/Private Wealth Private Capital Stock per Worker ($)

Sub-Saharan Africa 0.39 1,069South Asia 0.03 2,425East Asia 0.06 9,711Latin America 0.10 17,424Middle East 0.39 3,678

Source: Collier, Höffler, and Pattillo (1998), based on data for 43 countries of which 22 are African.


well above the supply price, especiallyfor unskilled labor. This also used to bethe case in Anglophone Africa, but in-flation has commonly eroded real wagesto levels well below those in the privatesector. Indeed, the low level of publicsector pay is now often seen as a policyproblem. Similarly, during the 1960sminimum wage legislation generallyraised entry wages in the private sector.Inflation and nonenforcement havegradually eroded this effect. Concernhas therefore shifted from excessiveentry-level wages to excessive wage dis-persion. This has been interpreted asreflecting rigidities due either to effi-ciency wages or rent-sharing. Wages arepositively correlated with firm size: forexample, Ann D. Velenchik (1997) findsa differential of over 4:1 between largeand small firms in Zimbabwe. In sup-port of the efficiency wage interpreta-tion, the returns to experience with thecurrent employer differ by firm size;workers in larger firms experiencesteeper wage growth and more rapidpromotion. Hence, reducing turnover ismore important for larger firms, and somay explain a wage premium.

Theories of rent sharing see the wagepremium as the result of above-averageprofits: in a noncompetitive labor mar-ket, workers may induce managementto share monopoly rents with them. Al-though African workers rarely have bar-gaining power, the real bargain may bebetween firms and governments: thegovernment generates rents for firmson the implicit understanding that theseare partly used to pay a wage premium.Since larger firms are more exposed togovernment, workers in larger firmswould better be able to exact a pre-mium. Velenchik tests this on cross-section data. There is Ghanaian andZimbabwean evidence for rent sharing:controlling for other firm and workercharacteristics, profits per worker are

highly significant in both cross-sectionand panel earnings functions. This hasbeen interpreted as showing that thestructure of wages is not competitive.However, rent-sharing is not due tounionization: on cross-section data,unionization is insignificant in the meanwages firms pay.59

The Flexible Labor Market View. TheHarris-Todaro view of wages as exoge-nous and unemployment as endogenousreversed the characterization of devel-oped-country labor markets proposedby David G. Blanchflower and AndrewJ. Oswald (1995). In the latter, wagesare endogenous to exogenously gener-ated unemployment, which varies be-cause of unobserved differences in theattractiveness of locations. According tothe former, the correlation between un-employment and wages will be positive;according to the latter it will be nega-tive. John Hoddinott (1996) used re-gional variations in unemployment andwage rates in Côte d’Ivoire to test theHarris-Todaro model against the “wagecurve” thesis of Blanchflower andOswald. He showed that the Ivorian ur-ban labor market conformed closely tothe Blanchflower-Oswald results fordeveloped countries: wages were infact slightly more responsive than indeveloped economies.

The wage premia hypothesized in theHarris-Todaro model were common atthe time it was formulated but provedto be temporary. They built up duringthe transition to independence whenunions had temporary political powerand when colonial authorities neededto give wage earners an interest inthe status quo. Usually, these premiaproved unsustainable after indepen-dence and were eroded by inflation.However, wages were flexible even inlow-inflation environments. Hoddinott’s

59 See Velenchik (1997) and Teal (1996).


results cannot be due to an inflation ef-fect since Côte d’Ivoire was subject tothe rules of the Franc Zone. VictorLevy and John L. Newman (1989)showed that real wages were neverthe-less remarkably flexible in Côted’Ivoire. They used panel data to esti-mate how the earnings function hadshifted during the recession of the1980s. They showed that mean earningsfigures, which appeared to show rigid-ity, were highly misleading becausenewly hired and therefore low-wageworkers were the first to lose their jobs.In fact, the earnings function hadshifted down by around 25 percent inonly four years.

Such urban wage adjustments werewidespread. Vali Jamal and John Weeks(1993) suggest that, for several Africancountries, the urban wage premiumover rural incomes had actually been re-versed during the 1980s. The time pathsof wages in Zimbabwe and South Africaare consistent with the premia beingtemporary during the transition to blackrule: real wages in Zimbabwe havefallen sharply from the mid-1980s, andcurrently much the largest wage pre-mium on the continent is that in SouthAfrica.

The Harris-Todaro model may noteven have been an accurate explanationof the unemployment that inspired it.Kenyan unemployment in the late 1960smay have been a response to rapidchanges in the supply of educated labor,as job seekers learned only with a lagthe extent to which selection criteriawere changing. Analyzing unemploy-ment rates among school leavers by ex-amination grade, Collier and DeepakLal (1986) showed how, over a period ofsix years, the peak unemployment rateshifted up the grade structure as thosewith poor grades realized that it wasnecessary to lower their reservationwages.

African labor markets are responsivemechanisms for the pricing of skills.Cross-section studies show that the re-turns to education reflect cognitiveskills rather than credentialism (M.Boissière, John Knight, and RichardSabot 1985, and Knight and Sabot 1990,for Kenya and Tanzania; Collier andGarg 1998 for Ghana). The comparisonof earnings functions and productionfunctions shows that workers earn theirmarginal products (Jones 1994). At thenational level, changes in the supply ofeducated labor change its price (Knightand Sabot 1990, and Simon Appleton etal. 1996).

Reconciling the Evidence. The evi-dence is conflicting: real wages havefallen steeply over time, and skills arecorrectly priced, yet there appear to belarge premia paid according to firmsize, unionization, and profitability.However, these three premia need notindicate imperfections.

Francis Teal (1996) stresses that OLSestimates of earnings functions may bemisleading because of the endogeneityof the profits variable. Allowing for thisin a 2SLS estimate, he finds that thefirm size variable is no longer signifi-cant: larger firms are more profitable,but firm size does not have an indepen-dent effect on wages. The focus in theliterature on the wage premium by firmsize may therefore have been mis-placed. Similarly, it is implausible that(other than in South Africa) unions canextract rents from firms. Teal testswhether unions influence the degree ofrent sharing and finds, contrary to someresults for developed countries, that inGhana it is unaffected by whether thefirm is unionized. Finally, although oncross-section evidence more profitablefirms pay higher wages, this mightreflect risk sharing rather than rentsharing. Worker-firm contracts mightbe state-contingent, akin to the


inter-household contracts found byChristopher Udry in Nigeria and theinter-firm contracts found by Faf-champs in Ghana. Such a sharing ofrisks between firms and workers may berational. Although wage rigidity in de-veloped countries has been explained asan efficient apportionment of risk-bear-ing between risk-neutral firms and risk-averse households, African firms are toosmall and have insufficient access to fi-nance to be risk neutral. Risk sharingcan only be distinguished from rentsharing using panel data. With risk shar-ing, wages will be more sensitive tochanges in profits than to the level ofprofits, whereas with rent sharing theconverse will hold. Teal found thatwages were three times more sensitiveto changes in profits than to the level ofprofits. Similarly, Velenchik finds thatchanges in profitability are a significantdeterminant of wage changes. Thus, therelationship between wages and profit-ability may be a further manifestationof the conduct of business in a shock-prone environment without adequatefinancial markets, rather than a puz-zling counterexample to the weightof evidence pointing to labor marketflexibility.

However, African labor markets arehighly distinctive in one important re-spect, namely their extent. Only a smallproportion of the labor force is allo-cated through the formal market, andthis proportion is declining: from 12percent in 1980 to 9 percent in 1990(Dipak Mazumdar 1994). Globally,there is a strong relationship betweenthe share of the labor force in nonagri-cultural wage employment and GDPper worker (World Bank 1995). Whilemuch of the correlation will not becaused by labor markets, presumablythey facilitate growth through special-ization and the division of labor, andhence skill formation, as well as reaping

the static gains of allocative efficiency.On our interpretation, the small size ofthe formal labor market, while a prob-lem for the growth process, is not dueto its malfunctioning. Rather, it reflectsthe constraints facing firms: high risks,poor infrastructure, and lack of socialcapital. Indeed, the evidence for risksharing in the labor market suggeststhat the problems that on other conti-nents are solved by financial marketsare in Africa partially resolved in thelabor market.

4.3 Product Markets

Many African governments were sus-picious of product markets as beingcontrolled by ethnic minorities. Theysought to Africanize them through stateownership and control, and to use themas instruments both to transfer incometo favored urban groups and as theirmain source of revenue. Product mar-kets were undermined through restric-tions on traders, high taxation, poor in-frastructure, and controls on prices andquantities.

First, some governments banned pri-vate traders. This policy was particu-larly applied in agricultural markets.Public monopsonies were created forexport crops. Sometimes there was seg-mented monopsony, a public marketingagency dealing with locally monopsonis-tic producer cooperatives. Some compe-tition existed between such monopso-nies, as farmers could smuggle crops toareas where higher prices were paid.Sometimes the monopsony was onlynominal, due to extensive illegal orinformal competition. This is typicalof food markets. Governments wereparticularly keen to ban interdistrictagricultural trade. For example, eventhe relatively market-friendly Kenyangovernment gave a monopoly of in-terdistrict trade in staple food to theMaize and Produce Board. The less


market-friendly Tanzanian governmentclosed all private village shops; privatetrading in urban areas was restricted; alldomestic nonagricultural trade was sup-posedly handled by the Board of Inter-nal Trade; and all international tradewas the monopoly of the Board of Ex-ternal Trade. The market-hostile Ango-lan and Ethiopian governments nation-alized virtually all trade (Jean-PaulAzam, Collier, and Andrea Cravinho1994).

Second, governments imposed hightaxation on international trade and be-came highly reliant upon it for revenue.More important than explicit exporttaxes were implicit taxes through ex-change rate overvaluation and widemargins for marketing parastatals. Forexample, in Tanzania the producerprice of cotton would have been 27 per-cent higher on average in the absenceof export taxation but 60 percent higherif the other two factors are taken intoaccount (Stefan Dercon 1993). By 1984coffee farmers were receiving only 40percent of the world price, comparedwith over 80 percent in Kenya, wherethe political elite had interests in coffeefarming. Despite the high tax revenue,governments restricted spending ontransport infrastructure. As a result, formany crops transport costs exceededproduction costs.

Third, governments introduced con-trols on prices and quantities. One ra-tionale was to reduce food prices to ur-ban consumers. For example, in Zaire,when price controls were abandoned in1982, real producer prices doubled forrice and trebled for maize and cassava(Tsishimbi wa Bilenga 1980). However,many governments applied price con-trols more widely to manufacturingfirms, agricultural producers, and con-sumers. In Tanzania the principle ofcost-plus pricing was extended frommanufacturing to smallholder agricul-

ture. Crop prices were set in cabinet onthe basis of calculations designed toequalize the returns to family labor be-tween crops. At the consumer level, al-though many African countries intro-duced price controls, they were seldomenforced. However, in a few countriesenforcement was more rigorous: tradersevading controls were imprisoned inTanzania and tortured and executed inEthiopia. At the peak, the Tanzaniangovernment was setting 2,000 prices. Afinal aspect of price control was pan-territorial pricing: transport costs weredisregarded. In some countries, mini-mum quantities of sales were imposedon producers. In manufacturing thesewere unenforceable, but in smallholderagriculture targets were enforced.60

Governments were suspicious of mar-kets and became frustrated with theevasion of controls: in Accra, as punish-ment for breaches of price controls, thegovernment blew up the central market.

This undermining of markets was det-rimental to growth. We now show thatit increased the costs of trade, reducedmarket integration, reallocated re-sources inefficiently, caused a retreatinto subsistence, increased risks, andcriminalized economic activity.

The costs of trade increased: in theTanzanian grain market, trading costsdoubled during the period of controls,and fell by over 60 percent once theywere removed. In turn, this reduced al-locative efficiency by reducing marketintegration. The coefficient of variationof Tanzanian maize prices in regionalcenters doubled between 1964 and1980, and sharply declined again oncethe market was liberalised. In Kenyathe Maize and Produce Board wasunable to handle the volume of trans-actions induced by the prices it set,

60 See Dercon (1995) for Ethiopia; Collier, Rad-wan, and Wangwe (1986) and Bevan et al. (1989)for Tanzania.


causing large differentials between dis-tricts in unofficial prices. In Benin andEthiopia during the control period,food markets were only weakly inte-grated, with increased integration afterliberalization.61

Farmers were responsive to mis-pric-ing. In Tanzania, pan-territorial pricingof grains generated a huge surplus inregions too inaccessible for it to be ofvalue. Conversely, there were huge re-ductions in export volumes. For exam-ple, Tanzanian cotton production wouldhave been around 50 percent higherwithout taxation (Dercon 1993).

We suggested above that part of thegrowth process in African agriculturewas that households moved up a ladderof opportunities, reflecting an initialdisequilibrium in which many exportactivities offered significantly higherreturns. Governments often changedincentives sufficiently to arrest thisprocess. For example, cost-plus croppricing was intended to make all exist-ing activities equally profitable. Moredrastically, the heavy taxation of agri-culture expanded subsistence at the ex-pense of the market: in some countries,such as Uganda, subsistence was for adecade the fastest growing sector of theeconomy.

In addition to the widespread re-source reallocations arising from croptaxation, generalized price controls,though less common, had more dra-matic growth-reducing effects. Pricecontrols of consumer goods, where en-forced, created shortages, and pan-territorial pricing shifted the most se-vere shortages to rural areas, sinceunrecoverable transport costs couldthereby be avoided. Rural shortages ofconsumer goods were a disincentiveto the sale of crops, and altered price

responses. Farmers reduced their salesof crops in reaction to reduced suppliesof consumer goods by a combinationof producing less and using a largerfraction for own consumption. Withinsuch a rationing regime, supply re-sponse to raising crop prices can beperverse so that the high taxation poli-cies can appear not to be reducingsupply.62

A further effect of the weakening ofmarkets was an increase in risks. Themonopolized food marketing channelsmade purchases unreliable. For exam-ple, inefficiency on the part of the Ken-yan Maize Marketing Board led to local-ized shortages, inducing farmers toretain subsistence production. Erraticsupplies of consumer goods induced in-creased money holdings in order to takeadvantage of occasional availability: thevelocity of circulation of rural moneyholdings approximately halved inMozambique and Tanzania in the fouryears coinciding with the onset ofshortages. Reduced integration in thelivestock market increased risks byundermining the use of livestockas an asset for consumption smooth-ing.63 Finally, the domestic price of im-ports fluctuated because of frequentchanges in trade policy. Where tradepolicy went through cycles of liberaliza-tion and protection, the liberalizationphases induced hoarding of imports inanticipation of policy reversal.64

Government interventions in marketswere to an extent by-passed, but thiscriminalized much trading activity. By

61 See Gordon (1990), Dercon (1995), and Lutz(1995).

62 See Azam and Faucher (1988) for Mozam-bique; Berthelemy (1988) for Madagascar; Bevanet al. (1987, 1989, 1990), Collier and Gunning(1991), and Dercon (1993) for Tanzania; Azam etal. (1994) for Angola; and Berthelemy and Morri-son (1987) and Bevan et al. (1987a, 1991) for theunderlying theory.

63 See Fafchamps and Gavian (1996) for Niger.64 See Reinikka (1996) for an application to

Kenya.


the mid-1980s, the illegal economy wasestimated to be 30 percent of Tanza-nian GDP. In Uganda, during the pro-longed period of high implicit coffeetaxation, some 27 percent of the cropwas smuggled out of the country. Al-though criminalization mitigated gov-ernment policies, it had growth-retardingeffects. It increased costs: contract en-forcement costs rose because the courtscould not be used, and because secrecyreduced the efficacy of sanctionsthrough social networks. As a resultthere was a gradual movement fromcredit to a cash economy, and tradingmargins widened. Criminalization keptprofitable enterprises small in orderto reduce visibility. Finally, it contrib-uted to the high capital flight noted inTable 8. Product markets were both amechanism for this flight and a cause.Over-invoicing of imports and under-invoicing of exports occurred on a mas-sive scale. Criminalization increasedcapital flight partly because illegally ac-quired profits were safer held abroad,and partly because, since letters ofcredit could not be used, imports had tobe paid for in advance.65

An Interim Assessment. In retrospect,the degree of hostility of many Africangovernments to private product marketsis surely astonishing. The main measureused in the aggregate-level growth lit-erature, the premium in the parallelforeign exchange market, is clearly onlya crude approximation for the numer-ous ways in which product marketswere undermined. However, the mainmessages of that literature, that Afri-can markets were radically less openthan those elsewhere and that thisseriously reduced growth, are amplysupported.

5. Implications of the Evidence

Does the Evidence Cohere? We nowassess whether the explanations of thegrowth regressions (Section 2) are sup-ported by the evidence on agents andmarkets (Sections 3 and 4). The aggre-gate growth regressions with which westarted found five significant causes ofAfrican stagnation, but this was notexhaustive since a significant Africadummy usually remained.

First, are the explanatory variablesused in the growth literature sup-ported? Section 4 confirmed that prod-uct markets had been heavily distortedand that this should have serious growtheffects. Financial markets were shownto be extremely limited, while labormarkets were not a source of majorproblems: the inflations that turned in-terest rate controls into acute financialrepression liberalized the wage con-trols. Thus, by the 1980s the main dis-tortions were in the product and creditmarkets. The variables used in thegrowth regressions include proxies forthe large distortions in these two mar-kets while ignoring the labor market,and this is consistent with the evidence.

However, Section 3 identified threeimpediments to growth that were inade-quately proxied in the growth regres-sions: high risk, inadequate social capi-tal, and inadequate infrastructure. Afourth, lack of finance, appeared tohave had more severe consequences forhouseholds than for firms. Betweenthem, these may well account for thepersistence of a significant Africadummy in the regressions. The reviewsof agents and markets both emphasizedthe high-risk environment. Where con-tracts could be designed so as to ensurerepeated transactions, there was evi-dence of risk-sharing between house-holds, between firms, and betweenfirms and households. The proxies for

65 See Maliyamkono and Bagachwa (1990) onTanzania; Henstridge (1995) on Uganda; Collierand Gunning (1995a) on trading margins; andYeats (1990) on mis-invoicing.


risk used in the growth regressionsmake an attempt at capturing systemicrisk, though omitting such obviously im-portant components as climatic volatil-ity, but completely miss the high levelof idiosyncratic risk which is such astriking feature of the continent. Forexample, aggregate price volatility un-derstates price volatility at the level ofthe individual agent because marketsare fragmented. Similarly, the effects ofdeficient social capital may be under-stated. Some growth regressions proxysocial capital by ethno-linguistic frac-tionalization, and while this is exoge-nous, it is not closely related to many ofthe deficiencies identified in Sections 3and 4. Growth regressions often includemore specific proxies for social capital,such as contract enforcement, however,enforceability is endogenous: if firmsincur high costs they can reduce en-forceability problems to levels not muchworse than those elsewhere. The prox-ies for public service delivery are per-haps most deficient relative to whatmight be feasible. Africa’s high trans-port costs, poor health care, and inef-fective police are essentially not cap-tured. Only the telephone system isproperly proxied, and its effect is foundto be massive, possibly because it prox-ies infrastructure more generally. Theproxy for financial depth is probablysatisfactory as a characterization of thelimited extent of finance for formal ac-tivities, but misses informal rural credit,which we have seen is also limited.

Despite the deficiencies discussedabove, there is reasonable agreementbetween what the growth regressionsfind to be important and the variablessuggested by the other literature. A lackof openness in product markets, a lackof social capital, high risk, and poorpublic services are the four factors thatboth find to be important. The lack offormal finance is found to be a minor

effect at both the aggregate level andthe firm level, but more important atthe household level. The disagreementwithin the aggregate-level evidence oneducation tends to be resolved by themicro-level evidence against it beingimportant. In agriculture most of theAfrican studies find it to have had littleor no effect, while in manufacturing thereturn to education is much lower thanto physical capital.

Thus, to a surprising extent, the ag-gregate and microeconomic analyses co-here. The microeconomic analyses sug-gest a possible interpretation of thefour main variables that explain stagna-tion in reduced form: openness, socialcapital, risk, and public services. In es-sence it is the following. Africa stag-nated because its governments werecaptured by a narrow elite that under-mined markets and used public servicesto deliver employment patronage.These policies reduced the returns onassets and increased the already highrisks private agents faced. To cope, pri-vate agents moved both financial andhuman capital abroad and diverted theirsocial capital into risk-reduction andrisk-bearing mechanisms.

Is Slow Growth Inevitable in Africa?Above, we have made two distinctionsin the causes of slow growth. The firstwas between those that are intrinsic,notably geography, and those that arepolicy-dependent. The second was be-tween those causes well-proxied in theregression analysis, which are predomi-nantly macro, and those identified atthe levels of agents and markets, whichare microeconomic. There are thusthree conceptually distinct causes ofslow growth: geography, macroeco-nomic policies, and microeconomicpolicies. However, in practice the ef-fects of adverse geography and adversepolicies are often difficult to disentan-gle: high risk, high transport costs, and


high dependence upon commodity ex-ports are all due to both. The relativeimportance of macroeconomic and mi-croeconomic policies cannot be quanti-fied, because the latter are so poorlyproxied in the regression literature. Asa result of these problems of attribu-tion, it is not possible to determinefrom the growth regressions whetherslow growth is inevitable in Africa. Inparticular, the Africa dummy could bedue either to unobserved geography orunobserved microeconomic policies.

In our view, while Africa indeed suf-fers from some geographic disadvan-tages, these are not sufficient to con-demn the continent to continued slowgrowth. One reason for this is that byvirtue of its long phase of poor policies,Africa now has a huge offsetting benefitin the form of “catch-up.” For example,consider the growth rates implied bythe Sachs and Warner regression in Ta-ble 1. While both tropical location andlandlockness are significantly negative,there is also a significant conditionalconvergence effect. Comparing a non-tropical coastal developing country suchas China, with a tropical landlockedcountry such as Uganda, the latter’sgeographic disadvantage is more thanoffset by its catch-up advantage. Hence,in the medium run, were the policy en-vironments the same, catch-up woulddominate geography: Uganda wouldhave a prolonged phase during which itwould grow more rapidly than China.

The effects of policy change caneither be inferred through the cross-country growth regressions or observeddirectly as countries change policies.The deceleration from the 1960s is po-tentially attributable to policy deterio-ration were climate change and terms oftrade deterioration properly included inthe analysis. The reforms of the “struc-tural adjustment” era offer the same po-tential. However, although during the

1980s many governments embarkedupon wide-ranging economic reformprograms under the auspices of theWorld Bank and the IMF, in practicethe programs were not implemented. Asof April 1993, only 5 out of a total of 26ESAF programs had been completedwithin their planned period and 8 hadapparently broken down altogether. Inthe period 1980–88 three-quarters ofWorld Bank adjustment loans had in-stalment tranche releases delayed be-cause of non-implementation of policyconditions (Tony Killick 1996, pp. 213–14). Only during the 1990s have a sig-nificant number of governments seri-ously embarked upon reform. Thelongest reforming country, Ghana, lib-eralized its exchange rate in 1987. Cur-rently the leading reform country isprobably Uganda, but its key reformsdate only from 1992: as late as March1992, the rate of inflation was 230percent and the country had the worstrisk-rating in Africa. Ethiopia, nowanother leading reformer, only endedits civil war in 1991. Zimbabwe, a thirdreformer, began its programme in1991. The Franc Zone only correctedovervaluation in 1994.

Potentially, the reforms constitute a“natural experiment” to test whetherslow growth is inevitable. However, be-cause the reforms are recent, any suchexperiment faces two problems. First, itis only possible to observe the initial re-sponse to reform: this may either un-derstate or overstate the ultimate ef-fects. Secondly, the composition of thereforms is biased towards those thatcan be implemented quickly. Thus,macroeconomic policy, which can beswiftly altered through the exchangerate, tariffs, taxes and the cancellationof infrastructure expenditures, has im-proved substantially in the reformingcountries, whereas public service de-livery, which can only be improved


slowly, has lagged behind and may evenhave deteriorated because of fiscaladjustment.

Bearing in mind these two limita-tions, we consider the evidence fromthe experiment. We sort the low-income African countries (those below$1,000 per capita) according to threecriteria that appear important forgrowth: peace, the avoidance of a highdegree of macroeconomic instability,and the avoidance of large allocative in-efficiencies. We compare the growthrates of output per worker for the mostrecent period, namely 1995–96.

During the 1990s, seven countrieshave been severely affected by civil war:Angola, Burundi, Liberia, Rwanda, Si-erra Leone, Somalia and Sudan. Be-tween them they account for 12 percentof the population of Sub-Saharan Af-rica. The data on GDP for such coun-tries are either highly unreliable ormissing. With this proviso, the averagegrowth rate was 0.8 percent.66

The next criterion concerns the mac-roeconomic environment. The appropri-ate fiscal stance for Africa is disputed.Bank–Fund programmes typically aimfor an inflation rate of 5 percent, butseldom achieve it. In support of a targetat around this level, Michael Sarel(1996) finds that inflation in excess of 8percent reduces growth. However, crit-ics argue for a higher inflation ceiling,typically 20 percent. We therefore setthe inflation standard at the fairly unde-manding level of an average rate below25 percent. The following African coun-tries that satisfy the condition of peacecurrently fail to meet the requirementof macroeconomic stability thus de-fined: Comoros, Equatorial Guinea,Madagascar, Malawi, Mozambique, Ni-ger, Nigeria, Sao Tome and Principe,

Tanzania, Togo, Zaire and Zambia. Thisgroup covers 222m people, which is 43percent of the population of Sub-Saha-ran Africa. As with social disorder, onecasualty of deep macroeconomic disor-der is economic statistics. With this pro-viso, the average growth rate was 2.7percent.

The remaining criterion concernspolicies that affect resource allocation.It is immensely difficult to get goodquantitative measures of the efficacy ofresource allocation policies. The onlyresource allocation indicator that is asreadily measurable as the major macro-economic indicators is the premium onthe parallel market for foreign ex-change. However, this is only one ofmany ways in which policy can affect re-source allocation. We assess five policyareas that concern resource allocation:trade and exchange rate systems; the fi-nancial sector; factor and product mar-kets; parastatals; and the composition ofpublic expenditure. We utilize a scoringsystem in which World Bank countryeconomists rate each policy area ac-cording to reasonably precise criteria(Deepak Bhattasali and A. Ray 1995).We classify countries as having mini-mum adequate resource allocation poli-cies if each of the five policies is abovea common low threshold. Eight coun-tries that have met the first two criteriado not satisfy this one: Cameroon,Chad, Congo, Eritrea, Guinea, Kenya,Lesotho and Zimbabwe. These coun-tries have a combined population of69m, or 12 percent of the population ofSub-Saharan Africa. The growth rate forthis group was 4.2 percent p.a.

The three criteria leave twelve low-income countries whose governmentsare currently providing at least modestlevels of social order, macroeconomicorder and resource allocation: Benin,Burkina Faso, Cape Verde, Côted’Ivoire, Ethiopia, Gambia, Ghana,

66 Growth rates are from World DevelopmentIndicators 1998 (World Bank forthcoming).


Guinea Bissau, Mali, Mauritania,Senegal, and Uganda. Between themthese countries account for 26 percentof the population of Sub-Saharan Af-rica. The average growth rate for thisgroup was 4.7 percent, which is quite ahigh annual growth rate of output perworker.

Thus, the “natural experiment” ap-pears to indicate that Africa’s slowgrowth has been the result of poor mac-roeconomic policies. However, such aconclusion would be premature. Thecurrent good growth rates of African re-formers are misleading as an estimateof the medium run pay-off to macro-economic policy reform in two respects.First, the correction of bad macro-economic policies enables output tobounce back without factor accumula-tion, which yields unsustainably rapidgrowth. Secondly, the reforms may failto induce an investment response. Thismay be because firms remain con-strained until microeconomic reformsare implemented, because of a debtoverhang which discourages investmentdue to fears of future tax liabilities, orbecause the recent macroeconomic re-forms as yet lack credibility. Since the“bounce back” effect and the invest-ment discouragement effect work in op-posite directions, current growth mayeither understate or overstate medium-run growth, and so is of limited use indetermining whether slow growth isinevitable.

There is evidence that both a lack ofinvestment response and “bounce back”are currently important. While thegrowth rates of the African reformersare currently similar to pre-crisis EastAsia, investment rates are around 9 per-centage points lower. The low rate ofinvestment is entirely due to low privateinvestment: public investment rates arecomparable to those elsewhere. In Sec-tion 3 we reviewed the firm-level evi-

dence on constraints to growth, andshowed that deficient public servicesare likely to be substantially retardinginvestment. There is some evidence fora debt Laffer curve with most Africancountries on the wrong side of it (El-badawi, Ndulu, and Ndung’u 1996).However, a problem with the analysis ofthe effects of debt is that it is likely tobe correlated with poor and persistentpolicies.

There is also evidence that the lim-ited credibility of macroeconomic re-forms is important. Survey evidenceshows that the main component of in-vestor risk is the fear of policy reversal.This may reflect both the history ofpoor policy and the policy reversals ofneighbours. Africa is rated as the mostrisky region in the world, and its posi-tion deteriorated sharply from a ratingof 31.8 in 1979 to 21.7 in 1995. Poten-tial investors may fear policy reversaland hold off until risk assessments haveimproved. In regression analysis, laggedrisk ratings significantly reduce privateinvestment during the 1990s: if the re-forming African countries had the riskrating of Botswana and Mauritius, pri-vate investment as a share of GDPwould have been around five percent-age points higher (Frederick Z. Jas-persen, Anthony Aylward, and A. Knox1998). If high risk is a major deterrentto investment, we would expect to findlow investment coexisting with high re-turns. As we showed in Section 2, in thepast this has not been the case: returnshave been low. However, currently for-eign investors identify risk as the majorobstacle to investment in Africa. Thereturn on foreign direct investment inAfrica during 1990–94 was around 60percent higher than that in other devel-oping regions (in the range of 24–30percent as against 16–18 percent). Yetthese high returns have been insuffi-cient to offset the high risks: private


capital flows to Africa remain far belowthose to all other regions. In 1995,flows to Africa (excluding South Africa)were less than 2 percent of all flows todeveloping countries and less than halfthose to the next-lowest region, theMiddle East (Amar Bhattacharya, PeterMontiel, and Sunil Sharma 1996).

The ratings of the three major risk-rating services are largely explicable interms of the economic characteristics ofa country, such as its level of reserves,but Africa is rated significantly worsethan warranted by these characteristics(Nadeem Ul Haque, Mark Nelson, andDonald Mathieson 1998). There is thusa significant Africa dummy in risk rat-ings. Jaspersen, Aylward, and Knox(1998) find that the Africa dummy per-sists in all their regressions of foreigndirect investment during 1990–94. Theyconclude: “This gives some support tothe conjecture that investors may be ir-rationally averse to committing FDI to

African countries, since the Africa ef-fect appears to dominate a range of fun-damental economic, political and socialrisk factors in the regression analysis.”Further, the reform countries werethose with the worst risk ratings, asshown in Figure 1, which relates therisk ratings to a World Bank classifica-tion of 26 countries into strong reform-ers, weak reformers, and those in whichpolicy has deteriorated (World Bank1994). The countries that became thestrong reformers post-1987 had consid-erably the worst risk ratings rightthroughout the 1980s. Although policyreform has gradually improved theirrisk ratings, they started from a verylow base.

To summarize, the “natural experi-ment” evidence on the growth responseto reform is directly of limited use be-cause growth rates are contaminated bythe “bounce-back” effect. However, in-directly, it points to a problem of a lack

Figure 1. Institutional Investor Risk Ratings for Africa by Policy Environment

Sep-81

30

20

10

0

Risk rating (higher rating = lower risk)

Large policyimprovement

Small policyimprovement

Policydeterioration

Source: Collier and Pattillo 1998.

Sep-83 Sep-85 Sep-87 Sep-89 Sep-91 Sep-93 Oct-95 Sep-97


of private investment response. This iseither because the returns on invest-ment remain low, or because they arenot high enough to compensate for per-ceived high risks and debt overhang.We have seen that there is clear evi-dence that perceived risks are indeedhigh. Potentially, the returns to invest-ment may still be low despite macro-economic policy reform because of theintrinsic disadvantages of geography,because macroeconomic policy is unim-portant, or because of the poor micro-economic policy environment. How-ever, we have seen that according to theregression evidence, the disadvantagesof geography are at least over the me-dium term more than offset by catch-up, and that macroeconomic policy ispotent. This suggests that medium-rungrowth is constrained by perceptions ofhigh risk and debt overhang, and bypoor microeconomic policies. In thelong run, geography may reassert itself,but only after a prolonged phase ofrapid growth due to catch-up.

We therefore finally consider theprospects for improved microeconomicpolicies, for reduced risk, and for debtforgiveness. All of these prospects de-pend upon the political economy of re-form. Underlying donor conditionalitywas the belief that aid could be usedboth to induce reform and to lock gov-ernments in. This has been discredited.There has been no clear relationshipbetween aid and policy change, and thephase of conditionality has not seen animprovement in risk ratings, but indeeda deterioration (Burnside and Dollar1997). Donors have increasingly recog-nised this, and the new Highly IndebtedPoor Countries debt relief initiative in-stead targets aid to countries that havean established and sustained record ofpolicy reform. There is scope for moreeffective lock-in mechanisms than do-nor conditionality, by constructing do-

mestic and external restraints upon pol-icy relapse. There are limits on the effi-cacy of domestic restraints due to thefragility of the rule of law in much ofAfrica. For example, the creation of le-gally independent national centralbanks might have little effect on credi-bility because of a history of autocraticbehavior by African presidents. Exter-nal restraints other than conditionalityinclude the WTO to bind their tariffs atlow levels, and the European Union inan arrangement similar to NAFTA (Col-lier and Gunning 1995). The FrancZone shows both the benefits and thecosts of external restraint. During thefirst post-independence decade it sig-nificantly raised the growth rate of itsmembers above the rest of Africa. How-ever, by the late 1980s this effecthad been reversed: governments hadlearned how to evade the budget re-straints while the economies sufferedfrom being locked into the same, in-creasingly overvalued exchange rates(Devarajan and de Melo 1991).

Finally, we consider the changingpower base of governments. As we haveseen, until recently African govern-ments have been undemocratic, insteadbeing responsive to their narrow urbanconstituencies:

Owners and workers in industrial firms, eco-nomic and political elites, privileged farmersand the managers of public bureaucracies—these constitute the development coalition incontemporary Africa. (Bates 1981, p. 121)

It is because governments were re-sponsive to this constituency that theysurvived, so that generalized economicdecline posed only a modest threat.Those reforms that depressed urbanconsumption significantly increasedstrikes and demonstrations, and govern-ments responded by reversing policies(Christian Morrisson, Jean-DominiqueLafay, and Sébastien Dessus 1994). In


effect, the policy equilibrium identifiedby Bates proved locally stable: whendisturbed by temporary aid, the inducedreforms were reversed. The wave ofdemocratization in Africa during the1990s has, however, weakened thepower of the old elites and so paved theway for politically sustainable reforms.To date, the reform of trade and ex-change rate policies has proved mucheasier than the reform of infrastructureand institutions. This may be becausethe latter inherently take more time toimplement. Alternatively, the politicalequilibrium may have changed suffi-ciently to permit macroeconomic re-form, but not sufficiently to permitmicroeconomic reform. The problemsand potential for economic policy mak-ing and improved service delivery un-der democratization constitute a newresearch agenda.

REFERENCES

Ablo, Emmanuel and Ritva Reinikka. 1998. “DoBudgets Really Matter? Evidence from PublicSpending on Education and Health in Uganda,”World Bank Policy Research Working PaperNo. 1926.

Amjadi, Azita and Alexander J. Yeats. 1995. “HaveTransport Costs Contributed to the RelativeDecline of Sub-Saharan African Exports?”World Bank Policy Research Working Paper1559.

Appleton, Simon; John Hoddinott, and John B.Knight. 1996. “Primary Education as an Inputinto Post-Primary Education: A NeglectedBenefit,” Oxford Bul. Econ. Statist., 58:1, pp.211–19.

Azam, Jean-Paul and Jean-Jacques Faucher. 1988.The Supply of Manufactured Goods and Agri-cultural Development: The Case of Mozam-bique. Paris: OECD.

Azam, Jean-Paul; Paul Collier, and Andrea Crav-inho. 1994. “Crop Sales, Crop Shortages andPeasant Portfolio Behaviour: An Analysis of An-gola,” J. Devel. Stud., 30:2, pp. 361–79.

Baland, Jean-Marie and Jean-Philippe Platteau.1996. Halting Degradation of Natural Re-sources: Is There a Role for Rural Communi-ties? Oxford: Oxford U. Press, Clarendon.

Barr, Abigail M. 1996. Entrepreneurial Networksand Economic Growth. PhD. dissertation, Ox-ford U.

Barro, Robert J. and Jong-Wha Lee. 1994. “Losersand Winners in Economic Growth,” in Proceed-

ings of the Annual World Bank Conference onDevelopment Economics, 1993, pp. 267–97.

Bates, Robert H. 1981. Markets and States inTropical Africa. Berkeley: U. California Press.

———. 1983. Essays in the Political Economy ofRural Africa. Cambridge: Cambridge U. Press.

———. 1990. “Capital, Kinship and Conflict:The Structuring Influence of Capital in KinshipSocieties,” Can. J. African Stud., 24, pp. 151–64.

Berthelemy, Jean-Claude. 1988. The Supply ofManufactured Goods and Agricultural Develop-ment: The Case of Madagascar. Paris: OECD.

Berthelemy, Jean-Claude and Christian Morris-son. 1987. “Manufactured Goods Supply andCash Crops in Sub-Saharan Africa,” World De-vel., 15, pp. 1353–67.

Besley, Timothy. 1995. “Property Rights and In-vestment Incentives: Theory and Evidence fromGhana,” J. Polit. Econ., 103:5, pp. 903–37.

Bevan, David; Arne Bigsten, Paul Collier, and JanWillem Gunning. 1987. East African Lessons onEconomic Liberalization, Thames Essay No. 48,Trade Policy Research Centre, London: Gower.

———. 1987a. “Peasant Supply Response in Ra-tioned Economies,” World Devel., 15, pp. 431–39.

Bevan, David; Paul Collier, and Jan Willem Gun-ning. 1989. Peasants and Governments: An Eco-nomic Analysis. Oxford: Oxford U. Press,Clarendon.

———. 1990. Controlled Open Economies: A Neo-classical Approach to Structuralism. Oxford:Oxford U. Press, Clarendon.

———. 1991. “Income and Substitution Effects inModels of Peasant Supply Response under Ra-tioning,” Oxford Econ. Pap., 43, pp. 340–43.

Bhattacharya, Amar; Peter J. Montiel, and SunilSharma. 1996. “Private Capital Flows to Sub-Saharan Africa: An Overview of Trends and De-terminants.” IMF Research Dept. mimeo.

Bhattasali, Deepak and A. Ray. 1995. “Sub-Saha-ran Africa: Policies and Outcomes.” World BankAfrica Region mimeo.

Biggs, Tyler; Mayank Raturi, and Pradeep Srivas-tava. 1996. “Enforcement of Contracts in an Af-rican Credit Market: Working Capital Financ-ing in Kenyan Manufacturing,” World BankRPED Discussion Paper.

Biggs, Tyler and Pradeep Srivastava. 1996. “Struc-tural Aspects of Manufacturing in Sub-SaharanAfrica: Findings from a Seven-Country Enter-prise Survey,” World Bank RPED DiscussionPaper.

Bigsten, Arne; Paul Collier, Stefan Dercon, Ber-nard Gauthier, Jan Willem Gunning, AndersIsaksson, Abena Oduro, Remco Oostendorp,Catherine Pattillo, Måns Soderbom, M. Sylvain,Francis Teal, and Albert Zeufack. Forthcoming.“Investment in the Manufacturing Sector in Af-rica: Evidence from Firm Surveys in Cameroon,Kenya, Ghana and Zimbabwe,” Oxford Bul.Econ. Statist.

———. 1998. “Exports and Firm-Level Efficiency


in the African manufacturing Sector,” Centrefor Study of African Economies mimeo, OxfordU.

Bigsten, Arne; Paul Collier, Stefan Dercon, Mar-cel Fafchamps, Bernard Gaultier, Jan WillemGunning, Anders Isaksson, Abena Oduro,Remco Oostendorp, Catharine Pattillo, MånsSoderbom, Francis Teal, and Albert Zeufack.1998. “Rates of Return on Physical and HumanCapital in Africa’s Manufacturing Sector,” Cen-tre for Study of African Economies mimeo, Ox-ford U.

Binswanger, Hans P. and John McIntire. 1987.“Behavioral and Material Determinants of Pro-duction Relations in Land Abundant TropicalAgriculture,” Econ. Devel. Cult. Change, 36,pp. 73–99.

Blanchflower, David G. and Andrew J. Oswald.1995. The Wage Curve. Cambridge, MA.: MITPress.

Bleaney, Michael; Norman Gemmell, and DavidGreenaway. 1995. “Tax Revenue Instability,with Particular Reference to Sub-Saharan Af-rica,” J. Devel. Stud., 31, pp. 883–902.

Boissière, M.; John B. Knight and Richard H. Sa-bot. 1985. “Earnings, Schooling, Ability, andCognitive Skills,” Amer. Econ. Rev., 75:5, pp.1016–30.

Bruce, John W. 1988. “A Perspective on Indige-nous Land Tenure Systems and Land Concen-tration,” in Land and Society in ContemporaryAfrica. R.E. Downs and S.P. Reyna, eds. Han-over, N.H.: University Press New England.

Burger, Kees. 1994. “Farm Households, Cash In-come and Food Production: The Case of Ken-yan Smallholdings,” Ph.D. thesis, Free Univer-sity, Amsterdam.

Burger, Kees; Paul Collier, and Jan Willem Gun-ning. 1996. “Social Learning: An Application toKenyan Agriculture,” Free University, Amster-dam mimeo.

Burnside Craig and David Dollar. 1997. “Aid, Poli-cies and Growth,” World Bank Policy ResearchWorking Paper No. 1777.

Byerlee, Derek. 1993. Review of Celis, Milimo,and Wanmali 1991. J. African Econ., 2, pp. 128–32.

Celis, Rafael; John T. Milimo, and Sudhir Wan-mali, eds. 1991. Adopting Improved Farm Tech-nology: A Study of Smallholder Farmers inEastern Province, Zambia. Washington, D.C.:Int. Food Policy Research Institute.

Chamley, Christophe and Patrick Honohan. 1990.“Taxation of Financial Intermediation: Mea-surement Principles and Application to Five Af-rican Countries,” World Bank Policy ResearchWorking Paper No. 421.

Chhibber, Ajay and Stanley Fischer, eds. 1991.Economic Reform in Sub-Saharan Africa.Washington, D.C.: The World Bank.

Cleave, John Hogarth. 1974. African Farmers: La-bour Use in the Development of SmallholderAgriculture. New York: Praeger.

Coe, David T. and Alexander W. Hoffmaister.

1998. “North-South Trade: Is Africa Unusual?”IMF WP 98/94.

Collier, Paul. 1998. “The Political Economy ofEthnicity,” in Proceedings of the Annual WorldBank Conference on Development Economics ,Boris Pleskovic and Joseph E. Stiglitz, eds.World Bank.

Collier, Paul and Ashish Garg. Forthcoming. “OnKin Groups and Wages in the Ghanaian LabourMarket,” Oxford Bul. Econ. Statist.

Collier, Paul and Jan Willem Gunning. 1991. “TheLiberalization of Price Controls: Theory and anApplication to Tanzania,” in Chhibber and Fis-cher 1991.

———. 1991a. “Money Creation and FinancialLiberalisation in a Socialist Banking System:Tanzania,” World Devel., 19, pp. 533–38.

———. 1995. “Trade Policy and Regional Integra-tion: Implications for the Relationship betweenEurope and Africa,” The World Economy,Vol.18, pp. 387–410.

———. 1995a. “War, Peace and Private Portfo-lios,” World Development, Vol. 23, pp. 233–41.

———. 1997. “Explaining African Economic Per-formance,” Centre for the Study of AfricanEconomies, Working Paper S97–2.1, Oxford U.

Collier, Paul and Jan Willem Gunning with associ-ates. 1998. Trade Shocks in Developing Coun-tries: Theory and Evidence. Oxford: Oxford U.Press. Clarendon.

Collier, Paul; Anke Höffler, and Catherine Pat-tillo. 1998. “Capital Flight as a PortfolioChoice,” World Bank Development ResearchGroup mimeo.

Collier, Paul and Deepak Lal. 1986. Labour andPoverty in Kenya, 1900–1980. Oxford: OxfordU. Press, Clarendon.

Collier, Paul and Catherine Pattillo, eds. 1998. In-vestment and Risk in Africa. London: Macmil-lan.

Collier, Paul; Samir Radwan, and SamuelWangwe. 1986. Labour and Poverty in RuralTanzania. Oxford: Oxford U. Press, Clarendon.

Collins, Susan M. and Barry P. Bosworth. 1996.“Economic Growth in East Asia: Accumulationversus Assimilation,” Brookings Pap. Econ. Act.,2, pp.135–203.

Craig, John. 1990. “Comparative African Experi-ences in Implementing Educational Policies,”World Bank Disc. Paper, Africa TechnicalDept. Series, 83.

Deaton, Angus. 1990. “Saving in DevelopingCountries: Theory and Review,” Proceedings ofthe Annual World Bank Conference on Develop-ment Economics, 1989, World Bank Econ. Rev.,Supplement, pp. 61–96.

———. 1992. “Saving and Income Smoothing inCôte d’Ivoire,” J. African Econ., 1, pp. 1–24.

———. 1994. Understanding Consumption. Ox-ford: Oxford U. Press, Clarendon Press.

Deaton, Angus and Ron Miller. 1996. “Interna-tional Commodity Prices, Macroeconomic Per-formance and Politics in Sub-Saharan Africa,” J.African Econ., 5:supp., pp. 99–191.


Dercon, Stefan. 1993. “Peasant Supply Responseand Macroeconomic Policies: Cotton in Tanza-nia,” J. African Econ., 2:2, pp. 157–94.

———. 1995. “On Market Integration and Liber-alisation: Method and Application to Ethiopia,”J. Devel. Stud., 32:1, pp. 112–43.

———. 1996. “Risk, Crop Choice and Savings:Evidence from Tanzania,” Econ. Devel. Cult.Change, 44:3, pp. 485–513.

———. 1998. “Wealth, Risk and Activity Choice:Cattle in Western Tanzania,” J. Devel. Econ.,55:1, pp. 1–42.

Dercon, Stefan and Pramila Krishnan. 1996. “In-come Portfolios in Rural Ethiopia and Tanzania:Choices and Constraints,” J. Devel. Stud., 32:6,pp. 850–75.

Devarajan, Shantayan and Jaime de Melo. 1991.“Membership in the CFA Zone: Odyssean Jour-ney or Trojan Horse?” in Chhibber and Fischer1991.

Dollar, David. 1992. “Outward-Oriented Develop-ing Economies Really Do Grow More Rapidly:Evidence from 95 LDCs 1976–85,” Econ. De-vel. Cult. Change, 40, pp. 523–44.

Easterly, William and Ross Levine. 1997. “Africa’sGrowth Tragedy: Policies and Ethnic Divi-sions,” Quart. J. Econ., 112:4, pp. 1203–50.

———. 1998. “Troubles with the Neighbours: Af-rica’s Problem, Africa’s Opportunity,” J. AfricanEcon., Vol. 7:1, pp. 120–42.

École des Hautes Études Commercials. 1993 “Pre-liminary Report on the Cameroon Survey,”Centre d’Etude et Administration Internation-ale, Montreal, mimeo.

Eicher, Carl K. and Doyle C. Baker. 1982. “Re-search on Agricultural Development in Sub-Sa-haran Africa: A Critical Survey,” Int. Devel.Pap. No. 1, Michigan State U.

Elbadawi, Ibrahim A. and Benno J. Ndulu. 1996.“Long Run Development and SustainableGrowth in Sub-Saharan Africa”, in New Direc-tions in Development Economics. M. Lundahland B.J. Ndulu, eds. London: Routledge.

Elbadawi, Ibrahim A.; Benno J. Ndulu, andNjuguna Ndung’u. 1996. “Debt Overhang andEconomic Growth in Sub-Saharan Africa,” Afri-can Economic Research Consortium, Nairobi,mimeo.

Elbadawi, Ibrahim A. and Klaus Schmidt-Hebbel.1997. “Macroeconomic Policies, Instability andGrowth in the World,” J. African Econ., 6:supp.

Ensminger, Jean. 1995. “Changing PropertyRights: Reconciling Formal and Informal Rightsto Land in Africa,” Washington U. Anthropol-ogy Dept. mimeo.

Fafchamps, Marcel. 1996. “Market Emergence,Trust and Reputation,” Stanford U. mimeo.

———. 1996b. “The Enforcement of CommercialContracts in Ghana,” World Devel., 24:3, pp.427–48.

Fafchamps, Marcel and Sarah Gavian. 1996. “TheSpatial Integration of Livestock Markets in Ni-ger,” J. African Econ., Vol. 5:3, pp. 366–405.

Fafchamps, Marcel; Jan Willem Gunning, and

Remco Oostendorp. 1998. “Inventories, Liquid-ity and Contractual Risk in African Manufactur-ing,” Stanford U. Econ. Dept. mimeo.

Fafchamps, Marcel; Christopher Udry, and Kath-erine Czukas. 1998. “Drought and Saving inWest Africa: Are Livestock a Bufferstock?” J.Devel. Econ., 55, pp. 273–305.

Fehr, Nils-Henrik M. von der; Finn R. Forsund,and S.A.C. Kittelsen. 1994. “The ZambianRPED Study: Final Report,” U. Oslo and U.Zambia, mimeo.

Ferree K.; S. Singh, and Robert Bates. 1996. “Po-litical Institutions and Economic Growth in Af-rica,” Harvard U. Dept. of Government.

Fielding, David. 1993. “Determinants of Invest-ment in Kenya and Côte d’Ivoire,” J. AfricanEcon., 2, pp. 299–328.

Fisher, Neil M. 1979. “Studies in Mixed CroppingIII. Further Results with Maize-Bean Mix-tures,” Experimental Agriculture, 15, pp. 49–58.

Food and Agriculture Organization. 1986. AfricanAgriculture, The Next 25 Years. Rome: FAO.

Free University Amsterdam (Econ. and Social In-stitute) and University of Zimbabwe, Harare(Econ. Dept.). 1993. “Regional Program on En-terprise Development: First Report on the Zim-babwe Survey,” mimeo.

———. 1995. “Regional Program on EnterpriseDevelopment: Report on Round III of the Zim-babwe Survey,” mimeo.

Gallup, John Luke and Jeffrey D. Sachs. 1998.“Geography and Economic Growth,” in Pro-ceedings of the Annual World Bank Conferenceon Development Economics. Pleskovic andStiglitz, eds. Washington, D.C.: World Bank.

Gelb, Alan; John B. Knight and Richard H. Sabot.1989. “Public Sector Employment, Rent Seek-ing and Economic Growth,” Econ. J., 101, pp.1186–99.

Gemmell, Norman. 1996. “Evaluating the Impactsof Human Capital Stocks and Accumulation onEconomic Growth: Some New Evidence,” Ox-ford Bul. Econ. Statist., 58, pp. 9–28.

Ghura, Dhaneshwar and Thomas J. Grennes.1993. “The Real Exchange Rate and Macro-economic Performance in Sub-Saharan Africa,”J. Devel. Econ., 42, pp. 155–74.

Ghura, Dhaneshwar and Michael T. Hadjimichael.1996. “Growth in Sub-Saharan Africa,” IMFStaff Pap., 43:3, pp. 605–34.

Giovannini, Alberto and Martha de Melo. 1993.“Government Revenue from Financial Repres-sion,” Amer. Econ. Rev., 83, pp. 953–63.

Gordon, H. 1990. “Marketing Liberalization,Household Participation in Markets and FoodSecurity in Tanzania,” background paper toWorld Bank report “Nutrition and Food MarketAspects of Food Insecurity in Tanzania.”

Greif, Avner. 1993. “Contract Enforceability andEconomic Institutions in Early Trade: TheMaghribi Traders’ Coalition,” Amer. Econ. Rev.,83, pp. 525–48.

Greif, Avner and Robert H. Bates. 1995. “Organis-


ing Violence: Wealth, Power, and Limited Gov-ernment,” Stanford U. mimeo.

Grove, Alfred Thomas. 1991. “The African Envi-ronment”, in Africa 30 Years On. Douglas Rim-mer, ed. London, James Currey.

Gunning, Jan Willem, ed. 1994. “The Manufactur-ing Sector in Zimbabwe: Dynamics and Con-straints,” Free U. Amsterdam and U. Zim-babwe, Harare mimeo.

Gunning, Jan Willem and Remco Oostendorp.1996. “What Drives Investment? Panel DataEvidence from Industrial Firms in Zimbabwe,”paper presented at Workshop on Investment inAfrica, Centre for Study of African Economies,Oxford U.

Gunning, Jan Willem and Clever Mumbengegwi,eds. 1995. “The Manufacturing Sector in Zim-babwe: Industrial Change under Structural Ad-justment,” Free U., Amsterdam and U. Zim-babwe mimeo.

Gunning, Jan Willem and Marc Pomp. 1995.“Growth, Adjustment and Resource Allocation,”ch. 2 in Gunning and Mumbengegwi 1995..

Hadjimichael, Michael T. and Dhaneshwar Ghura.1995. “Public Policies and Private Savings andInvestment in Sub-Saharan Africa: An Empiri-cal Investigation,” IMF Working Paper 95/19.

Hadjimichael, Michael; Dhaneshwar Ghura, Mar-tin Muhleisen, R. Nord, and E. Murat Ucer.1995. “Sub-Saharan Africa: Growth, Savings andInvestment, 1986–93,” IMF Occasional Paper118.

Haque, Nadeem Ul; Mark Nelson, and Donald T.Mathieson. 1998. “Creditworthiness Ratings:Their Political and Economic Content,” in Col-lier and Pattillo 1998.

Harris, John R. and Michael P. Todaro. 1970. “Mi-gration, Unemployment and Development: aTwo-Sector Analysis,” Amer. Econ. Rev., 60:1,pp. 126–42.

Helleiner, Gerald K. 1986. “Outward Orientation,Import Instability and African EconomicGrowth: an Empirical Investigation”, in Theoryand Reality in Development. S. Lall and F. Ste-wart, eds. London: Macmillan.

———. 1992. “The IMF, the World Bank and Af-rica’s Adjustment and External Debt Problems:An Unofficial View,” World Devel., 20, pp. 779–92.

Henstridge, Mark. 1995. “Coffee and Money inUganda: An Econometric Analysis,” PhD. the-sis, Oxford U.

Hoddinott, John. 1996. “Wages and Unemploy-ment in an Urban African Labour Market,”Econ. J., 106, pp. 1610–26.

Hoff, Karla Ruth; Avishay Braverman, and JosephE. Stiglitz, eds. 1993. The Economics of RuralOrganization: Theory, Practice, and Policy. Ox-ford: Oxford U. Press for the World Bank.

Ikhide, Sylvanus I. 1992. “Financial Deepening,Credit Availability and the Efficiency of Invest-ment: Evidence from Selected African Coun-tries,” UN Econ. Commission for Africa, Devel.Research Paper No.2.

Isham, Jonathan and Daniel Kaufmann. 1999.“The Forgotten Rationale for Policy Reform,”Quart. J. Econ., forthcoming.

Isham Jonathan; Daniel Kaufmann, and LantPritchett. 1995. “Government and the Returnson Investment,” World Bank Policy ResearchWorking Paper 1550.

Jamal, Vali and John Weeks. 1993. Africa Misun-derstood: Or Whatever Happened to the RuralUrban Gap? London: Macmillan.

Jaspersen, Frederick Z.; Anthony H. Aylward,and A.D. Knox. 1998. “The Effects of Risk onPrivate Investment: Africa Compared withOther Developing Areas,” in Collier and Pattillo1998.

Jones, Patricia. 1994. “Are Manufacturing WorkersReally Worth Their Pay?” Centre for Study ofAfrican Econ. mimeo., Oxford U.

Jones, William O. 1960. “Economic Man in Af-rica,” Food Research Institute Stud., 1, pp. 107–34.

Kasekende, Louis and M. Atingi-Ego. 1997. “Fi-nancial Liberalisation and its Implications forthe Domestic Financial System: The Case ofUganda,” Nairobi: African Economic ResearchConsortium.

Kenya Central Bureau of Statistics. 1977. Inte-grated Rural Survey 1974–75: Basic Report.Nairobi.

Killick, Tony. 1996. “Principals, Agents and theLimits of Bretton Woods Conditionality,” WorldEcon., 16, pp. 211–29.

King, Robert G. and Ross Levine. 1993. “Finance,Entrepreneurship and Growth: Theory and Evi-dence,” J. Monet. Econ., 32, pp. 513–42.

Kinsey, Bill; Kees Burger, and Jan Willem Gun-ning. 1998. “Coping with Drought in Zim-babwe: Survey Evidence on Household Re-sponses to Risk,” World Devel., 26, pp. 89–110.

Knack, Stephen and Philip Keefer. 1995. “Institu-tions and Economic Performance: Cross-Coun-try Tests using Alternative Institutional Mea-sures,” Econ. Pol., 7, pp. 207–27.

Knight, John B. and Richard H. Sabot. 1990. Edu-cation, Productivity and Inequality: The EastAfrican Natural Experiment. Oxford: Oxford U.Press.

Lavy, Victor and Jean-Marc Germain. 1994. “Qual-ity and Cost in Health Care Choice in Develop-ing Countries,” World Bank LSMS Working Pa-per No. 105.

Lee, Kyu Sik and Alex Anas. 1991. “Manufactur-ers’ Responses to Infrastructure Deficiencies inNigeria: Private Alternatives and Policy Op-tions,” in Chhibber and Fischer 1991.

Leonard, David K. 1977. Reaching the PeasantFarmer: Organisation Theory and Practice inKenya. Chicago: U. Chicago Press.

Levy, Victor and John L. Newman. 1989. “WageRigidity: Micro and Macro Evidence on LabourMarket Adjustment in the Modern Sector,”World Bank Econ. Rev., Vol. 3, pp. 97–117.

Lutz, C. 1995. The Functioning of the Maize Mar-ket in Benin: Spatial and Temporal Arbitrage


on the Market of a Staple Food Crop. Ph.D.thesis, U. Amsterdam.

Maliyamkono, T.L. and Mboya S.D. Bagachwa.1990. The Second Economy in Tanzania. Lon-don, James Currey.

Mazumdar, Dipak. 1994. “Wages in Africa,” WorldBank Africa Technical Dept., mimeo.

Mauro, Paulo. 1995. “Corruption and Growth,”Quart. J. Econ., 110, pp. 681–712.

Migot-Adholla, Shem; P. Hazell, B. Blarel, and F.Place. 1993. “Indigenous Land Right Systems inSub-Saharan Africa: A Constraint on Produc-tivity,” in Hoff, Braverman, and Stiglitz 1993,pp. 269–91.

Morrisson Christian; Jean-Dominique Lafay, andSébastien Dessus. 1994. “Adjustment Pro-grammes and Politico-Economic Interactions inDeveloping Countries: Lessons from an Empiri-cal Analysis of Africa in the 1980s,” in FromAdjustment to Development in Africa. G.A.Cornia and Gerald K. Helleiner, eds. London:Macmillan.

Mwabu, Germano; Martha Ainsworth, and AndrewNyamete. 1993. “Quality of Medical Care andChoice of Medical Treatment in Kenya: An Em-pirical Analysis,” J. Human Res., 28, pp. 838–62.

Narayan-Parker, Deepa and Lant Pritchett. 1996.“Cents and Sociability: Household Income andSocial Capital in Rural Tanzania,” World BankResearch Dept. mimeo.

Navaretti, Giorgio Barba; Bernard Gauthier, Jaimede Melo, and James R. Tybout. 1996. “How DoIndustrial Enterprises Respond to Policy Re-forms? Recent Evidence from Cameroon,”World Bank mimeo.

Ncube, Mthuli; Paul Collier, Jan Willem Gunning,and Kupukile Mlambo. 1998. “Trade Liberaliza-tion and Regional Integration in Zimbabwe”, inRegional Integration and Trade Liberalizationin Sub-Saharan Africa. Vol. 2: Country CaseStudies. A. Oyejide, Benno Ndulu, and JanWillem Gunning, eds. London: Macmillan.

Nehru, Vikram; Eric Swanson, and AshutoshDubey. 1995. “New Data Base on Human Capi-tal Stock in Developing Countries: Sources,Methodology and Results,” J. DevelopmentEcon., Vol. 46, pp. 379–401.

Nissanke, Machiko and Ernest Aryeetey. 1998. Fi-nancial Integration and Development: Liberali-zation and Reform in Sub-Saharan Africa. Lon-don: Routledge.

Odedokun, M.O. 1996. “Foreign Creditors’ Per-ception of Risks in Lending to African Coun-tries: Evidence from Interest Rate Spread,” De-vel. Policy Rev., 14, pp. 391–408.

Oehmke, James F. and Eric W. Crawford. 1996.“The Impact of Agricultural Technology in Sub-Saharan Africa,” J. African Econ., 5, pp. 271–92.

Oshikoya, Temitope W. 1994. “MacroeconomicDeterminants of Domestic Private Investmentin Africa,” Econ. Development and CulturalChange, Vol. 42, pp. 573–91.

Pattillo, Catherine. 1998. “The Impact of Uncer-tainty on the Investment Behaviour of Ghanaian

Manufacturing Sector Firms,” in Collier andPattillo 1998.

Pinckney, Thomas C. and Peter K. Kimuyu. 1994.“Land Tenure Reform in East Africa: Good,Bad or Unimportant,” J. African Economies,Vol. 3, pp. 1–28.

Platteau, Jean-Phillippe. 1994. “Behind the Mar-ket Stage where Real Societies Exist–Part I: theRole of Public and Private Order Institutions,”J. Development Stud., Vol. 30, pp. 533–77.

Posner, Richard A. 1980. “A Theory of PrimitiveSociety, with Special Reference to Law,” J. LawEcon., 23, pp. 1–53

Pradhan, Sanjay Kumar. 1996. “Evaluating PublicSpending,” World Bank Discussion Paper 323,Washington DC.

Pritchett, Lant. 1991. “Measuring Real ExchangeRate Instability in Developing Countries: Em-pirical Evidence and Implications,” Policy Re-search and External Affairs Working Paper No.791, World Bank.

Pritchett, Lant. 1996. “Where Has All the Educa-tion Gone?” World Bank Policy Research Work-ing Paper No. 1581.

Reardon, Thomas, P.M. Matlon and ChristopherL. Delgado. 1988. “Coping with Household-Level Food Insecurity in Drought-Affected Ar-eas of Burkina Faso,” World Devel., 16, pp.1065–74.

———. 1992. “Determinants and Effects of In-come Diversification amongst Farm Householdsin Burkina Faso,” J. Devel. Stud., 28.

Reardon, Thomas; A. Fall, V. Kelly, C. Delgado,P. Matlon, J. Hopkins, and O. Badiane. 1994.“Is Income Diversification Agriculture-Led inthe West African Semi-Arid Tropics?” Ch. 11 inEconomic Policy Experience in Africa. A. At-sain, S. Wangwe and A.G. Drabek, eds. Nairobi:African Economic Research Consortium.

Reinikka, Ritva. 1996. “The Credibility Problem inTrade Liberalisation: Empirical Evidence fromKenya,” J. African Econ., 5, pp. 444–68.

Rimmer, Douglas. 1990. Staying Poor: Ghana’sPolitical Economy 1950–90. Oxford: Pergam-mon.

Roberts, Mark J. and James R. Tybout. 1997.What Makes Exports Boom? Washington, DC:World Bank.

Rodrik, Dani. 1998. “Trade Policy and EconomicPerformance in Sub-Saharan Africa,” NBERWork. Paper No. 6562.

Sachs, Jeffrey D. and Andrew M. Warner. 1995.“Natural Resources and Economic Growth,”Development Discussion Paper 517a, HarvardInstitute Int. Devel., Cambridge, MA.

———. 1995a. “Economic Reform and the Pro-cess of Global Integration,” Brookings Pap.Econ. Act., 1995:1, pp. 1–118.

———. 1997. “Sources of Slow Growth in AfricanEconomies,” J. African Econ., 6, pp. 335–76.

Sahn, David, ed. 1994. Economic Crisis and PolicyReform in Africa: An Introduction. Ithaca andLondon: Cornell U. Press.

Sarel, Michael. 1996. “Nonlinear Effects of Infla-


tion on Economic Growth,” IMF Staff Pap., 43,pp. 199–215.

Seck, Diery and Yasim H. el Nil. 1993. “FinancialLiberalisation in Africa,” World Devel., 21, pp.1867–81.

Shipton, Parker MacDonald. 1985. “Land, Credit,and Crop Transactions in Kenya,” unpub. PhDthesis, Columbia U.

Shleiffer, Andrei and Robert Ward Vishny. 1993.“Corruption,” Quart. J. Econ., 108, pp. 599–617.

Singer, J.D. and M. Small. 1994. “Correlates ofWar Project: International and Civil War Data1816–92,” Ann Arbor, Michigan, Inter-Univer-sity Consortium Political and Social Research.

Stiglitz, Joseph E. 1974. “Alternative Theories ofWage Determination and Unemployment inLDCs: The Labor Turnover Model,” Quart. J.Econ., 88, pp. 194–227.

Teal, Francis. 1996. “The Size and Sources of Eco-nomic Rents in a Developing Country Manufac-turing Labour Market,” Econ. J., 106, pp. 963–76.

Temple, Jonathan. Forthcoming. “Initial Condi-tions, Social Capital and Growth in Africa,” J.African Econ.

Tsishimbi, wa Bilenga, with Peter Glick and ErikThorbecke. 1980. “Missed Opportunity for Ad-justment in a Rent-Seeking Society: The Caseof Zaire,” ch. 4 in David E. Sahn 1994.

Udry, Christopher. 1993. “Credit Markets inNorthern Nigeria: Credit as Insurance in a Ru-ral Economy,” in Hoff et al. 1993.

———. 1994. “Risk and Insurance in a RuralCredit Market: An Empirical Investigation inNorthern Nigeria,” Rev. Econ. Stud., 61, pp.495–526.

University of Gothenburg and University of Nai-robi. 1994. “Limitations and Rewards inKenya’s Manufacturing Sector: A Study of En-terprise Development,” mimeo.

University of Leuven, Catholic and University ofBurundi. 1994. “Regional Program on Enter-prise Development: First Report on the Bu-rundi Survey,” mimeo.

Velenchik, Ann D. 1997. “Government Interven-tion, Efficiency Wages, and the Employer SizeWage Effect in Zimbabwe,” J. Devel. Econ., 53,pp. 305–38.

Webb, Patrick and Thomas Reardon. 1991.“Drought Impact and Household Response inEast and West Africa,” Quart. J. Int. Agricul-ture.

Wheeler, David. 1984. “Sources of Stagnation inSub-Saharan Africa,” World Development, 12,pp. 1–23.

Widner, Jennifer A. 1998. “The Courts as Re-straints,” in Collier and Pattillo 1998.

Wood, Adrian and Jörg Mayer. Forthcoming. “Af-rica’s Export Structure in Comparative Perspec-tive,” in the UNCTAD series Economic Devel-opment and Regional Dynamics in Africa:Lessons from the East Asian Experience.

World Bank. 1994. Adjustment in Africa: Reforms,Results, and the Road Ahead. Oxford: Oxford U.Press.

World Bank. 1995. World Development Report.New York: Oxford U. Press.

World Bank. Forthcoming. World DevelopmentIndicators 1998. Washington, D.C.

Yeats, Alexander J. 1990. “On the Accuracy ofEconomic Observations: Do Sub-Saharan TradeStatistics Mean Anything?” World Bank Econ.Rev., 4, pp. 135–56.


Documents

Collier and Gunning: Explaining African Economic ... of Economic Literature Vol. XXXVII (March 1999), pp. 64–111 Collier and Gunning: Explaining African Economic Performance Explaining