THE WORLD BANK ECONOMIC REVIEW, VOL. 5, NO. 1: 155-175 FIL E CO 10"P Y ' Indigenous Land Rights Systems in Sub-Saharan Africa: A Constraint on Productivity? Shem Migot-Adholla, Peter Hazell, Benoit Blarel, and Frank Place This article uses cross-sectional evidence from Ghana, Kenya, and Rwanda in 1987-88 to examine the question, Are indigenous land rights systems in Sub-Saharan Africa a constraint on productivity? The evidence supports the hypothesis suggested by histori- cal studies, that African indigenous land rights systems have spontaneously evolved from systems of communal control towards individualized rights in response to in- creases in commercialization and population pressure. Cross-sectional data on the incidence of land improvements and on land yields provide little support for the view that limitations under indigenous law on the right to transfer land are a constraint on productivity. Within the context of Africa's rapid population growth and the need for in- creased productivity of land, there is a growing debate about whether the indige- nous land tenure systems are a constraint on agricultural transformation. Some authors, such as Dorner (1972), World Bank (1974), and Harrison (1987), see the indigenous tenure systems as static constraints on agricultural development, providing insufficient tenure security to induce farmers to make necessary land- improving investments. Others, however, such as Cohen (1980), Boserup (1981), Noronha (1985), and Bruce (1988), have countered that the indigenous tenure arrangements are dynamic in nature and evolve in response to changes in factor prices. In particular, it is argued that there is a spontaneous individualiza- tion of land rights over time, whereby farm households acquire a broader and more powerful set of transfer and exclusion rights over their land as population pressure and agricultural commercialization proceed. The issue is far from academic, because it brings into question the very need for expensive land registration and titling programs at this stage in the economic development of Sub-Saharan Africa. If the indigenous tenure systems are dy- namic, then it is relevant to ask if there are more useful things that governments can do to facilitate the process of adaptation. Up to now, the debate has been carried out without benefit of empirical tests of the performance of indigenous land tenure systems or of the effect of land The authors, except for Benoit Blarel, are in the Agriculture and Rural Development Department of the World Bank. Blarel is in the Bank's Africa-South Central and Indian Ocean Department. @ 1991 The International Bank for Reconstruction and Development / THE WORLD BANK. 155 156 THE WORLD BANK ECONOMIC REVIEW, VOL. 5, NO. I titliiig on agricultural output. This article takes a first step in providing empiri- cal testing of the relationship between indigenous tenure arrangements and agri- cultural productivity. New data are presented from a recently completed study of land rights systems in Ghana, Kenya, and Rwanda. We use the results from that study, together with historical evidence, to argue that, at least in rainfed cropping areas, indigenous African tenure systems have so far been flexible and responsive to changing economic circumstances. Where population pressure and commercialization have increased, the indigenous tenure systems have autono- mously evolved from a system of communal property rights towards one of individualized rights. Controlling for differences in land quality and farmer characteristics, there is at best a weak relationship between individualization of land rights and land yields in the regions surveyed. The article is organized as follows. The next two sections describe the charac- teristics and evolution of indigenous land rights systems in Sub-Saharan Africa. Section III presents results from our cross-sectional study in Kenya, Ghana and Rwanda. Section IV discusses government interventions in land tenure systems, and section V concludes with directions for policy. I. THE NATURE OF INDIGENOUS TENURE Africa has a wide variety of ecological conditions, cultural systems, and politi- cal structures. It is therefore somewhat dangerous to make general assertions about the nature of indigenous land rights systems. But some common features may still be discerned. Several interrelated themes have influenced the standard conceptualization of African land tenure systems and will be summarized here. Until recently, indigenous African land rights systems have been incorrectly presented by most foreign anthropologists, colonial administrators, and nation- alist idealists as static polar contrasts to Western property rights systems. It is asserted that indigenous tenure systems assign land rights to the community, and that communal control discourages long-term investment in land improvements. The argument is that individual farmers, not having secure private rights to the land, may not be able to claim fully the returns on their investment. To the extent that investments are required for conservation purposes, indigenous ten- ure arrangements will also potentially promote land degradation. Further, it is asserted that because land is an integral part of the social system and legitimate use is determined by birth, affinity, common residence, and social status or some combination of these, transactions are limited to the members of the lineage. This encumbers the emergence of market transactions in land in which access would be determined by supply and demand factors and entrepre- neurial ability. Contrasted to this picture is the idea that "modern" (implicitly "Western") property rights systems should be founded on principles of contrac- tual law and economic efficiency. But often ignored is the fact that, except in very isolated cases, communal control over land under indigenous tenure systems today occurs mainly in areas Migot-Adholla and others 157 characterized by relative land abundance and low intensification; but even then, farmers typically have secure use and inheritance rights. Historical records sug- gest that indigenous tenure had demonstrated remarkable flexibility in adapting to new farming technology or methods of exchange long before the colonial period (Morgan 1969; Hill 1963; Jones 1980; Bates 1986). The contrast between indigenous African tenure and Western property rights systems should be perceived not in terms of polar extremes but as points along a continuum, the particular location of which is determined by population pres- sure, and the degree of commercialization of agriculture. Evidence from differ- ent locations in Africa confirms instances of autonomous intensification and privatization of rights in land since the beginning of the century (Netting 1968; Lagemann 1977; Ruthenberg 1985). Section III of this article presents new cross-sectional evidence of evolution of indigenous land rights systems in Ghana, Kenya, and Rwanda. II. AGRICULTURAL INTENSIFICATION AND THE EVOLUTION OF INDIGENOUS LAND RIGHTS SYSTEMS Reduced to its bare essentials, agricultural intensification entails a multi- dimensional process of response to increasing population density, technological change, and commercialization, or to any combination of these. It is charac- terized by substitution of labor for land in the initial stages and a shift from forest fallow through bush and grassland fallow. This is followed by more continuous cropping and systems of crop rotation and soil improvement (includ- ing green and animal manures, and compost), followed by additional modern yield-enhancing inputs such as chemical fertilizers, insecticides, and high- yielding seed varieties (Boserup 1965; Geertz 1963; Netting 1968; Waddell 1972; Pingali and Binswanger 1986). Agricultural intensification has often occurred in response to the creation of new markets following the separation of consumption and reproductive objec- tives from those relating to production. In the context of African development, this process, which was described in Great Transformations by Polanyi (1944), may be identified with the transformation of land, labor, and capital into com- mercial commodities, particularly after the introduction of colonial rule. Al- though the onset of these transformations may have predated formal colonial rule in some parts of Africa and were probably autonomous, later changes during the colonial and post-colonial periods were either directly or indirectly the outcome of state intervention. Historically, African agrarian systems have been characterized by relative abundance of land and critical seasonal scarcities of labor. The relative abun- dance of land, and the need to shift the location of farm plots in order to regenerate soil fertility naturally, implied less concern at the collective level with guaranteeing rights over specific pieces of land and instead emphasis upon con- trol over large unutilized areas. It is still true of many areas today that individ- 158 THE WORLD BANK ECONOMIC REVIEW, VOL. 5, NO. 1 uals gain use rights by initial clearing or inheritance and maintain such rights through continuous exploitation, making allowance for suitable fallow periods. Thus, households and individuals are able to claim perpetual interests in farm- ing land and improvements in which they have invested labor or capital. These interests are inheritable and may be bequeathed to the next generation. Provided farmland is not under crops, it is also subject to secondary use (for example, for grazing) by members of the social group even though it is clearly not perceived to belong in the same category of common property as uncultivatable pasture land, forests, or fisheries. Although investment of labor or capital in land is often necessary to exclude others, the secondary right of exclusion is generally condi- tioned by the seasonality of farming operations. Agricultural intensification, which typically involves more continuous use of land, thus enhances the process of privatization of rights over land. One implication of social determination of rights to land is that, even during long periods of absence, eligible individuals retain claims to land within the territorial areas occupied by their kinship or residential groups. The ability of African urban labor migrants to fall back on the kinship network as a source of security may have contributed to relatively low formal sector wages and poor job security for many urban workers, which in turn perpetuate the need to maintain claims over rural land. So long as land is abundant and the technology of production remains land extensive, claims of rights to rural land by absentee migrant workers need not result in production inefficiencies. But as land be- comes scarce and absentee claimants seek to assert their rights through intermit- tent cultivation, it is likely that their plots would be farmed less intensively than others. However, this may not always be the case. By investing cash remittances from household members engaged in urban employment into agriculture, house- holds with nonfarm income sources can experience higher levels of productivity. The earliest individualization of a broad range of transfer and exclusion rights over land in Africa arose largely in response to the cultivation of commercial crops, primarily oil palm, cocoa, groundnuts, cotton and coffee (Hill 1963; Jones 1980; Moore 1986). After initial hesitation concerning the suitability of plantation farming, the colonial powers did not undertake any direct interven- tion to restructure indigenous land tenure arrangements in order to accommo- date these developments, except where there was significant white settlement. In Angola, Kenya, Mozambique, eastern Zaire, and Zimbabwe, for instance, land tenure legislation was enacted to provide freehold rights to individual European farmers and corporations involved in plantation agriculture. In addition, as was the case in Kenya until the mid-1950s, cultivation of certain profitable crops by Africans was prohibited or severely restricted. This, combined with other policy instruments such as forced labor or the imposition of a head tax payable in cash, helped to guarantee cheap labor to white farmers. Elsewhere, the most important policy support to agricultural commercializa- tion, and unintentionally to the individualization of land rights, was the con- struction of communications and transportation infrastructure. Later, other in- Migot-Adholla and others 159 struments such as price support, credit, and extension were used with fairly significant results in promoting agricultural commercialization. Based on the simplified sketch of the process of intensification as a result of the interplay between population pressure, technological changes, and agri- cultural commercialization outlined above, three broad categories of tenure regimes may be identified. At the earliest stage, characterized by the predomi- nance of pastoral and sylvan economies, all land is communally owned, the group of authorized users is clearly defined, and there are rules specifying their rights and obligations with respect to the land and its resources. Following Feder and Feeny in this issue, we identify such regimes as common property and distinguish them from open-access, a regime of unrestricted privilege but no duties. The latter appears to represent the breakdown of rules governing group management of common property resources. As population pressure increases, the period of fallow shortens and shifting cultivation is replaced by systems of rotation and soil improvement. These changes may also be precipitated by the introduction of commercial tree crop production, which tends to enhance rights of exclusion of individuals even though the basic control over outsiders' access to the land continues to be exercised by the community. Voluntary transfers of land to nonmembers are virtually nonexistent. It is prescribed that transfers are only to members of the landowning group and may be made only by bequests, gifts, or rent. But the latter two remain relatively rare, becoming significant mainly in areas of ex- treme land shortage or where there is high demand for land of particular types. Even then, prior approval may be required from leaders of the lineage or community. As increased intensification leads to virtual exclusion by farmers of other individuals, there may be localized exchanges in productive factors leading to land or labor tenancies and credit markets. But only rarely do individuals ac- quire the untrammeled right of disposal of land. This partly relates to the significance of land as a source of security for all members of the lineage, even for those who spend extended periods working away. The distinguishing feature of different tenure regimes may thus be said to revolve around restrictions on the individual holder's ability to transfer land (only among family members, within the lineage or community, or to outsiders; and with or without approval from other lineage or community members), which also tends to coincide with the mode of transmittal (inheritance, gifts or bequests, and sale). III. LAND RIGHTS, POPULATION PRESSURE, COMMERCIALIZATION, AND AGRICULTURAL PRODUCTIVITY: EMPIRICAL EVIDENCE The historical overview above suggests two key hypotheses that need to be tested. The first is that land rights do evolve toward full privatization in the presence of increasing commercialization and population pressure. The second is that as long as the evolution of land rights remains unrestricted, then, except 160 THE WORLD BANK ECONOMIC REVIEW, VOL. 5, NO. 1 in short-run situations of dramatic economic or social change, indigenous land tenure systems are not likely to be a constraint on agricultural productivity. Survey Profile To test these hypotheses, we undertook farm surveys in 1987-88 in a total of ten regions in Ghana, Kenya, and Rwanda. Ghana was chosen to represent a situation of relative land abundance, whereas Rwanda represents a case of land scarcity. Within Ghana, we have chosen a highly commercialized shallot grow- ing region with a dense, homogeneous population (Anloga), a land-abundant, cocoa-growing area with a significant migrant population (Wassa), and a land- abundant, food-growing area with a significant migrant population (Ejura). All three Rwandan regions are densely populated, and each produces a similar mix of crops (predominantly coffee, sorghum, and sweet potatoes). However, the commercial activity of farmers is noticeably greater in Ruhengeri than in either Butare or Gitarama. We included Kenya in the survey because it is the only Sub-Saharan African country with more than thirty years of experience of a national land registration program and thus provides a test case for land tenure reform. Within Kenya, two traditional and densely populated African areas were chosen, the maize- producing region of Madzu and the commercialized area of Kianjogu (coffee). Two post-independence resettlement areas (where farmers received about five hectares of land) were also selected. Lumakanda exhibits more commercial activity than does the other resettlement area of Mweiga. Each of the ten regions is dominated by agriculture, and few off-farm income-earning opportunities exist. More than 100 households were surveyed in each of the Ghanaian and Ken- yan regions and about 80 in each of the Rwandan areas. Among the variables measured were characteristics of the household (for example, number, age, education, and occupation of all individuals, number of regular workers, non- farm income, wealth, use of credit, and so on), characteristics of the household head (for example, farming experience, place of birth, office holding in local organizations), characteristics of the farm (for example, farm size and number of parcels), and characteristics of the parcels (mode of acquisition, soil fertility, parcel size, rights of transfer and use, land improvements made, type of docu- ment held, incidence of dispute, and, in the last season, the crops grown, inputs used, and output). To provide an overview of tenure regimes predominant in our study regions, table 1 displays the percentage of parcels acquired under various methods. In the Rwandan regions and in Wassa and Ejura, numerous methods of acquisition are common. In Anloga and the Kenyan regions, nearly all parcels are acquired by three or fewer methods. The majority of parcels in all regions are acquired through nonmarket channels such as inheritance, gift, government allocation, and appropriation (initial clearing and use of part of the pool of communal land). Inheritance is by far the most common, whereas appropriation is becom- Migot-Adholla and others 161 ing rare as unused land disappears. Purchases of land are much less common but account for at least 17 percent of operated parcels in Madzu, Lumakanda, Wassa, and Ruhengeri. Markets for leaseholds (fixed rentals, sharecropping arrangements, and pledges') are relatively rare in all regions except the Anloga region of Ghana. Although one would perhaps expect more land transactions, especially given that in many of the regions commercial crops are produced and land is relatively scarce, the lack of off-farm income opportunities greatly re- duces the willingness of households to alienate land. Effect of Population Pressure, Commercialization, and Government Intervention on the Individualization of Land Rights Data were collected on whether a head of household could exercise numerous rights over particular parcels of land and whether, to exercise some of these rights, prior approval from another individual or institution was required.2 The specific land rights enumerated for each parcel included use rights (rights to grow annual crops, to grow perennial crops, to be buried, to make permanent improvements, and to collect firewood, among others) and transfer rights (rights to lend, rent, pledge, mortgage, give, bequeath, and sell). The first three rows of table 2 categorize land parcels according to the breadth of accompanying trans- fer rights. The first group, "limited transfer" land, includes those parcels for which the farmer has no permanent transfer or alienation rights, but may have some temporary transfer privileges. Second, the "preferential transfer" category describes parcels that may be permanently transferred but only within the family or lineage (that is, through gift or bequest). Third, "complete transfer" lands are those that may be alienated outside the lineage through the right to sell.3 This simple classification scheme produced mutually exclusive groups, each compris- ing parcels which were very homogeneous with respect to the breadth of rights, with more land rights found on "complete transfer" parcels. The data in table 2 provide the basis for simple tests of the effect of population pressure and agricultural commercialization on the individualization of land rights. The effect of population pressure is best tested by comparing regions in Ghana. Although each Ghanaian region is fairly commercialized (Anloga, shal- lots; Wassa, cocoa; Ejura, large food crop farms), only Anloga experiences high population pressure. From the sum of the fifth and sixth rows of table 2, we see that for permanently held land, the percentage of "complete transfer" parcels is 82.5 percent in Ejura, 76.7 percent in Anloga, and 70.3 percent in Wassa. If the necessity of approval is also taken into atcount, however, the results change 1. Pledges are possessory mortgages in which the lender of cash takes possession of the land for a specified period and earns, as interest, the proceeds from the land. 2. Farmers' perceptions about who they must clear land transfers with often do not correspond with legal requirements, but the data we report here are respondents' perceptions of their rights. 3. There may still be restrictions on transferring land to members from other tribes or cultural groups, but these are much less constraining than restrictions on other transfers at the family or immediate lineage level. Table 1. Methods of Acquisition of Operated Parcels in the Study Regions (percent) Ghana Rwanda Kenya Luma- Mode of acquisition Anloga Wassa Ejura Ruhengeri Butare Gitarama Madzu Kanda Kianjogu Mweiga Purchased 0.9 18.0 6.3 17.4 3.2 8.1 28.6 26.0 13.2 6.0 Inherited 57.5 6.3 22.1 46.6 43.4 49.5 68.3 31.3 82.1 35.0 Given 0.8 44.2 43.1 8.3 5.6 4.1 0.0 0.0 0.0 1.0 Appropriated 0.0 22.1 14.7 0.5 3.7 1.5 0.0 0.0 0.0 0.0 Allocated from government 0.0 0.0 0.0 4.0 4.9 15.6 0.0 33.6 1.9 58.0 Other permanent acquisitions 0.0 1.7 0.4 6.3 4.1 0.2 0.8 1.6 0.0 0.0 Rented 21.5 7.0 9.6 7.4 15.7 7.9 1.5 7.6 1.9 0.0 Pledged 19.3 0.6 0.4 0.0 0.0 0.0 0.0 0.0 0.0 0.0 Borrowed 0.0 0.1 0.0 9.5 19.4 13.2 0.8 0.0 0.9 0.0 Squatter land 0.0 0.0 3.7 0.0 0.0 0.0 0.0 0.0 0.0 0.0 Source: World Bank data, available for a nominal reproduction charge upon written request to the author. Table 2. The Prevalence of Land Rights across Study Regions (percent) Ghana Rwanda Kenya Luma- Land right Anloga Wassa Ejura Rubengeri Butare Gitarama Madzu Kanda Kianjogu Mweiga All parcels Limited transfer 52.4 6.0 21.0 15.5 36.7 21.1 26.2 9.9 23.6 53.0 N Preferential transfer 2.1 29.1 6.6 3.1 16.5 21.3 6.3 14.5 67.9 10.0 Complete transfer 45.4 64.9 72.4 81.5 46.7 57.6 67.5 75.6 8.5 37.0 Permanently held parcels No right to sell 23.3 29.7 17.5 2.4 28.2 22.4 30.9 18.2 91.3 63.0 Right to sell with approval 14.0 55.6 73.5 19.7 37.3 10.6 39.0 51.2 4.9 4.0 Right to sell without approval 62.7 14.7 9.0 77.8 34.5 67.0 30.1 30.6 3.9 33.0 Source: World Bank data, available for a nominal reproduction charge upon written request to the author. 164 THE WORLD BANK ECONOMIC REVIEW, VOL. 5, NO. 1 considerably. The percentage of permanently held parcels which may be sold without approval is much higher in Anloga (62.7 percent) than in either Wassa (14.7 percent) or Ejura (9.0 percent). If the ability to transfer land without approval is indeed the paramount measure of the individualization of land rights, then the Ghanaian data strongly support the notion that increased popu- lation pressure brings forth a higher degree of privatization of land rights. Our best basis for testing the effect of commercialization on the privatization of land rights is Rwanda. In Rwanda, each of the three regions is characterized by high population density, but Ruhengeri is the most commercialized and Bu- tare is the least. In Ruhengeri, 97.5 percent of permanently held parcels are "complete transfer" lands as opposed to 77.6 percent in Gitarama and 71.8 percent in Butare. If approval to sell is considered, the Butare region exhibits by far the lowest percentage of freely saleable parcels. These patterns provide sup- port for the argument that increased commercialization hastens the individualiz- ation of land rights. Our Kenyan regions do not provide the basis for such controlled tests. Madzu and Kianjogu differ somewhat both in population density and in degree of commercialization, with Madzu being denser and less commercialized.4 Because land rights are more individualized in Madzu, if all else were equal, it suggests that population pressure may have been more influential than commercializa- tion on land rights. The Kenyan regions of Madzu and Kianjogu can also be used to test the effectiveness of government land policies, in this case a land registration and titling program, on the individualization of land rights. Of the 97 parcels with land titles in the two regions, only 23 can be sold by the current operator (23.7 percent) and only 24.5 percent of 94 parcels which may be sold are also titled. The apparent persistence of indigenous control over land transfers demonstrates the difficulty in altering custom by government decree. Another example of the ineffectiveness of government intervention is in Rwanda, where despite laws forbidding land sales without government authorization, farmers claim the right of sale over most of the sampled parcels. Land Rights and Agricultural Productivity Following Feder and others (1988), the relationship between land rights and productivity is hypothesized to proceed as follows. Increased individualization of rights improves farmers' abilities to reap returns from investments on land. This leads to a greater demand for land improvements as well as for complemen- tary inputs. Increased individualization of rights may also improve the credit- worthiness of the farmer and enhance his chances of receiving formal credit. Both of these demand- and supply-side mechanisms interact to increase invest- ments in land and input use, which in turn lead to greater land productivity. 4. The Lumakanda and Mweiga regions are less useful for testing purposes because they have been settled only since the 1960s, and thus sufficient time has not passed to evaluate the evolution of land rights. Migot-Adholla and others 165 Table 3. Aspects of Credit Markets in the Study Regions Percentage of households receiving: Formal sector loans Percent secured Country and region Any credit Formal credit Number with land Ghana Anloga 47.8 37.4 43 19.S Wassa 22.0 12.7 19 26.3 Ejura 5.7 4.4 7 0.0 Rwanda Ruhengeri 41.0 0.0 0 n.a. Butare 12.7 0.0 0 n.a. Gitarama 13.8 8.8 12 41.7 Kenya Madzu - 4.0 5 50.0a Lumakanda - 10.7 13 69.2 Kianjogu - 8.7 9 0.0 Mweiga - 1.0 1 100.0 - Not available. n.a. Not applicable. Note: The figures for Ghana are for 1987; for Rwanda and Kenya, they represent the 1987-88 period. a. The type of collateral was not reported by the respondents for all loans. Source: World Bank data, available for a nominal reproduction charge upon written request to the author. Credit. The use of formal credit is limited in the study regions, reflecting the poor development of formal rural banking institutions. Table 3 shows that less than 13 percent of the farms received formal credit during 1987-88 in nine of the ten study regions. The Anloga region of Ghana was the only exception- 37.4 percent of households received formal credit there. Furthermore, in Ghana and Rwanda, all formal credit loans were short-term, none being extended for more than one year. It is therefore not surprising to find a weak relationship between land rights and the use of formal credit. For instance, we did not find any significant relationships between the use of formal credit and the proportion of land held with "complete transfer" rights. Given the low incidence of formal borrowing, the absolute number of cases in which land is used to secure loans is very low (see table 3), and in fact none were reported in four of the regions. Group loans, where a group of borrowers is jointly responsible for repayment in lieu of collateral, are common in obtaining formal credit in Anloga. In each region, loan maturities averaged under one year. Although this may be partly influenced by lack of formal collateral, it is characteristic of banks not to favor long-term lending. The case of Kenya differs considerably from that of Ghana or Rwanda in that formal titles to land are held by many farmers and could be obtained by count- less others. However, we did not find a significant relationship between the possession of title and use of formal credit. The use of land titles also did not imply an increase in loan maturity or loan size. For instance, among all formal 166 THE WORLD BANK ECONOMIC REVIEW, VOL. 5, NO. I sector loans, the average maturity of loans secured by land was 24.1 months, but those secured by other means (cosignatory, group guaranty, and agricultural produce) averaged 31.5 months in length (the difference is not statistically sig- nificant). In light of these findings, we conclude that there is little relation between land rights or land title and the use of formal credit. The low incidence of formal credit in Kenya also suggests that land titles alone will not lead to the development of active rural credit markets. Land improvements. In terms of their effect on investment in land improve- ments, land rights have mixed results.5 The ability to bequeath land is the important distinguishing right in each Rwandan prefecture. Parcels which can- not be bequeathed ("limited transfer" parcels) are much less likely to be im- proved by farmers in any manner or with long-term investments.6 For example, 78.7 percent of parcels which may be bequeathed were improved as opposed to 26.7 percent for those which could not be bequeathed. Among permanently held parcels, there is no difference in the incidence of land improvements between "preferential transfer" and "complete transfer" land, nor does the requirement of prior approval matter. In Anloga (only permanently held parcels were surveyed for investments), 61.8 percent of "complete transfer" parcels were improved (by drainage, mulching, or excavation) as opposed to only 5.4 percent of "limited transfer" parcels. Moreover, the parcels which could be transferred freely were more likely to have been improved than those requiring prior approval. In Ejura and in Wassa, the incidence of investment was not related to land rights. A lack of relationship was also found for each Kenyan region. From these varied find- ings, it is not possible to make any general assertions regarding the effect of land rights on land improvements.7 Productivity. To test the relationship between land rights and land produc- tivity, we estimated reduced-form yield equations using parcel level data for selected crops." In general form, the equation is: Y = f(X,, Xp, S) + e, where Y = yield, measured as the gross value per hectare of all crops harvested on the parcel; Xh = a vector of household characteristics (to capture resource and skill levels); X, 5. In Rwanda, numerous types of land improvements were classified as either boundary, short-term, or long-term. In Kenya, only terracing, drainage, and tree crop improvements were examined. In Anloga, the improvements were drainage, mulching, and excavation. In Wassa, the only improvements made by farmers were tree crops. Finally, in Ejura, tree crops and destumping investments were analyzed. 6. The difference is most marked when the improvements made by the current operators are com- pared. But land that cannot be bequeathed by the current operators also has less accumulation of improvements, irrespective of who made them. 7. For a more sophisticated analysis of these relationships, see Place, Hazell, and Lau (1990). 8. Yield regressions were confined to the more important crops in each region, and we tried to avoid crops that were extensively intercropped. Production was physically measured at harvest in Rwanda but was obtained by farmer recall elsewhere. Because the major crops were mostly cash crops (for example, cocoa and shallots in Ghana), farmers also had less trouble recalling the amounts produced. Migot-Adholla and others 167 = a vector of parcel characteristics (to capture size and land quality differences); S = a set of dummy variables for the land rights categories; and e = an error term. In estimating this equation, we need to be sure that X and the land rights S are not correlated with the error term; otherwise the estimated coefficients would be biased and inconsistent. There are two possibilities to worry about. The first is that the choice of land parcels (and hence X, and S) are determined simultaneously with yields. Given that most parcels were acquired many years ago and that transfer rights, which are the basis of our land rights categories, are virtually predetermined by the mode of acquisition,9 there is little argument for a causal relationship from Y to S or X,,. It is more appropriate to view farmers' decisions as recursive; first they select parcels and then they farm them. A more worrying possibility is that there may be unobservable variables which affect both the choice of parcels and current yields, even though decisions are made at different points in time (this will lead to the collapse of the recursive framework). These unobservables may be household effects (for example, man- agerial skill) or parcel effects (for example, quality of soil). Because we have several parcel observations for each household, we are able to correct for unobservable household effects using either dummy variables (fixed household effects) or an error components model. Unfortunately, we do not have multiple observations on each parcel and cannot treat parcel effects in the same way. But we are able to control for a number of parcel characteristics, including soil fertility, which are included in X,. Given the numerous observable parcel quality controls, it is doubtful that unobservable parcel quality is signifi- cant in either the determination of yields or the choice of land rights and X,.10 We found no relationship between land rights and parcel yields in Kenya and Ghana (see table 4). This is true in both the fixed effects (where feasible) and error component models. In additional regressions, we also found that the mode of acquisition had no effect on parcel yields. But this is to be expected because there are strong correlations between the mode of acquisition and the land rights held. In Gitarama and the pooled Rwandan sample, we found that 'short-term use rights"' parcels were more productive than parcels in all other land rights categories. This perverse result can be explained as a multicollinearity problem. Land rights categories are highly related to the year of acquisition in Rwanda, primarily because 'short-term use rights' parcels are almost always rented and 9. Although not reported here, cross-tabulations of land rights categories against mode of acquisition show a high correlation. This is not to deny that land rights evolve over time, but the changes probably occur when land changes hands. Moreover, evolutionary changes will not quickly lead to the reclassifica- tion of parcels from one of our broad land rights categories to another. 10. We also find no evidence of any relationship between parcel quality variables and the land rights categories. 11. "Short-term use rights" parcels are those "limited transfer" parcels whose use rights are restricted to a specified number of seasons (mainly rented parcels). 168 THE WORLD BANK ECONOMIC REVIEW, VOL. 5, NO. 1 Table 4. Summary of Parcel Yield Regression Results Complete Complete transfer transfer Regional Preferential Complete with without regression transfer transfer approval approval Other Ghana Error component Anloga -0.348 -0.012 Wassa 0.211 -0.106 Ejura 0.441 -0.082 Fixed dummy Ejura 0.302 -0.196 Long-term Rwandaa use rights Error component Pooled Rwanda -0.210 -0.362** -0.234* Ruhengeri -0.148 Butare -0.236 -0.319 -0.285 Gitarama 0.001 -0.683** Fixed dummy Pooled Rwanda -0.306 -0.527** -0.320* Ruhengeri -0.122 Butare -0.327 -0.382 -0.344 Gitarama 0.156 -0.847* Kenya Land titleb Error component Madzu 0.146 0.051 0.283 Lumakanda -0.134 -0.149 -0.142 Kianjogu 0.127 0.337 0.155 Mweiga 0.580 0.223 -0.167 *Significant at a 10 percent level. *Significant at a 5 percent level. Note: The most limited land rights group (that is, "limited transfer rights") is omitted from the regression and is the base against which all included land rights variables are compared. Other explanatory variables included in the regressions but not reported here are (1) parcel characteris- tics: size, fertility, topographical location, slope, distance to house, year acquired, prior improvements, documents held, and cropping pattern; (2) household characteristics: age, education, occupation, gender, and farming experience of the household head; (3) farm characteristics: size, level of fragmentation, wealth, size of household, and nonfarm income; village where household resides. a. The Rwandan results are from the model containing the years since acquisition variable. See the text for a discussion of the results for the land rights variables once years since acquisition is removed. b. The land title coefficients are from separate regressions which exclude the land rights variables. Source: World Bank data, available for a nominal reproduction charge upon written request to the author. hence have been held for much less time than parcels acquired under other modes. Moreover, the time since acquisition variable is positively and signifi- cantly related to yield, and when this variable is dropped in a separate regres- sion, the results show that the more secure land rights categories are no longer significantly less productive than 'short-term use rights' parcels. The negative but insignificant coefficients remain, but they might be explained by differences in households who rent in land. Farmers who rent in land may generally be in Migot-Adholla and others 169 dire need of land resources and apply greater amounts of labor in order to provide subsistence for their families. We also tested the effect of land title on productivity in Kenya by including it alongside the land rights variables and by itself in separate regressions. In all cases, the possession of land title was not significantly related to yields. This is likely explained by the limited use of credit in the Kenyan study regions. IV. IMPACT OF RECENT GOVERNMENT INTERVENTIONS IN LAND TENURE Individual Tenure Perhaps the most comprehensive case of land tenure change aimed at estab- lishing alienable individual rights is that initiated in Kenya during the 1950s and continued to the present date with some modifications. Although there is agree- ment that Kenya has enjoyed among the highest productivity increases in Africa during this period, the causal relationships between individual tenure and in- creased agricultural output have been seriously questioned. Many believe that the removal of prohibitions against Africans to produce high value commodities (tea, coffee, and dairy products), coupled with substantial investment in com- munication and transport infrastructure during the 1950s and 1960s, improve- ments in extension services, and establishment of credit institutions, have been more important than changes in land tenure arrangements (Heyer, Maihta, and Senga 1976; Anthony and others 1979; Migot-Adholla 1985). Indeed, this ap- pears plausible when account is taken of the substantial increase in the produc- tion of these crops (cocoa, coffee, tea, and oil palm) in countries such as Cam- eroon, C6te d'Ivoire, and Ghana, which have not undertaken any fundamental legal tenure change. Critics of the Kenyan tenure reform have faulted it on grounds of increased landlessness. It should be realized, though, that an important objective of the program was concentration of land in the hands of the more efficient producers (Swynnerton 1954). Given Kenya's failure to industrialize and its very limited land resources, the current level of landlessness is as much a product of the exceptionally high rates of population growth as it is a reflection of the existing land tenure system. Paradoxically, one consequence of the emergence of the nation-state in Africa, and particularly colonial and post-colonial reinterpretations of tribal authority and "indigenous" tenure, has been the freezing of ethnic boundaries, which has restricted opportunities for expansion of extensive land use. Where marginal land has already been utilized, output may be increased only by intensification. In such situations, we have argued that individualized transferable rights of land use will have evolved, even without introduction of formal registration. But given the centrality of ethnicity in national politics, it is inconceivable that a national land market would evolve merely as a result of the introduction of privatized tenure system. Evidence from other studies of Kenya indicates that although there is a weak market in land nationally, it is more severely restricted 170 THE WORLD BANK ECONOMIC REVIEW, VOL. 5, NO. I in the former African reserve (Non-Scheduled Areas), where it operates mainly among members of the same ethnic group. In contrast, a significant level of transactions in land occurs in the former white settled (Scheduled) areas and in urban peripheries, where individuals are not bound by strong kinship identity. The consequence is that titles to agricultural land are perceived to have a greater commercial value within the urban periphery and former scheduled areas, than in the former African reserves, largely because of difficulties in enforcing con- tracts. For although some banks have accepted titled land as collateral and auctioned it off in cases of default, in some cases purchasers were not able to take occupation of the land for fear of reprisals. Thus, rather than stimulating a land market in which more efficient farmers acquired land, this situation had created a market in land titles which were often used to secure loans for non- agricultural investment (Okoth-Ogendo 1986). But this development has been complicated even further by recent legislative amendments which require that disputes over land be referred to elders in the first instance-a condition which underlines official ambivalence with respect to full implementation of the provi- sions of the Registered Land Act for fear of promoting dispossession of poor and indigent peasants by their richer neighbors. More generally, our results on credit access and productivity, as well as those from parallel studies undertaken by the Land Tenure Center of the University of Wisconsin (1990), suggest that large-scale land registration and titling programs are unlikely to be economically worthwhile for much of Sub-Saharan Africa at this stage of its economic development. But there are circumstances when titling might be worthwhile, for example: * When the indigenous tenure systems are absent or very weak: this is fre- quently the case in land settlement areas, but it can also arise elsewhere following periods of major economic or political upheaval, particularly if traditional lines of authority have been severed * In areas where the incidence of land disputes is high: this may occur in areas where large numbers of migrants or strangers have settled and established rival claims to land owned by indigenous groups * Where major project interventions are planned that either require full priva- tization of land rights for their success, or are likely to weaken the land rights of some vulnerable groups: some irrigation and tree crop projects provide good examples. Where tenure reform through land registration is to be undertaken, it is important to recognize the possibly overlapping rights of different individuals on the same piece of land. For example, herdsmen may have dry-season grazing rights on land that is cropped by others. If some of these rights are lost without adequate compensation, vulnerable groups may lose an important source of livelihood. The rights of access to land by all lineage members, even when working or resident elsewhere, has also been shown to provide an important element of security, particularly in old age, as well as being an important factor Migot-Adholla and others 171 in explaining the low incidence of landlessness in many heavily populated rural areas of Sub-Saharan Africa. Again, loss of these rights without adequate com- pensation could prove economically disastrous for some vulnerable groups. Land registration programs that attempt to record and protect the full range of customary rights run the risk of ossifying land tenure institutions and reducing their ability to evolve with economic circumstances. Where titling is required, it seems better to seek full privatization of land rights but within the context of an adequate system of compensation for those who lose some or all of their rights over land. Strengthening Indigenous Land Tenure The indigenous land tenure systems appear to be adapting efficiently to changes in relative factor scarcities. This is reflected in the emergence of markets for the sale and rental of land, and in the trend toward increased privatization of land rights. For this process to continue, it is important that governments not impose unnecessary restrictions on the ownership or transfer of land. For exam- ple, restrictions on land sales and rentals are often legislated in the belief that this will prevent excessive concentration of land amongst the rich and the dis- possession of the poor. In practice, this concern is rarely justified, and it might anyway be more easily addressed through farm-size ceilings without inhibiting land markets and preventing more efficient allocations of land between farmers with different management skills or factor endowments. Some governments (for example, those of Botswana, Nigeria, and Swaziland) are also unnecessarily restricting land transactions through policies that seek to "retribalize" land in the pursuit of rather nostalgic idealizations of African rural society (Cohen 1980; Bruce 1988). Rather than restricting land markets, governments should create an "en- abling" legal and institutional environment for more efficient transactions. This might entail establishing a voluntary system to simply record the details of land transactions and the interests of the different parties, and providing or reinforc- ing channels for the enforcement of all duly recorded contractual arrangements. These kinds of interventions could go a long way toward resolving many of the disputes that arise over land, particularly in areas where significant migrant or stranger farmers have settled, yet would be much cheaper than formal registra- tion and titling programs. V. CONCLUSION We have argued that the contrast between indigenous African tenure and Western property rights systems should be perceived not in terms of opposite extremes but as points along a continuum between communal rights systems and privatized rights systems. In response to population pressure, agricultural com- mercialization, and technological change, indigenous African tenure systems have moved along that continuum in the direction of greater individualization of 172 THE WORLD BANK ECONOMIC REVIEW, VOL. 5, NO. 1 land rights. This conclusion is supported not only by historical evidence, but also by our cross-sectional study of land rights in Ghana, Kenya, and Rwanda. Controlling for differences in land quality and household characteristics, our regression analysis indicates no relationship between cross-sectional variations in land rights and productivity. Thus, it undermines the conventional view that land rights are a constraint on productivity. Our study also indicates that land titling is not sufficient to increase access to formal credit; land as collateral is of little value in the study areas of Kenya, primarily because land transfers to outsiders through sale (or foreclosure) are not always recognized as legitimate. Moreover, in the absence of better nonfarm opportunities, farmers are reluctant to mortgage land. We are unable to determine precisely the degree of efficiency with which indigenous systems in our study regions have evolved. We do find some success stories associated with indigenous tenure systems: multiple cropping and high use of inputs in Anloga, adoption of cocoa production in Wassa, use of mecha- nization in Ejura, and adoption of some high-yielding varieties and soil conser- vation improvements in Rwanda. However, it is difficult to judge whether the pace of these changes could have been hastened by government intervention in land tenure arrangements. Evidence from Boserup (1981) and Pingali and others (1987) suggests that without accompanying improvements in infrastructure, rural health and education, and price incentives, the rate of change in agri- cultural intensification and technological change would be constrained under any tenure regime. Needless to say, our results are subject to some qualifications. The most important is that our study regions are, like much of Sub-Saharan Africa, poorly endowed with physical infrastructure and effective credit and marketing institu- tions. Factor markets for land, labor, and capital are also poorly developed and the available technologies have mostly been stagnant for quite some time. Within this context, one should not expect to find much new investment in land improvements or much use of modern inputs, even when full land rights are assured. Not only will the demand for investments and modern inputs be muted, but even where it exists, it may not be realized because of failures on the supply side for credit and inputs. As new technologies become available, credit, input, and product markets improve, and rural infrastructure is developed, then more significant relationships between tenure security and land productivity may be- gin to emerge. A second limitation of the study is that we covered only areas of rainfed agriculture. Questions remain about the suitability of indigenous land rights for irrigated farming, extensive pastoral and livestock systems, and communal for- estry areas. A third, more subtle limitation of our argument is that it extrapolates from past performance of indigenous land rights to future evolution. These systems are predicated on the solidarity of small-scale social entities. These entitles may cease to function if national institutions or national and international mobility Migot-Adholla and others 173 of labor destroy their legitimacy, or if loyalties to lineage and communal groups become weak. Given the wide variety of African rural economies and politics and differences in resource endowments, a pragmatic approach that promotes the adaptability of existing land tenure institutions appears preferable to radical reform either of an individualist or collectivist type. But this is not to argue for the reinvention of an idealized African past. Rather, it suggests intervention only where demand for change has been proved genuine. In the case of land registration, such a demand should arise following increased commercial agricultural opportunities and the acceptance of the legitimacy of impersonal forms of land transfers. In the meantime, governments have much to do in providing rural infrastructure, promoting market efficiency, and disseminating information about production technology, which are the real constraints on agricultural productivity. REFERENCES Anthony, K. B., R. M. Bruce, F. Johnston, W. 0. Jones, and V. C. Uchendu. 1979. Agricultural Change in Tropical Africa. Ithaca, N.Y.: Cornell University Press. Bates, Robert H. 1986. "Some Contemporary Orthodoxies in the Study of Agrarian Change." In Atul Kohli, ed., The State and Development in the Third World. Prince- ton, N.J.: Princeton University Press. Berry, Sarah. 1985. Fathers Work for Their Sons: Accumulation, Mobility and Class Innovation in a Yoruba Community. Berkeley: University of California Press. Binswanger, Hans, and Mark Rosenzweig. 1986. "Behavioral and Material Determi- nants of Production Relations in Agriculture." Journal of Development Studies 22, no. 3: 503-39. Boserup, Ester. 1965. The Conditions of Agricultural Growth. Chicago: Aldine. _. 1981. Population and Technological Change. Chicago: University of Chicago Press. Bruce, John. 1988. "A Perspective on Indigenous Land Tenure Systems and Land Con- centration." In R. E. Downs and S. Reyna, eds., Land and Society in Contemporary Africa. Hanover, N.H.: University Press of New England. Chanock, Martin. 1985. Law, Custom and Social Order: The Colonial Experience in Malawi and Zambia. Cambridge, England: Cambridge University Press. Cohen, John. 1980. "Land Tenure and Rural Development in Africa." In R. Bates and M. Lofchie. Agricultural Development in Africa. New York: Praeger. Dorner, Peter. 1972. Land Reform and Economic Development. Harmondsworth, En- gland: Penguin. Feder, Gershon, Tongroj Onchan, Yongyuth Chalamwong, and Chira Hongladaron. 1988. Land Policies and Farm Productivity in Thailand. Baltimore, Md.: Johns Hop- kins University Press. Geertz, Clifford. 1963. Agricultural Involution: The Process of Ecological Change in Indonesia. Berkeley: University of California Press. Gershenberg, T. 1971. "Customary Land Tenure as a Constraint on Agricultural Devel- opment: A Re-evaluation." East Africa Journal of Rural Development 4: 51-62. 174 THE WORLD BANK ECONOMIC REVIEW, VOL. 5, NO. 1 Harrison, Paul. 1987. The Greening of Africa, London: Paladin Grafton Books. Heyer, Judith. 1981. "Agricultural Development Policy in Kenya from the Colonial Period to 1975." In J. Heyer, P. Roberts, and G. Williams, eds., Rural Development in Tropical Africa. London: Macmillan. Heyer, Judith, J. Maitha, and W. Senga. 1976. Agricultural Development in Kenya. Nairobi, Kenya: Oxford University Press. Hill, Polly. 1963. Migrant Cocoa Farmers of Southern Ghana: A Study in Rural Capital- ism. Cambridge, England: Cambridge University Press. Hyden, Goran. 1980. Beyond Ujamaa in Tanzania: Under-development and an Uncap- tured Peasantry. Berkeley: University of California Press. Jones, W. 0. 1980. "Agricultural Trade within Tropical Africa: Historical Background." In R. H. Bates and M. E Lofchie, eds., Agricultural Developments in Africa: Issues of Public Policy. New York: Praeger. Lagemann, J. 1977. Traditional African Farming Systems in Eastern Nigeria. Afrika- Studien 98, Munich: Ifo-Institut, Weltforum-Verlag. McHenry, Dean E. 1976. "The Ujamaa Village in Tanzania: A Comparison with Chi- nese, Soviet and African Experience in Collectivism." Comparative Studies in Society and History 18 (July): 347-70. Migot-Adholla, S. E. 1985. "Rural Development Policy and Equality." In Joel D. Bar- kan, ed., Politics and Public Policy in Kenya and Tanzania. New York: Praeger. Moore, Sally Falk. 1986. Social Facts and Fabrications: "Customary" Law on Kiliman- jaro, 1880-1980. Cambridge, England: Cambridge University Press. Morgan, W. B. 1969. "Peasant Agriculture in Tropical Africa." In M. E Thomas and Whittington, eds., Environment and Land Use in Africa. London: Methuen. Netting, Robert. 1968. Hill Farmers of Nigeria: Cultural Ecology of the Kofyar of Jos Plateau. Seattle: University of Washington Press. Noronha, Raymond. 1985. "A Review of the Literature on Land Tenure Systems in Sub- Saharan Africa." Report ARU 43. World Bank Agriculture and Rural Development Department. Washington, D.C. Processed. Okoth-Ogendo, H. W. 0. 1986. "The Perils of Land 'Tenure' Reform." In J. W. Arntzen, L. D. Ngcongco, and S. D. Turner, eds., Land Policy and Agriculture in Eastern and Southern Africa, Tokyo: United Nations University. Pingali, P. L., and Hans P. Binswanger. 1986. "Population Density, Market Access, and Farmer Generated Technical Change in Sub-Saharan Africa." ARU Report 58. World Bank Agriculture and Rural Development Department. Processed. Washington, D.C. Pingali, Prabhu, Hans P. Binswanger, and Yves Bigot. 1987. Agricultural Mechanization and the Evolution of Farming Systems in Sub-Saharan Africa. Baltimore, Md.: Johns Hopkins University Press. Place, Frank, Peter B. R. Hazell, and Lawrence J. Lau. 1990. "An Econometric Analysis of the Efficiency of Indigenous Land Tenure Systems in Sub-Saharan Africa." World Bank Agriculture and Rural Development Department. Processed. Washington, D.C. Polanyi, Karl. 1944. Great Transformations. New York: Holt, Rinehart and Winston. Richards, Audrey I., Ford G. Sturrock, and Jean M. Fort, eds. 1973. Subsistence to Commercial Farming in Present-day Buganda: An Economic and Anthropological Survey. Cambridge, England: Cambridge University Press. Ruthenberg, Hans. 1985. Innovation Policy for Small Farmers in the Tropics: The Migot-Adholla and others 175 Economics of Technical Innovations for Agricultural Development. Oxford, England: Clarendon Press. Swynnerton, R. J. M. 1954. A Plan to Intensify the Development of African Agriculture in Kenya. Nairobi, Kenya: Government Printer. Uchendu, Victor C. 1970. "The Impact of Changing Agricultural Technology on African Land Tenure." Journal of Developing Areas 4 (July) 9: 477-85. University of Wisconsin. 1990. "Security of Tenure in Africa.' Land Tenure Center of University of Wisconsin-Madison. Madison. Processed. Waddell, Eric. 1972. The Mound Builders: Agricultural Practices, Environment, and Society in the Central Highlands of New Guinea. Seattle: University of Washington Press. Woodman, Gordon R. 1987. "How State Courts Create Customary Law in Ghana and Nigeria." In B. W. Morse and G. R. Woodman, eds., Indigenous Law and the State. Dordrech, Netherlands: Foris Publications. World Bank. 1974. "Land Reform:' World Bank Development Series. Washington, D.C. Processed.