Journal of Development Economics xxx (2010) xxx–xxx

DEVEC-01567; No of Pages 13

Contents lists available at ScienceDirect

Journal of Development Economics

j ourna l homepage: www.e lsev ie r.com/ locate /devec

Land tenure differences and investment in land improvement measures: Theoreticaland empirical analyses

Awudu Abdulai a,⁎, Victor Owusu b, Renan Goetz c

a University of Kiel, Kiel, Germanyb Kwame Nkrumah University of Science and Technology, Kumasi, Ghanac University of Girona, Girona, Spain

⁎ Corresponding author. Department of Food EconomUniversity of Kiel, Olshausenstrasse 40, 24118 Kiel, Germfax: +49 431 880 7308.

E-mail address: aabdula@food-econ.uni-kiel.de (A. A

0304-3878/$ – see front matter © 2010 Elsevier B.V. Aldoi:10.1016/j.jdeveco.2010.08.002

Please cite this article as: Abdulai, A., etempirical analyses, J. Dev. Econ. (2010), do

a b s t r a c t

a r t i c l e i n f o

Article history:Received 16 July 2008Received in revised form 26 July 2010Accepted 1 August 2010Available online xxxx

JEL classification:O13O55Q12O1

Keywords:Land tenureProperty rightsInvestmentOptimal controlFarm productivity

This article develops a theoretical framework to examine the relationship between land tenure arrangementsand households' investment in soil-improving and conservation measures. It then analyzes this relationshipwith a multivariate probit model based on detailed plot-level data from villages in the Brong Ahafo region ofGhana. A major hypothesis tested is that investment in productivity-enhancing and conservation techniquesare influenced by land tenure arrangements. The theoretical analysis and empirical results generally revealthat land tenure differences significantly influence farmers' decisions to invest in land-improving andconservation measures. The findings also show that tenure security does affect farm productivity.

ics and Consumption Studies,any. Tel.: +49 431 880 4426;

bdulai).

l rights reserved.

al., Land tenure differences and investmenti:10.1016/j.jdeveco.2010.08.002

© 2010 Elsevier B.V. All rights reserved.

1. Introduction

The role of land tenure on investment in productivity-enhancingmeasures in developing countries has been widely documented in theeconomic literature. Since land is central to the social and economicdevelopment of a vast majority of the people living in Sub-SaharanAfrica, the link between indigenous tenure arrangements andproductivity-enhancing investments has attracted the attention ofboth researchers and policy makers. While studies by Dorner (1972)and Harrison (1987) argued that indigenous tenure systems provideinsufficient security to induce farmers to undertake soil-improvinginvestments, Noronha (1985) pointed out that these arrangementsare dynamic and evolve in line with factor prices. The significance ofthis debate has attracted a great deal of attention among economists(Binswanger and Rosenzweig 1986; Besley, 1995; Quisumbing et al.,2001a; Brasselle et al., 2002; Place and Otsuka, 2002; Jacoby et al.,2002; Banerjee and Ghatak, 2004; Goldstein and Udry, 2008;Deininger and Ali, 2008).

A central issue of the related empirical investigations is the effect oftenure security on investment and productivity. On theoreticalgrounds, three main arguments have been advanced for a positivelink between tenure security and investment. First, secured propertyrights are expected to provide a guarantee for farmers to undertakelong-term investment in land-improving and conservation measures,since there would be no fear of expropriation. As noted by Banerjeeand Ghatak (2004), given that the result of land-improving invest-ments is normally realizedwith a one period lag, if the tenant is evictedwith some possibility during this period, he will be enjoying only afraction of the benefit from the investment in expected terms. Thismay cause the tenant to supply a lower level of investment in effort forthe same crop share, a reason why security of tenure is thought to begood for investment. Second, it has been argued that secured landrights make it easier to use land as collateral to obtain loans to financeagricultural investments (Feder and Feeny, 1991). The third effectoperates through better possibilities for trade. If improved transferrights enhance factor mobility bymaking it easier for farmers to sell orrent their land, investment in land-improving measures may befacilitated. An issue that has gained increasing significance in recentempirical analysis is the endogeneity of land rights in estimating theeffect of tenure security on agricultural investment. Authors likeBesley (1995), Quisumbing et al. (2001a), Place and Otsuka (2002),

in land improvement measures: Theoretical and

2 A. Abdulai et al. / Journal of Development Economics xxx (2010) xxx–xxx

Brasselle et al. (2002) have rightly noted that farmers may undertakeland-improving investments in order to gain tenure security.1

As rightly noted by Besley and Ghatak (2009), the empiricalfindings from the land rights–investment relationship appear to beinconclusive. Studies on Africa by Migot-Adholla et al. (1994) andPinkney and Kimuyu (1994) reveal that the impact of land rights onsoil-improving investment and planting of tree crops is quite low.On the other hand, studies by Besley (1995) on Ghana, Jacoby et al.(2002) on China (2002), as well as Carter and Olinto (2003) onParaguay show that tenure security exerts a positive and significantimpact on investments. Banerjee et al. (2002), who found a positiveimpact of tenure reform on farm productivity in India, partlyattributed this to higher investment due to improved tenuresecurity. Brasselle et al. (2002) report that land tenure security isinfluenced by investment, and that once the endogeneity bias isproperly controlled, increased land rights do not appear to stimulateinvestment. Place and Otsuka (2002) also found that coffee plantingis used by farmers to enhance tenure security, supporting the notionthat farmers consider tenure implications when making investmentdecisions.

The more recent studies appear to be showing positive impacts oftenure security on investment. Deininger and Ali (2008) show thatfull land ownership, compared to mere occupancy rights, exerts astatistically significant and economically large effect on investmentand productivity of land-use in Uganda. Goldstein and Udry (2008)also find that insecure land tenure in Ghana is associated with greatlyreduced investment in land fertility, while Jacoby andMansuri (2008)find in their study on Pakistan that farmers invest less in their leasedplots than they do in their owned plots.

This article contributes to the tenure-security-investment debateby developing a framework that captures the impact of different landtenure arrangements on investment decisions of farmers. The modelembodies behavioral assumptions consistent with investment deci-sions that characterize investment in productivity-enhancing inputsin the agricultural sectors of most sub-Saharan African countries. First,we develop a model to examine the effects of 4 different land tenurearrangements on investment decisions of farmers on theoreticalgrounds. The investments include planting trees, mulching, andapplication of organic manure and mineral fertilizers. We then usevariations in tenure arrangements between 560 plots obtained from asurvey of 246 farmers from 6 villages in the Brong Ahafo region ofGhana to analyze the impact of land tenure arrangements oninvestment in soil-improving and conservation measures. Theempirical part of the article also examines the relationship betweenland tenure arrangements and farm productivity.

The main contributions of the article reside in the fact that theresults from the theoretical analysis hold for a wide range ofsituations and are as such independent of case specific data. Theempirical analysis considers a) endogeneity between land rightsand investment decisions and b) interdependence between thedifferent investment decisions. Our empirical evidence shows thatland tenure differences significantly influence farmers' decisions toinvest in land-improving and conservation measures, and thattenure differences do affect farm productivity, even after accountingfor household fixed effects.

The rest of the paper is organized as follows. Section 2 discussesland tenure in Ghana. In Section 3, we present a theoretical model onsoil capital and forest use on plots where farmers can undertake short-term and long-term investments in land improvements. In Section 4,the estimation and identification strategy employed in the empiricalanalysis is outlined. Section 5 discusses the survey data. The empirical

1 Quisumbing et al. (2001a) document in their study that when primary forestswere abundant in the Akan area in Ghana, they were usually appropriated by young,unmarried males for food-crop production. In return, they obtained relatively strongland rights for the substantial labor input that was required to clear forests.

Please cite this article as: Abdulai, A., et al., Land tenure differencesempirical analyses, J. Dev. Econ. (2010), doi:10.1016/j.jdeveco.2010.08.

results are discussed in Section 6, while the final section presentssome concluding remarks.

2. Land tenure in Ghana

As in several other African countries, land is traditionally owned bythe community in Ghana. Control of the land is transmitted throughthe elders, who are custodians of the land. Each headman thereforesees to it that all members of his lineage have portions to farm (Gildea,1964). With the development of the cocoa industry in the country,practices of landholding have become individual ownership in contrastto family control of segments of the community land. Particularly inthe Ashanti and Brong Ahafo regions, the procurement of large bundlesof land by wealthy investors changed the old order. These investorseither moved into previously unclaimed land or acquired securedrights to community land in exchange for money or influence. Some ofthese large-scale farmers sometimes reside somewhere else andsupervise the operations on their land (Benneh, 1989).

Even among migrants, land rights have become more clearlyindividualized, with members of the family qualifying for inheritanceof land in the event of the death of the family head (Quisumbing et al.,2001b). There is a complex system of communally owned land in therural northern regions of the country, withmany local variations. Landtenure is generally based on the community's social organization, andthe basic unit of ownership is the family or clan. Institutions of landtenure are now evolving towards individualized ownership throughinvestments in tree planting and management, transfer of land asgrants and gifts, and the coming into effect of the Interstate SuccessionLaw (PNDC 111) in 1985 (Otsuka et al., 2003). The law is a legalframework that provides equal rights of inheritance between spousesand increased rights for children.2

Given that full ownership of rights over land traditionally resideswith the community, one becomes less concerned with overall landtenure security than with rights that the individual holds over specificland parcels (Place and Hazell, 1993). We therefore focus on the long-term interests farmers have on land parcels, in terms of their rights tocultivate the land for long periods of time and their ability to rent orsell the land. As argued by Place and Hazell (1993), these features ofland control are best captured by tenure measures based on theindividual use and transfer rights that farmers possess over land. Tocapture these features, we collected detailed information on individ-ual rights – basically use rights and transfer rights – for each parceloperated by the farmers in the sample.3

The four main types of land tenure arrangements identified inthe survey area are owner-operated with full property rights,owner-operated with restricted property rights, fixed-rent andsharecropping contracts. The owner-operated with full rightsinvolves farmers owning and cultivating their own plots. Farmerscultivating these parcels have transfer rights, including rights to sellthe parcels, although in some cases family approval has to beacquired before the land can be sold. Owner-operated withrestricted rights involves plots that are acquired as grants, butcannot be transferred or inherited, although they may be rented out.The fixed-rent arrangement involves land owners renting outparcels to tenants, who are normally migrants from other areas.Under a sharecropping contract, an arrangement is made betweenthe landlord and the operator, such that part of the output is givento the landlord as compensation for using the land. The two forms

2 Otsuka et al. (2003) have observed that instead of the stipulations of the law, somelocal communities prefer a formula that gives one-third of the property to each spouse,children and maternal family.

3 As rightly noted by an anonymous reviewer, the more commonly used term for apiece of land with a unique mode of land acquisition is parcel. Given that some authorsrefer to it as plots, we use the terms synonymously in this paper.

and investment in land improvement measures: Theoretical and002

3A. Abdulai et al. / Journal of Development Economics xxx (2010) xxx–xxx

are abunu (one-half of the output for the tenant and one-half for theowner) and abusa (one-third for the tenant).4

3. Theoretical analysis

The model that will be presented later analyzes the link betweenland tenure arrangements and investment in land-improving andconservation measures within a dynamic framework. The previousliterature, considers standing forest (Angelsen, 1999; Barbier, 2004)or soil capital (Ehui et al., 1990) as a renewable or non-renewableresource. In this article, we model soil capital and trees as arenewable resource and analyze their interdependencies withagricultural production. It is assumed here that farmers combineinvestments in both mineral fertilizers, XM(t), such as NPK andorganic fertilizers, XOr(t), such as mulch and organic manure wheret indicates calendar time.5 We control for cultivated plots and plotswhose areas of production have been used for tree planting.Farmers are also assumed to choose production methods thatimprove soil fertility and increase productivity. Although crop yieldsnormally increase with higher rates of mineral fertilizer, yields maydecline with time, if no organic fertilizer or any other organicmaterial is added. The decline in yields may result from soildegradation, which then erodes the original purpose of investment.

Given the potential physical and chemical degradation of the soilas a result of continuous application of mineral fertilizer, profitmaximizing farmers normally invest in organic fertilizers that buildup the soil structure and naturally replenish nutrients in the soil withrelatively low cost. Moreover, the nutrients supplied by organicfertilizer are available over a longer time horizon compared tonutrients supplied by mineral fertilizer (Besley, 1995; Jacoby et al.,2002).6 Let us assume now that the production function is defined for1 ha. Under this assumption, the agricultural production function perhectare can be defined as f(S(t),XM(t),XOr(t)), where S(t) representssoil capital, and XM(t) and XOr(t) are as defined previously. Theapplication of organic fertilizers augments soil capital according to thefunction h(XOr(t)), with h′(⋅)N0. Moreover, since the mineral andorganic fertilizers are “perfect” substitutes in the short run, we canwrite the function f(⋅)as a sum of expressions that reflect individuallythe effect of XM(t), and XOr(t). Hence, we can assume that the crossderivative, fXM

XOr is equal to zero.In contrast, soil capital and fertilizers are subsitutes to a lower

degree. As explained previously, the continuous application of XM(t), orXOr(t) has opposite effects on the evolution of S(t) over time. Anincrease in S(t) results in a decrease in the marginal productivities ofXM(t) and XOr(t), while a decrease in S(t) yields the opposite effect.Therefore, we assume that fXMSb0 and fXOrSb0. The volume of thebiomass of trees (wood) is given by W(t). The farmers have the choiceto plant young trees and to cut trees, whose volume is denoted by P(t)and C(t) respectively. The planted trees grow according to the logisticgrowth function g(W(t)), with g′(⋅)N0. Standing trees increase soilcapital specified by the function y(W(t)), with y′(⋅)N0. This is becausetree management practices can improve the availability of nutrients in

4 As noted by Jacoby and Mansuri (2008), landlords normally offer fixed-rentcontracts only to tenants with sufficiently high wealth. For tenants with bindingwealth constraints, landlords normally go into share contracts. However, it issignificant to note that share-cropping contracts usually give rise to the well-knownMarshallian inefficiency, in which both current production effect and investment areprovided below their first best level (Banerjee and Ghatak, 2004).

5 In order to simplify notation of the theoretical model we denote mulch andorganic manure by the single variable XOr because both affect soil capital in a verysimilar way. However, in the empirical part of the paper we distinguish between thesetwo organic fertilizers because farmers usually apply only one of them.

6 In a recent paper, Jacoby and Mansuri (2008) indicate that field trials in Pakistanshow that the marginal effects of manure on grain yields persist for at least three yearsfollowing the initial application, while the productivity effects of mineral fertilizers areessentially limited to the season of application.

Please cite this article as: Abdulai, A., et al., Land tenure differencesempirical analyses, J. Dev. Econ. (2010), doi:10.1016/j.jdeveco.2010.08.

the soil, and also protect agricultural crops against water and winderosion. In addition, standing trees may also provide fruits. The fruityield is assumed to be proportional to the biomass of the trees, and isspecified as ϑW(t). However, planting trees reduces the acreage that isavailable for crop production. This reduction in acreage can beexpressed asαW(t) with αN0. The size of the entire farm is normalizedto one and the share of the land that is used for crop production isdenoted by L(t). Under the restriction that 0≤L(t)=1−αW(t),W=0implies that the entire land will be used for crop cultivation, whereasW=1/α implies using the entire land for growing trees.

Since current decisions tend to affect the evolution of the naturalresources over time, we analyze the farmer's decision problemwithina dynamic context and take into consideration the fact that theplanning horizon of the farmer depends on the land rights. We furtherassume that agricultural households maximize farm net benefitssubject to agronomic and biophysical constraints (the evolution of thesoil and forest) over a planning horizon of length tk− t0 where t0denotes the initial point and tkthe final point of the planning horizon.The residual value of the trees and soil capital is given by r(S(tk),W(tk)). The function r(⋅)will be zero for fixed-rent tenants andsharecroppers, because they do not have the possibility to sell theland. Owner-operated with restricted property rights have thepossibility to rent out the land, but not sell or bequeth it, so that thevalue of the function will be strictly positive. However, it is smallerthan that of the owners with full property rights because they can sellor bequeth the land. Given that restricted property rights are usuallynot limited over time, it is assumed that owners with full or withrestricted rights have the same long-term perspective,7 while tenants(fixed-rent or sharecropping) have a planning horizon that corre-sponds to the stipulated tenure duration. Given these assumptions,the farmer's decision problem can be stated as

maxL;XM ;XOr ;P;C

J ≡∫tk

t0

e−φt ½ðpf S tð Þ;XM tð Þ;XOr tð Þð Þ−pMXM tð Þ

−pOrXOr tð ÞÞL tð ÞÞ−pL ⋅ð Þ + pWC tð Þ + pFϑ W tð Þ

−pPP tð Þ�dt + e−φtk r S tkð Þ;W tkð Þð Þ

ð1Þ

subject to

S tð Þ = h XOr tð Þð ÞL tð Þ + y W tð Þð Þ−δf S tð Þ;XM tð Þ;XOr tð Þð ÞL tð Þ;

with S t0ð Þ = St0 ;W tð Þ = g W tð Þð Þ−C tð Þ + P tð Þ;

with W t0ð Þ = Wt0;0≤L tð Þ = 1−α W tð Þ and XOr tð Þ;XM tð Þ;C tð Þ≥0;

where pL ⋅ð Þ = 1−θð Þ pL + θγ1pf ⋅ð ÞL tð Þ + θγ2pWC tð Þ represents thecost of the land cultivated with agricultural crops and planted trees,withθ=0,1. In the case of sharecropping θ=1 and pL=γ1pf(⋅)L(t)+γ2pWC(t)), where γ1 and γ2 indicate the share of the yields that accrueto the owner of the land.8 In the case of no sharecropping (owner andfixed-rent tenant), θ=0, the cost of the land is given by the constantpL, which denotes the annual land rent in the case of a tenant, or theannual cost of capital (own and borrowed) in the case of an owner.

All parameters, except t0 , tk and St0 are grouped in a vector named υ.The components of this vector are given by p = price of the cultivatedcrop, pM = price of mineral fertilizer, pOr = price of organic fertilizer,pW = price of the wood minus the logging and transportation cost, pFthe price of the fruit, pP = price of the seedlings of the trees and its

7 Since the time perspective of owner-operated is independent of type of propertyright, we simply use the term owner in the theoretical section of the article to refer toboth types of owners.

8 Inherent to sharecropping is the question of sharing the risk between the landlordand the farmer. However, as we focus on the issue of different tenure regimes we useexpected values and do not analyze the variation in crop yields.

and investment in land improvement measures: Theoretical and002

10 As pointed out by a reviewer, the current formulation allows the landlord toextract the increase in soil capital as a result of a “hold-up problem”, i.e., the landlordas the owner of the asset can withhold the benefits of an increase in the asset quality.Alternatively, one can think of different contract arrangements that allow the tenant to

4 A. Abdulai et al. / Journal of Development Economics xxx (2010) xxx–xxx

plantation, δ = degradation of the soil capital, and φ = discount rate.The previously introduced parameters ϑ,γ1,γ2 and pL also belong tothe vector υ.

Let the solution of problem (1) be given by

J L� tð Þ;X�Or tð Þ;X�

M tð Þ; P� tð Þ;C� tð Þ; t0; tk; S t0ð Þ;�υ� �= J� t0; tk; S t0ð Þð Þ;

where the superscript * indicates the evaluation of the variable alongits optimal trajectory given the parameter values of t0, tk,S(t0) and υ.Hence, J*(t0, tk,S(t0)) indicates the maximized discounted farm netbenefits aggregated over the time horizon of length tk− t0 given theinitial soil capital of S(t0)=St0.

As indicated in the Introduction, observations by some authorsshow that farmers may undertake land-improving measures such asplanting of trees in order to gain land rights, resulting in endogeneityof land rights in an investment specification.We analyze this potentialendogeneity effect by examining the extent to which soil-improvinginvestment choices, L(t),XOr(t),XM(t),P(t),C(t), may affect land tenurearrangements.9 For this purpose, we consider two cases: a) invest-ment in land (land acquisition) and b) investment in soil capital (soilimprovement): Since the decision model (1) identifies farm returns,the costs of fertilizer, young trees, and land, the farm net benefitsalready take into account the costs and benefits associated withdifferent land tenure arrangements. Thus, it is possible to compare thediscounted stream of future farm net benefits of the different landrights. Define nN0 as the number of potential contracts and k as theduration of each contract (in years). Hence, the discounted farm netbenefits of a tenant (Te) or share cropper (Sh) aggregated over a time

span of n k years, are given by: J�Te;Sh = ∑n−1

i=0J� tik; t i + 1ð Þk; S tikð Þ� �

. Thus,

the discounted farm net benefits are maximized for each contract oflength k not taking into account the other n−1 contracts. Thedifferent periods are connected through the stock variable since theterminal value of period i becomes the initial value of the soil capital ofperiod i+1. Consequently, a tenant or sharecropper maximizes the n-sequence of farm net benefits of k-years, but does not maximize overthe time horizon of n k. In contrast, an owner (O) maximizes the farmnet benefits over the entire planning horizon n k. Thus, the discountedmaximized farm net benefits over the entire planning horizon aregiven by JO

* = J*(t0, tnk,S(t0)).

Observation 1. The condition JTe, Sh* b JO

* indicates that it is optimal fora tenant or sharecropper to acquire land permanently, whereas JTe, Sh

* N JO*

indicates that it is not optimal.

Observation 1 points out that a tenant or sharecropper whomaximizes farm net benefits over a planning horizon of length n kshould acquire land on a permanent basis, if the costs of investing inland and soil capital are lower than the sum of the additional farm netreturns resulting from the improved soil and the saved payments incash or kind for the land-use. In this case, land improvement and landacquisition decisions occur simultaneously. Finally, land acquisitionmay also be considered as a mean of securing a stream of future farmnet returns in order to avoid a situation where the tenurearrangement is not renewed after m contracts, 0bmbn, in whichcase the analysis would have to consider the expected value of JTe, Sh* ,rather than its certainty value.

In the second case, we consider investment in soil capital (soilimprovement). In particular, we analyze the situation where tenantsor sharecroppers invest in soil capital during the first tenure period oflength k in order to achieve an extended tenure arrangement of length2k instead of k in the subsequent period. We assume that the landlordis willing to offer an extended duration of the tenure contract

9 As we discuss in the Empirical analysis section of the paper, Jacoby and Mansuri(2008) argue that most models of agrarian contracts result in endogeneity of landtenure variables.

Please cite this article as: Abdulai, A., et al., Land tenure differencesempirical analyses, J. Dev. Econ. (2010), doi:10.1016/j.jdeveco.2010.08.

provided that the tenant or sharecropper has conserved or invested inthe soil, so that soil capital exceeds a threshold value Stk at the end ofthe initial lease arrangement, i.e., S(tk)≥Stk. The farm net benefits of atenant or sharecropper with no extended tenure arrangement over aplanning horizon of 3k is given by

J�Te;Sh = ∑2

i=0J� tik; t i + 1ð Þk; S tikð Þ� �

:

while the farm net benefits of an extended tenure arrangement afterthe first period can be expressed as

˜J�Te;Sh = J� t0; tk; S t0ð Þ; S tkð Þ N Stk

� �+ J� tk; t3k; S tkð Þð Þ:

Observation 2. The condition J�Te;Shb

˜J�Te;Sh indicates that it is optimalfor a tenant or sharecropper to conserve or invest in the soil in exchangefor an extended tenure arrangement of length 2k, whereas J

�Te; Sh N ˜J�Te; Sh

indicates that it is optimal to continue with tenure arrangements oflength k.

Observation 2 suggests that a farmer who maximizes farm netbenefits over a planning horizon of length 3k should conserve orinvest in soil capital if the additional farm net benefits from improvedsoil are greater than the associated investment costs. In this case, soilconservation or soil improvement leads to a change in the conditionsof the land tenure arrangement.10

After analyzing the extent to which soil investment choices mayaffect land tenure rights, we now examine the extent to which landrights may affect soil investment choices. For this purpose, weevaluate the first order conditions of the farmer's decision problemgiven in Eq. (1). To simplify notation, we suppress the argument t onthe variables as well as those of the costate variables and Lagrangemultipliers to be introduced later, and define the current valueLagrangian, ℒ, by

ℒ = pf S;XM ;XOrð Þ−pMXM−pOrXOrð ÞL−pL ⋅ð Þ + pWC + pFϑ W

−ppP + μ1L + μ2 1−αW−Lð Þ + ξ1XM + ξ2XOr + ξ3P + ξ4C

+ λS h XOrð ÞL + y Wð Þ−δf S;XM ;XOrð ÞLð Þ + λW g Wð Þ−C + Pð Þ;ð2Þ

where λS and λW are the corresponding costate variables, μ1 and μ2 areLagrange multipliers associated with the restrictions related to theavailability of land, and ξ1 to ξ4 are Lagrange multipliers related to thenon-negativity of the control variables. The first order conditions aregiven by

∂ℒ∂L = pf−pMXM−pOrXOr−θγ1pf + λS h−δfð Þ + μ1−μ2 = 0 ð3Þ

∂ℒ∂XM

= pfXM−pM− θγ1p + λSδð ÞfXM

� �L + ξ1 = 0 ð4Þ

∂ℒ∂XOr

= pfXOr−pOr−θγ1pfXOr

+ λS h′−δfXOr

� �� �L + ξ2 = 0 ð5Þ

internalize part of the increase in the asset quality, for instance, if the landlordcommits to renegotiate the conditions of the contract at a certain point in time, whilethe contract is in place. See the paper by Jacoby and Mansuri (2008) for a discussion ofthe importance of imperfect commitment.

and investment in land improvement measures: Theoretical and002

( )h′− + >

(owner)S

pM

pM +

pM + fXM

fXM

λ δ

(owner)Spor+ porλ fXorδ

( )h′− + < (owner)Spor+ porλ fXorδ

(sharecropper)pγ

(fixed-rent tenant)

pM(fixed-rent tenant)

pfXM ,with S(t) = S0

pf Xor , S(t) = S0

pf Xor , S(t) > S0

pf Xor , S(t) < S0

pfXM ,with S(t) > S0

pfXM ,with S(t) < S0

OMX Te

MX MXTeMXO

MX ShMXSh

MX

OX OXTeXTeXOor or or or or or or

XorX ShXShX

1

pM + fXor (sharecropper)pγ1

A

B

Fig. 1. A. The optimal amount of mineral fertilizer. B. The optimal amount of organicfertilizer.

5A. Abdulai et al. / Journal of Development Economics xxx (2010) xxx–xxx

∂ℒ∂P = −pP + λW + ξ3 = 0 ð6Þ

∂ℒ∂C = 1−θγ2ð Þp′W + ξ4−λW = 0 ð7Þ

λS = φλS− p−θγ1p−λSδð ÞfSLð Þ ð8Þ

λW = φλW−pFϑ + αμ2−λSy′−λWg′ Wð Þ: ð9Þ

If we assume that the initial soil stock S(0)=S0 is identical for allthree tenure regimes, we can determine the optimal short-runbehavior for all three types of tenure arrangements. However, thesoil stock will not be identical for all three tenure regimes over timeand we also need to account for individual changes in S(t). Thesechanges are significant in determining the optimal long-run behaviorof the different types of farmers. The short-run and long-runbehaviors of farmers will be discussed later.

3.1. Short-run behavior

For an interior solution, (ξ1=0), and pfXM(⋅)|S(t)= S0 the solution of

Eq. (4) is presented in Fig. 1A.11 Whereas owner-cultivated willconsider the shadow cost of the soil (λS), tenants or sharecroppers donot consider these costs. Hence, fixed-rent tenants, sharecroppers andowners apply XM

Te, XMSh, and XM

O levels, respectively.

Observation 3. Given that the initial soil capital is identical for allthree types of farmers, Fig. 1A shows that tenants initially apply moremineral fertilizer than sharecroppers and owner-cultivators. A directcomparison between sharecroppers and owner-cultivators is not possiblein the present analysis.

The behavior of sharecroppers will depend on the value of γ1. Incase the owner chooses γ1=λSδ /p, owners and sharecroppers wouldtend to apply the similar levels of mineral fertilizer. The optimal levelsof organic fertilizer, which can be derived from an interior solution ofEq. (5) is presented in Fig. 1B.

Observation 4. Given that the initial soil capital is identical for allthree types of farmers, Fig. 1B shows that owners apply initially moreorganic fertilizer than tenants, if the soil improvement effect of organicfertilizer outweighs its degradation effect. Sharecroppers apply lessorganic fertilizer than tenants.

Given that owners consider the shadow cost of the soil (λS), theyapply more organic fertilizer than a tenant, provided that the soilimprovement effect of organic fertilizer, h′, is greater than the soildegradation effect (δfXOr

) of cultivation. This situation is depicted inFig. 1B by comparing XOr

Te with XO

Or . It is however significant to notethat tenants may apply more organic fertilizers than owners underspecific conditions. Such a situationmay arise, if the soil improvementeffect of organic fertilizer is lower than the soil degradation effect,which compares XOr

Te with XOrO . Fig. 1B also indicates that sharecroppers

apply less organic fertilizer, XOrSh, than fixed-rent tenants. A direct

comparison between sharecroppers and owners is however notpossible. If the share that accrues to the landlord is equal toλSp

−h′fxOr

+ δ� �

, owners and sharecroppers would apply similar levelsof organic fertilizer.

11 Since the cross derivative of fXMXOr is zero, we can graph pfXM

independently of thetenure arrangement, although XM and XOr vary with the tenure arrangement. However,pfXM

varies with S, since fXMS and fXOrS decrease with an increases in S.

Please cite this article as: Abdulai, A., et al., Land tenure differencesempirical analyses, J. Dev. Econ. (2010), doi:10.1016/j.jdeveco.2010.08.

3.2. Long-run behavior

For the determination of the long-run behavior of the farmers, weneed to establish the optimal evolution of soil capital (stock variable).For this purpose, we distinguish between the situations where theinitial soil capital is above the long-run soil capital and where it isbelow. Although fixed-rent tenants and sharecroppers usually haveshort-term contracts, we include their behavior in the analysis byinterpreting the sequence of their short-term behavior as their long-run behavior. The principal results can be summarized in thefollowing observation.

Observation 5. Provided that owners build up soil capital, Fig. 1Aand B shows that owners will tend to reduce the application rate ofmineral and organic fertilizer over time, whereas sharecroppers andfixed-rent tenants increase the application of mineral and organicfertilizer over time, provided that their soil capital decline.

The demonstration of Observation 5 and additional analysis arepresented in Appendix A. Condition (6) suggests that it is optimal toplant young trees if their in-situ value, λW, is equal to their plantingcost. Otherwise, ξ3 presents the difference between planting costand in-situ value of the young trees, and it is optimal to plant notrees. Cutting of trees results in CN0, and therefore ξ4=0 in Eq. (7).Hence, for tenants and owners, the unit price of wood needs to beequal to the in-situ value of the standing trees, while an additionalγ2pW has to be subtracted from the net price of wood forsharecroppers. If the net price minus θγ2PW is not equal to the in-situ value, the difference will be reflected in the value of ξ4. It is alsoevident in Eq. (7) that λW is equal to pW(1−θγ2) for cases whereCN0, indicating that λW is a constant, and hence λW = 0. Thiscondition holds at the steady state equilibrium by definition.Moreover, the condition λW = 0 holds outside the steady stateequilibrium when farmers cut trees. Hence, the following discussion

and investment in land improvement measures: Theoretical and002

6 A. Abdulai et al. / Journal of Development Economics xxx (2010) xxx–xxx

holds for the two described situations. In this case, utilizing thedefinition of λW in Eq. (7), (9) can be written as:

0 = φ−g′ Wð Þ� �

1−θγ2ð ÞpW−pFϑ + αμ2−λSy′ Wð Þ: ð10Þ

where land is in abundance, the opportunity cost of land tends to bezero, i.e., μ2=0. Given that both tenants and sharecroppers do notconsider the shadow cost of the soil, Eq. (10) reduces to

0 = φ−g′ Wð Þ� �

1−θγ2ð ÞpW−pFϑ: ð11Þ

Let us first consider the case where the trees only produce woodwithout fruits, i.e. ϑ=0. In this situation, Eq. (11) holds, if wechoose W such that g′(W) is equal to φ, in which case the marginalgrowth rate of the biomass will be equal to the discount rate. Thiscase is depicted in Fig. 2 for W=W*, a result that is standard innatural resource economics for common property resources, sincethe optimal stock is on the left side of the stock that supports themaximum sustainable yield, WMSY. In the situation where treesproduce wood and fruit, ϑN0, Eq. (11) can only hold if the term(φ−g′(W) is strictly positive. In other words, the term g′(W) has todecrease in comparison with the situation where trees do notproduce any fruit. Consequently, as it can be seen from Fig. 2, theoptimal amount of W has to increase and will be situated to the leftof W*. Hence, if trees produce fruit, farmers find it optimal toincrease the “number of trees”.

However, if agricultural land is scarce, the opportunity cost of landis given by μ2=pf−pMXM−pOrXOr−θγ1pf+λS(h−δf), according toEq. (3). Hence, the optimal W for owners is given by

0 = φ−g′ Wð Þ− pFϑpW

+αμ2pw

−λSy′

pW

!pW

= φ−g′ Wð Þ + αpW

−pFϑα

+ pf−pMXM−pOrXOr + λS h−δf− y′

α

! ! !pW

ð12Þ

and for sharecroppers and tenants by

0 = φ−g′ Wð Þ + αμ2−pFϑ1−θγ2

� �2pW

!1−θγ2ð ÞpW

= φ−g′ Wð Þ + α1−θγ2

� �2pW

− pFϑα

+ pf−pMXM−pOrXOr−θγ1pf� � !

1−θγ2ð ÞpW :

ð13Þ

It needs to be noted that the optimal W cannot be unambiguouslydetermined from Eqs. (12) and (13), since the term μ2 is not identicalfor the different tenure regimes. Hence, the graphical solution ofEqs. (12) and (13) can only be obtained under the assumption that the

Fig. 2. The optimal “number of trees” in the presence of agricultural production.

Please cite this article as: Abdulai, A., et al., Land tenure differencesempirical analyses, J. Dev. Econ. (2010), doi:10.1016/j.jdeveco.2010.08.

differences in μ2 are relatively small and do not alter the ranking of theoptimal W for the different tenure regimes. The empirical analysisundertaken with primary data addresses the situation where theprevious assumption is not applicable.

Observation 6. If the opportunity costs of land are initially zero(land abundance) and increase thereafter, for instance as a result of anincrease in population pressure; owners and tenants grow fewer trees ifthey are used exclusively for wood production. However, if theopportunity costs of land are strictly positive from the beginning, theinfluence of land tenure rights on tree planting can only be derivedempirically.

The demonstration of Observation 6 is presented in Appendix A. Theforegoing analysis has considered the case where trees are cut, that is,CN0. In the specific case where C=0 we know from the relationshipW = g′ Wð Þ−C + P, that W is defined by dW

dt = g′ Wð Þ = ∫dW =∫g′ Wð Þdt, in which caseW tð Þ = WO + ∫

t

0g′ Wð Þdt. Trees grow during

this phase without being cut.

4. Empirical analysis

In this section, we employ plot-level data to examine therelationship between land tenure arrangements and investments inproductivity-enhancing measures, to complement the theoreticalanalysis presented in the previous section.12 We also investigate howland tenure arrangements affect farm productivity.

4.1. Econometric specification

The first order conditions (4)–(9) imply that farmers invest inland-improving or conservation measures if it leads to an increase inthe expected farm net benefit aggregated over the planning horizon.However, the expected farm net benefit is not observable, since it issubjective. What is observed is the decision to invest or not to invest,i.e. the planting of trees, application of mineral fertilizer, as well asorganic fertilizer such as mulch and organic manure. The empiricalanalysis focuses on the factors that influence the likelihood of farmersengaging in these investments. In linewith themaximization problemoutlined in Eq. (1), farmers invest in soil-improving and naturalresource management measures, if it augments the farm net benefit,i.e. ∂ J /∂P,∂ J /∂XM,∂ J /∂XOrN0.

Unfortunately, changes in J are not observable, but can beexpressed as a function of observable elements. Let us define theunderlying latent propensity variable for investment on plot j, ownedby farmer l, for the soil-improving and natural resource managementstrategy m as Jjlm. The underlying propensities can then be related tothe plot's observed characteristics and farmer related variables, Zjlm, aswell as land tenure arrangements, Rjlm and unobserved characteristics,εjlm, in the following latent variable model:

Jjlm = Zjlmβm + Rjlmγm + εjlm m = Trees; Fertilizer;Mulch;Manureð Þ:ð14Þ

Denoting tree planting, mineral fertilizer application, mulchingand manure application as P, XM, XOr

Mu, and XOrMa, respectively, Eq. (14)

12 As noted by a reviewer, the theoretical analysis provides general results in terms ofrelative values, of course, at the cost of some simplification of the complex real world.The empirical analysis complements the theoretical analysis with plot-level data offarm households. However, the results presented are in terms of probabilities due tothe fact the changes in the farm net benefits are not observable. Hence, the empiricalstudy should be considered as complementary analysis rather than an exactspecification of the theoretical model.

and investment in land improvement measures: Theoretical and002

16 Rivers and Vuong (1988) point out that the usual probit standard errors and teststatistics are not strictly valid if the null hypothesis of exogeneity of the variable isrejected. In such a case, they suggest the use of an M-estimator to derive the

7A. Abdulai et al. / Journal of Development Economics xxx (2010) xxx–xxx

can be transformed into a binary probit equation for participation foreach investment option under the following mapping from the latentvariable to its observed realization:

ˆJjlm =

1 if Jjlm N 0;

0 if Jjlm≤0:m = P; XM ; XMu

Or ; XMaOr

� �:

8>><>>: ð15Þ

Let us assume that εjlm(m=P,XM,XOrMu,XOr

Ma) jointly follow amultivariate normal distribution with mean zero and variance 1,and the covariance matrix Σ.13 This can be expressed as

εP ; εXM; εXMu

0; εXMa

0

� �′eMVN 0;Σð Þ:

Maximum likelihood method can then be employed to estimatethe parameters and the four correlations of the error terms (Greene,2008). However, because the probabilities that enter the likelihoodare functions of high dimensional multivariate normal distributions,they are simulated using GHK algorithm (Greene, 2008, p. 582). Somestudies on investment in soil-improving and natural resourcemanagement measures have employed single-equation techniques,with the assumption that εjlm independently follows univariatedistributions.14 However, because of the substitutability or comple-mentarity between these investment options, and the fact that theplots in the sample are similar across equations, it is most likely thatthe error terms of these equations will be correlated.

As indicated earlier, land tenure rights may be influenced byinvestment decisions, resulting in endogeneity of the land tenurearrangement variables in the investment specification. Jacoby andMansuri (2008) also point out that most models of agrarian contractsimply a correlation between contractual choice and unobservedcultivator characteristics, resulting in endogeneity of land tenurearrangements. A properly specified two-stage instrumental variableapproach will produce consistent estimates of the parameters inEq. (14). The first-stage equation specifies land rights as a function ofexogenous variables, including those in Eq. (14) and others that affectland tenure arrangements,

Rjlm = α0 + Zjlmα1 + Vjlmα2 + ξjlm; ð16Þ

where Vjlm is a vector of instrumental variables that is correlated withland tenure arrangements but uncorrelated with εjlm, the residual inEq. (14), and is therefore excluded from Eq. (14). The predicted valuesfrom Eq. (16) are then used in the second stage estimation of Eq. (14).However, when the dependent variable is discrete as in the presentstudy, the usual two-stage approach described previously will not beable to address the endogeneity problem.15 Wooldridge (2002)argues that the most useful two-step approach to examine endo-geneity in a probit model is the Two-Stage Conditional MaximumLikelihood (2SCML) proposed by Rivers and Vuong (1988).

Rather than using the predicted values from the first-stage linearprobability regression, the 2SCML approach involves specifying theinvestment equation as

Jjlm = β0 + β1Zjlm + β2Rjlm + β3Ujlm + μjlm m = P; XM ; XMuOr ; XMa

Or

� �;

ð17Þ

13 As pointed out by Greene (2008), the magnitude of the variance of the disturbanceterm cannot be identified for each probit equation, as such the variance has normallybeen assumed as 1.14 For example, Marenya and Barrett (2007) employed single probit models for theinvestment options in their study on Western Kenya.15 The non-linearity of the probit model will result in estimates of standard errorsthat are downward-biased and coefficients that are not normally distributed(Wooldridge, 2002).

Please cite this article as: Abdulai, A., et al., Land tenure differencesempirical analyses, J. Dev. Econ. (2010), doi:10.1016/j.jdeveco.2010.08.

where Rjlm is a vector of the observed tenure arrangement variables andUjlm is a vector of the residual terms from Eq. (16). The probit estimatesof β2 in Eq. (17) are consistent (Blundell and Smith, 1989; Wooldridge,2002).16 A significant feature of the approach is that the usual probit t-statistics on β3 are valid tests of the null hypotheses that the variablesare exogenous, i.e., H0:β3=0. However, if β3≠0, then the probitstandard errors and test statistics are not strictly valid, and we wouldhave only estimated β1 and β2 up to scale (Wooldridge, 2002). Aspointed out by Wooldridge (2002), under H0:β3=0, ξjlm=μjlm, andhence the distribution of ξjlm plays no role under the null hypothesis.Therefore the test of exogeneity is valid without assuming normality orhomoskedasticity of εjlm. However, when Rjlm and εjlm are correlated,normality of Ujlm is crucial. As noted by Brasselle et al. (2002), a jointWald test can also be performed on the vector of β3's to examine theexogeneity of tenure arrangements as a whole.

While the specification in Eq. (17) controls for farmer and plotcharacteristics, it does not control for family-level unobservables thatcould be correlated with both tenure status and the decision to adoptland-improving measures.17 To deal with this concern, we allowed forhousehold fixed effects in Eq. (17) to yield the following specification

Jjlm = λlm + β1Yjlm + β2Rjlm + β3Ujlm + ζjlm ð18Þ

where λlm is the household-specific intercept and represents theintrinsic propensity (based on variables unobserved by the researcher)of farmer l for activity m; Y is vector of plot-level variables. Thus, inestimating the household fixed-effects model, all household level anddistrict level variables drop out of the regression. A linear probabilitymodel is employed in the first-stage estimation of the four tenure rightsvariables.18 Specifically, three tenure arrangement variables areestimated, with owner-operated without transfer rights being used asthe default variable.

4.2. Identification strategy

To ensure identification in the estimation of the investmentspecification, some of the variables included in the first-stageestimation of tenure rights are excluded from the multivariate probitestimation. As suggested by Jacoby and Mansuri (2008), a suitableidentification strategy is to employ a variable that strongly influencescontractual choice but is orthogonal to unobserved plot character-istics. Distance of the plot from the landlord's home is a suitablevariable, given that plots that are located further away from home aremore costly to self-cultivate. Now to serve as an excluded instrument,a question that needs to be addressed is whether distance to thelandlord is related to plot quality.19 To examine the relationshipbetween proximity of plot from landlord and plot quality, weestimated reduced-form regressions of soil quality, controlling forplot characteristics and district dummies (Jacoby and Mansuri, 2008).Proximity is defined by whether landlord resides in the village theplot is located, and the distance of plot from the landlord' residence.The results, which are not presented for the sake of brevity show thatproximity does not significantly influence soil quality.

asymptotic variance of the two-step estimator.17 We are grateful to an anonymous reviewer for drawing our attention to this pointand suggesting the use of household fixed-effects analysis.18 Brasselle et al. (2002) also employed the 2SMCL in their study on Burkina Faso,using a linear probability model in the first stage, while Besley (1995) employed thelinear probability model to estimate the investment specification in his study.19 As pointed out by Jacoby and Mansuri (2008), landlords may purchase poorquality land far away from their homes with the intention of leasing it out. This is notlikely to be the case in this sample, since land is mainly acquired through inheritance,with only about 1% acquired through purchases.

and investment in land improvement measures: Theoretical and002

8 A. Abdulai et al. / Journal of Development Economics xxx (2010) xxx–xxx

The variables that are excluded from the second stage regression ofinvestment decisions for owner-cultivated plots are mode ofacquisition of the plot, a dummy variable indicating whethercultivator resides in the village where the plot is located or not, andthe distance of the plot from the cultivator's home. The variables usedas instruments for fixed-rent tenants and sharecroppers are a dummyvariable indicating whether the landlord resides in village where theplot is located or not, and the distance of the plot from the landlord'sresidence.20 The validity of the exclusion restrictions are tested withthe approach suggested by Lee (1992), who presents an overidenti-fication test statistics, distributed as χ2 with degrees of freedom equalto the number of excluded instruments.21 The test involves estimatingan alternative version of Eq. (17) that includes the instruments22

Jjlm = β0 + β′1Zjlm + β′

2Rjlm + β′3Ujlm + β′

4I + μ ′jlm m = P; XM ; XMu

Or ; XMaOr Þ:

�ð19Þ

The insignificance of β′4 then provides direct evidence that theinstruments can be excluded from Eq. (17). As indicated in thetheoretical analysis, the magnitude of the influence of different tenurearrangements on investment decisions of farmers, as well as the signof the influence on some investment decisions cannot be determineda priori, and therefore needs to be determined empirically, which willbe undertaken later.

5. Data and definition of variables

The data used in the analysis were collected during January andOctober 2003 in six villages in twodistricts– Techiman andNkoranza–in the Brong Ahafo region of Ghana. A stratified random sample of 246farm households with 560 plots was selected from four villages inTechiman District and two villages in Nkoranza District, with severalhouseholds cultivating multiple plots with different land tenurearrangements. The locations sampled in Techiman include Twimea–Nkwanta, Aworopata, Woraso and Nkwaeso. In Nkoranza District,Dromankese and Ayerede were sampled.

The sample was taken to ensure representation of the various landtenure arrangements in the area. The survey requested eachhousehold to report its land tenure arrangements on each plot itcultivated. Specifically, it consisted of 202 owner-operated withtransfer rights plots, 104 owner-operated without transfer rightsplots, 159 plots under fixed-rent contracts, and 95 plots undersharecropping contracts. A total of 194 households were identifiedwith multiple plots. Farmers were also asked about investments theyhad undertaken in the past five years to improve the land they werecultivating. The investments include tree planting, mulching, manureand mineral fertilizer.23

Land purchases are very rare in this area, where they account forjust 1% of all parcel acquisitions in the sample. On the other hand,inheritance through the family and borrowing are more common inthe area. All land acquisitions for owner-operated without transferrights were through borrowing and gifts, while inheritance and giftswere the main channels of land acquisition for owner-operated withtransfer rights. Gifts were mainly from family members and friends.

20 It is important to note that besides raising the cost of self-cultivation, greaterdistance between landowner and plot may also increase the cost of monitoring theplot, if leased out, resulting in potentially biased estimates (Jacoby and Mansuri, 2009).However, sharecroppers, who are likely to be monitored, are normally not supervisedin the study area. Hence, this problem does not arise in our estimation strategy.21 Davidson and Mackinnon (1993) explain that this statistic tests the jointhypothesis that the excluded instruments are not inappropriately excluded and areuncorrelated with the error term in the investment specification.22 This involves specifying individual probit models for the various investmentactivities and then employing the χ2 test for the validity of the instruments.23 Very few farmers constructed ditches on their plots. Given the insignificantnumber of farmers that were engaged in this investment, we deleted it from theanalysis.

Please cite this article as: Abdulai, A., et al., Land tenure differencesempirical analyses, J. Dev. Econ. (2010), doi:10.1016/j.jdeveco.2010.08.

Information on household characteristics, such as number of years ofschooling and age of farmer, value of livestock owned by household,and value of farm implements were included.

Differences across plots in terms of quality and location also affectthe suitability of the plots for various investments. Information onplot characteristics was therefore collected to address this issue. Theplot-level characteristics gathered include plot size, distance of plotfrom home, gender of farmer cultivating plot, whether the cultivatorof the plot lived in the village where the plot is located, plot fertility,as well as the crops grown on the plot. The descriptive statistics ofthe variables used in the analysis are provided in Table 1. Theincidence of investment is measured by dummy variables that takeon the value of one when a particular investment was undertaken ona given plot and zero otherwise. Four variables are employed in thestudy to examine tenure security. As mentioned previously, theseinclude owner-operated with full rights, owner-operated withrestricted rights, fixed-rent cultivation and sharecropping contract.24

ese variables are measured with dummy variables. Total farm outputvalue, measured in Ghanaian Cedis, includes revenues from bothcrops and trees.

6. Estimation results

6.1. Investment specification

Although our primary interest is on the impact of tenurearrangements on investments, we first estimated the determinantsof land tenure arrangements to account for potential endogeneitybetween the two variables. The results of the first-stage regression arereported in Table A1 in Appendix A. In all three specifications, the p-values of the F-statistics for the instruments used indicate that theyplay an important role in land tenure arrangements. Also reported inthe table are estimates that account for household fixed effects. Theestimates show that the signs of the coefficients are similar to theestimates without household fixed effects, although the magnitudesof the coefficients vary slightly.

The empirical results for the investment specifications arepresented in Table 2. It can be observed from the results that allestimated correlation coefficients are positive and significantlydifferent from zero at the 1% level of significance, indicating thatunobserved variables involved in each investment option aresignificantly positively related, and confirms that it is more efficientto model investment decisions jointly rather than separately. The χ2

statistics for the validity tests of the overidentifying restrictionspresented in Table 2 fail to reject the exclusion restriction that theinstruments affect investment only via land tenure arrangements.

Noteworthy is the fact that all the variables representing theresiduals derived from the first-stage regressions for tenure arrange-ments are not statistically significant at conventional levels, indicatingno simultaneity bias and that the coefficients have been consistentlyestimated (Brasselle et al., 2002; Wooldridge, 2002). Also shown inthe table are the χ2 statistics for the joint Wald tests on the vector ofthese residuals from the first-stage estimations. These values alsoreveal that for each investment equation, the null hypothesis that theresiduals are jointly equal to zero could not be rejected, againconfirming the results of the individual t-statistics.

The variable representing owner-operated with full rights ispositive and significantly different from zero in all four investmentoptions, suggesting that relative to owner-cultivated without transferrights, the owner-cultivated with transfer rights are more likely toinvest in trees, organic and mineral fertilizers, confirming theimportance of secured property rights. Consistent with the theoretical

24 Although the typical fixed-rent tenure duration for the sample was 2 years, someof the fixed-rent contracts went as high as to 4 years, while some of the 2 year contactshad been renewed a number of times.

and investment in land improvement measures: Theoretical and002

25 To examine the effect of tenure duration (number of years current tenant hascultivated the plot) on investments, we re-estimated the specifications in Table 2, andin the spirit of Jacoby and Mansuri (2008), included the log of tenancy duration in thespecifications, with zero for owned plots. The results revealed positive and significantcoefficients for mulch, trees and organic manure, suggesting that farmers with longercontract durations are more likely to invest in soil-improving measures. Thus, thelonger the tenant stays on a leased plot, the higher the probability of his increasedinvestment in measures that improve soil quality over an extended period, comparedto short durations contracts that only encourage investment in measures that increasecrop productivity in a planting season.

Table 1Descriptive statistics of variables used in the regression models.

Variable Definition of variables Mean S.D.

Dependent variablesTREES 1 if farmer plants trees, 0 otherwise 0.43 0.50FERT 1 if farmer applies fertilizer, 0 otherwise 0.42 0.49MULCH 1 if farmer applies mulch, 0 otherwise 0.35 0.48MANURE 1 if farmer applies organic manure, 0 otherwise 0.14 0.34YIELD Total output value per acre (1000 Cedis) 96.17 28.20

Tenure variablesOWNER 1 if land is under own-operated with full rights 0.36 0.48FIXRENT 1 if land is under fixed-rent contract 0.28 0.45SHARECROP 1 if land is under sharecropping contract 0.17 0.38OTHER 1 if land is under owner-operated without rights 0.19 0.26

Household characteristicsAGE Age of farmer (years) 49.98 13.67EDUCN Years of formal education of farmer 3.76 4.88LIVEST Value of livestock wealth (¢×10−6) 11.20 26.11IMPLTS Number of implements owned by farmer 13.47 8.78EXTEN 1 if farmer received extension visit, 0 otherwise 0.38 0.49

Plot characteristicsFEMALE 1 for female held plot, 0 otherwise 0.52 0.21PCREDIT 1 if household has access to credit, 0 otherwise 0.43 0.32PLTDIST Distance of plot from landowner's residence (km) 2.33 1.91PLTLOC 1 if plot outside village of landowner, 0 otherwise 0.19 0.37PLTSIZE Plot size in acres 2.94 2.03PLTYEARS Number of years the plot has been under use 16.58 9.46PLOTFERT 1 if plot is on fertile land, 0 otherwise 0.14 0.35

CropsPLANTAIN If farmer cultivates plantain on plot, 0 otherwise 0.09 0.29CASSAVA If farmer cultivates cassava on plot, 0 otherwise 0.06 0.24VBEANS If farmer cultivates beans on plot, 0 otherwise 0.46 0.49OTHERS If farmer cultivates other crops, 0 otherwise 0.39 0.28

Location dummiesTWIMEA 1 if farmer resides at Twimea–Nkwanta 0.22 0.41AWOROPAT 1 if farmer resides at Aworopata 0.13 0.34WORASO 1 if farmer resides at Woraso 0.18 0.38AYEREDE 1 if farmer resides at Ayerede 0.27 0.44DROMA 1 if farmer resides at Dromankese 0.08 0.28

Note: Exchange rate: US $1=¢8500 in 2003. ¢=Ghanaian Cedis.

Table 2Multivariate probit regressions of investment in land improvement measures.

Variable Trees Mulch Fertilizer Manure

CONSTANT −0.518⁎⁎⁎

(3.26)−0.296⁎

(1.67)−0.408⁎⁎

(2.29)−0.482⁎

(1.78)OWNER 0.718⁎⁎⁎

(4.19)0.663⁎⁎⁎

(2.83)0.229⁎⁎

(2.07)0.335⁎⁎

(2.32)SHARECROP 0.241

(1.28)−0.339⁎⁎

(2.08)−0.094(0.91)

−0.381⁎

(1.96)FIXRENT −0.365⁎⁎⁎

(3.24)−0.181⁎⁎

(2.36)0.277⁎

(1.93)−0.159⁎

(1.88)PLTSIZE 0.034⁎⁎

(2.19)−0.025(0.68)

0.103⁎⁎⁎

(2.66)−0.115⁎⁎

(2.14)PLOTFERT 0.281

(1.54)0.573⁎⁎⁎

(3.08)−0.148(0.74)

0.625⁎⁎⁎

(2.81)PLTYRS 0.082

(1.16)0.196(0.88)

0.106(1.27)

0.073(0.92)

FEMALE −0.481(1.46)

−0.9764⁎⁎

(2.10)0.326(1.04)

−0.311(0.64)

ETHNIC 0.272(1.41)

0.064(1.09)

0.177(1.56)

0.008(0.93)

AGE −0.052⁎⁎⁎

(2.85)0.018(0.86)

0.005(0.56)

−0.027⁎

(1.69)EDUCN 0.092⁎⁎⁎

(4.34)0.056⁎⁎⁎

(2.79)0.026⁎

(1.87)0.014⁎

(1.79)HHSIZE 0.017

(0.64)−0.028(1.20)

−0.025(1.12)

0.021⁎

(1.75)LIVEST −0.006

(1.11)−0.002(0.03)

0.012⁎

(1.82)0.032⁎⁎⁎

(2.73)IMPLTS 0.536⁎⁎⁎

(3.63)0.018(0.71)

0.565⁎

(1.82)0.318(1.56)

RESOWNER 0.172(1.29)

0.014(1.37)

0.056(1.19)

0.096(1.53)

RESFIXED 0.034(1.06)

0.016(1.41)

0.024(1.32)

0.043(1.09)

RESSHARE 0.134(1.08)

0.082(1.22)

0.064(1.41)

0.116(1.43)

Village dummy variables Yes Yes Yes Yesχ2-statistic for jointsignificance of residuals

1.14[0.38]

0.87[0.46]

0.91[0.52]

1.02[0.39]

χ2-statistic foroveridentification

0.73(0.41)

0.44(0.33)

0.62(0.48)

0.67(0.51)

Cross-equation correlationsρTF 0.256⁎⁎ ρTM 0.317⁎⁎⁎ ρTO 0.412⁎⁎⁎

ρFM 0.218⁎⁎⁎ ρMO 0.336⁎⁎⁎ ρFO 0.225⁎⁎⁎

Mc Fadden R2 0.267

Note: Absolute t-values in parentheses. RESOWNER, RESFIXED and RESSHARE denoteresiduals from the first-stage regressions for owner-operated with full rights, fixed-rentand sharecropping contracts, respectively.

⁎ Significant at 10%.⁎⁎ Significant at 5%.⁎⁎⁎ Significant at 1%.

9A. Abdulai et al. / Journal of Development Economics xxx (2010) xxx–xxx

analysis, the variable for fixed-rent is negative and significant fortrees, mulch and organic manure, but positive and significant formineral fertilizer. This indicates that relative to owner-cultivatedwithout transfer rights, plots on fixed-rent contracts are less likely toattract investment in trees, mulch and organic manure, but are morelikely to attract investment in mineral fertilizer for short-termbenefits.

Although the variables for sharecropping are negative, only thosefor mulch and manure are significantly different from zero, alsoindicating that farmers under sharecropping are less likely to invest inthese activities, relative to owner-cultivated. The evidence clearlyreveals that mulch and organic manure are more likely to be used onplots that are owner-cultivated with full rights and owner-cultivatedwithout full rights than leased plots with the same characteristics.Trees are also more likely to be planted on plots that are owner-cultivated with full rights and owner-cultivated without full rightsthan on leased plots. The finding that owner-cultivated plots are morelikely to apply organic fertilizers than both sharecroppers and fixed-rent tenants suggests that the short durations of the contracts thattend to reduce the benefits obtained by tenants from the applicationof manure and mulch may be serving as a disincentive for them toinvest in these measures. Moreover, it may grow stronger the closerthe contract is to its expiration date.

The empirical results also indicate that fixed-rent tenants are morelikely to invest in mineral fertilizer than owner-cultivated without full

Please cite this article as: Abdulai, A., et al., Land tenure differencesempirical analyses, J. Dev. Econ. (2010), doi:10.1016/j.jdeveco.2010.08.

rights. As explained earlier, farmers on fixed-rent contracts normallytry to maximize net benefits from the plots within a very short time,which usually includes applying relatively high levels of mineralfertilizer.25 The variable for sharecropping is negative in three of thespecifications, but statistically significant only for manure. Thefindings here clearly show that owner-operated with full rights aremore likely to invest in these activities than sharecroppers. Thepositive and significant impact of secured tenure rights on investmentis in line with the results reported by Besley (1995) and Goldstein andUdry (2008) for Ghana. In particular, the finding that farmers invest

and investment in land improvement measures: Theoretical and002

Table 3Multivariate probit regression of investment in land improvement measures withhousehold fixed effects.

Trees Mulch Fertilizer Manure

CONSTANT −0.432⁎⁎⁎

(3.17)−0.220⁎

(1.81)0.336(1.47)

−0.568(1.39)

OWNER 0.616⁎⁎

(3.02)0.541⁎⁎

(2.27)0.165⁎

(1.98)0.329⁎⁎

(2.19)SHARECROP 0.173

(1.36)−0.305⁎

(1.97)−0.076(1.63)

−0.268⁎⁎

(2.15)FIXRENT −0.221⁎⁎⁎

(2.68)−0.218⁎

(1.82)0.234⁎

(1.78)−0.184⁎

(1.92)FEMALE −0.217⁎⁎

(2.33)−0.53⁎

(1.81)0.036(0.93)

0.114(1.07)

PLTSIZE 0.161⁎⁎

(2.24)0.075(0.91)

0.118⁎⁎

(2.35)−0.066(1.29)

PLOTFERT 0.043(1.26)

0.528⁎

(1.83)0.012(1.14)

0.137⁎⁎

(2.16)PLTYRS 0.012

(1.02)0.031(0.98)

0.019(1.47)

0.024(1.35)

RESOWNER 0.137(1.43)

0.015(1.19)

0.118(1.28)

0.046(1.06)

RESFIXED 0.019(1.22)

0.027(1.44)

0.014(1.38)

0.052(1.29)

RESSHARE 0.135(1.19)

0.077(1.38)

0.104(1.12)

0.137(1.04)

χ2-statistic for jointSig. of residuals

0.86[0.41]

0.68[0.59]

1.12[0.36]

0.94[0.43]

χ2-statistic forOveridentification

0.76[0.55]

0.57[0.48]

0.65[0.47]

0.84[0.52]

Cross-equation correlationsρTF 0.344⁎⁎

ρFM 0.216⁎⁎⁎

ρTM 0.279⁎⁎⁎

ρTO 0.326⁎⁎⁎

ρFO 0.238⁎⁎⁎

ρMO 0.315⁎⁎⁎

Mc Fadden R2 0.252

Note: Absolute t-values in parentheses and p-values in squared brackets. RESOWNER,RESFIXED and RESSHARE are as defined in Table 2.

⁎ Significant at 10%.⁎⁎ Significant at 5%.⁎⁎⁎ Significant at 1%.

10 A. Abdulai et al. / Journal of Development Economics xxx (2010) xxx–xxx

less in organic fertilizers (mulch and manure) in plots that are leasedthan in those that are owned are consistent with the recent findingsby Jacoby and Mansuri (2008).

Trees are more likely to be planted by farmers with highereducation, more assets, and larger plot sizes. In particular, educationappears to have a positive and significant impact on all fourinvestment options, a finding that is in line with the human capitaltheory. Age exerts a negative and significant effect on both treeplanting and the application of manure. Thus, controlling for tenurearrangements and other farmer's and plot-level characteristics olderfarmers appear to be less likely to invest in trees and manure. This isprobably because younger farmers cultivate the land for a longer time,and as such are in a better position to benefit from the returns frominvestments in soil-improving measures even in the distant future. Inparticular, if farmers are not credit constrained and take futuregenerations into account, younger farmers will be more likely toinvest in conservationmeasures than older ones. Almost all the villagedummies are significantly different from zero, indicating significantcluster effects, and probably revealing agro-climatic variation andaccess to infrastructure.26 As noted by Besley (1995), they could alsobe representing village-level variation in tenure arrangements.

The estimates with household fixed effects are presented inTable 3. The results indicate that the magnitudes of the coefficientsrepresenting the tenure arrangement variables are slightly lower thanthose without household fixed effects. However, the positive impactof the secure tenure rights on investment in the productivity-enhancing measures, as well as the negative effects of shared-cropping and fixed-rent tenancy on applying organic fertilizer remain.This suggests that unobserved household effects are not responsiblefor the positive and significant impact of secured property rights oninvestment decisions.

Given that the coefficients presented in Tables 2 and 3 indicate theimpact of the explanatory variables on the probability of each choicebut not the marginal effects, we compute the marginal contributionsof the explanatory variables on the probability of investing inproductivity-enhancing measures. We are particularly interested inthe marginal effects of the land tenure arrangement variables. Weemploy the estimates from Table 3 which account for household fixedeffects. The marginal effects are evaluated at the means of theexplanatory variables. The standard errors of the marginal effects areestimated using the DELTA method (Greene, 2008). All the significantmarginal effects have the expected signs. Looking at the tenurearrangement variables, the evidence shows that being an owner-cultivator increases the probability of investing in soil-improvingmeasures between 11% and 49%. On the other hand being a fixed-renttenant tends to decrease the probability of investing in trees by 16%,mulch by 24%, andmanure by 12%. However, being a fixed-rent tenantincreases the probability of investing in mineral fertilizer by 16%,which is even higher than the 11% for an owner-cultivator.Furthermore, being a sharecropper reduces the probability ofinvesting in both manure and mulch by 27% and 32%, respectively,which are much higher than the reduction in probability of investingby fixed-rent tenants. These findings are consistent with the notionthat secured rights matter for investment in productivity-enhancingmeasures (Table 4).

6.2. Tenure arrangement and farm productivity

As pointed out by Jacoby and Mansuri (2008), it is worth askinghow yields would be affected by land ownership, given the resultspresented previously. We investigate the impact of tenure arrange-ments on plot-level productivity in this section. The results from this

26 The joint test of the null hypothesis that all district effects are equal using alikelihood ratio test gives a sample chi-squared value of 72.48 and a critical value atthe 1% level of 16.8.

Please cite this article as: Abdulai, A., et al., Land tenure differencesempirical analyses, J. Dev. Econ. (2010), doi:10.1016/j.jdeveco.2010.08.

instrumental variable analysis are presented in Table 5. Given thesignificant diversity of crops and intercrops on the plots, we employedvalue of crop output per acre as the dependent variable (Place andHazell, 1993; Place and Otsuka, 2002). Separate analysis for eachcropping pattern was not undertaken because of the relatively smallsample sizes that arise from the data set. Dummy variables forcropping patterns were however introduced in the regression tocapture the effects of the individual crops.

Given the potential endogeneity of the access to credit variable, itwas instrumented by first estimating a probit model of determinantsof access to credit and then using the predicted values in theproductivity estimation. This is because in some cases, land or a cropitself can be used as collateral to obtain credit. The results from thisfirst-stage regression are not presented for the sake of brevity, but areavailable from the authors upon request. The estimates in Table 5indicate a positive and statistically significant effect of the owner-operated with rights variable, suggesting that ownership of landresults in higher output. The results actually reinforce the finding thatsecured tenured rights facilitate investments in land-improvingmeasures or yield-enhancing inputs. This finding is consistent withresults reported by Banerjee et al. (2002), who found a positiveimpact of tenure reform on agricultural productivity for West Bengalin India. The results are also consistent with the findings by Goldsteinand Udry (2008) for Ghana, who showed that a great deal of potentialoutput is lost in the study area because land tenure is insecure.However, the findings contrast with those reported by Place andHazell (1993) and Place and Otsuka (2002), who found no significantrelationship between tenure and crop productivity in their studies.

and investment in land improvement measures: Theoretical and002

Table 4Marginal effects on the marginal probability of investment (in %).

Trees Mulch Fertilizer Manure

OWNER 0.3946(0.0817)

0.3853(0.0938)

0.1135(0.0421)

0.4877(0.1029)

SHARECROP 0.1108(0.0721)

−0.3190(0.0992)

−0.0523(0.0214)

−0.2756(0.0808)

FIXRENT −0.1609(0.0386)

−0.2375(0.0938)

0.1579(0.0362)

−0.1179(0.0392)

FEMALE −0.1190(0.0613)

−0.0761(0.0396)

+0.1476(0.0548)

0.0730(0.0317)

PLTFERT 0.0275(0.0108)

0.0735(0.0313)

0.0770(0.0269)

0.0878(0.0365)

PLTSIZE 0.1031(0.0792)

+0.1089(0.0421)

0.0812(0.0367)

−0.0443(0.0219)

PLTYRS 0.0077(0.0015)

0.0182(0.0098)

0.0179(0.0086)

0.0152(0.0084)

Notes: Standard errors of the estimated marginal effects are presented in parentheses.

11A. Abdulai et al. / Journal of Development Economics xxx (2010) xxx–xxx

The fixed-rent variable also showed a positive sign, but is notsignificantly different from zero at conventional levels, while thesharecropping variable is negative, but not significant. It is significantto note that the investments considered are either land-conserving orproductivity-enhancing inputs, and ownership tends to positivelyinfluence investment in these productivity-enhancing measures. Theresults also indicate positive and statistically significant effects ofaccess to credit and extension services. Plots farther away, as well asthose planted with crops such as cassava, beans and plantain alsoindicate positive and significant effects on productivity. The resultsalso show that incorporating household fixed effects does not changethe positive and significant impact of secured land rights on farmproductivity.

Table 5Instrumental variable estimates of determinants of productivity at plot level.

Pooled cross-section Household fixedeffectsa

Variable Coefficient t-value Coefficient t-value

CONSTANT 1.172⁎⁎⁎ 3.56 0.8632⁎⁎⁎ 2.71OWNER 0.4719⁎⁎ 2.39 0.2687⁎⁎ 2.26FIXRENT −0.2930 1.36 −0.1262 1.28SHARECROP −0.0176 1.18 −0.0527 1.02PLTSIZE −0.2268 1.52 −0.2479 1.61PLOTFERT 0.0768⁎ 1.91 0.0612⁎ 1.97PLANTAIN 0.1816⁎⁎ 2.26 0.0481⁎⁎ 2.24CASSAVA 0.4418⁎⁎⁎ 2.49 0.3316⁎⁎ 2.16VBEANS 0.3261⁎⁎⁎ 2.88 0.2488⁎⁎⁎ 2.87FEMALE 0.0378 0.93 0.026 1.49PCREDITb 0.6879⁎⁎ 2.31EXTEN 0.2714⁎⁎ 2.46LIVEST 0.0251 0.89HHSIZE 0.0187⁎ 1.87ETHNIC 0.2516 0.96AGE −0.0217⁎ 1.82TWIMEA −0.5843⁎⁎⁎ 3.48WORASO −0.1457 0.96AWOROPAT −0.2163⁎⁎ 2.13AYEREDE −0.1372 0.76DROMA 0.0178 1.38Adjusted R2 0.267F-statistic foroveridentification

6.89[0.00]

Number of observations 560 194

The p-values for the test statistic for the validity of the exclusion restrictions for theinstruments are given in square brackets.

a Household and village-level variables are dropped due to the inclusion ofhousehold fixed effects.

b Predicted values of credit from a first-stage credit regression used in the estimation.⁎⁎⁎ Significant at 1%.⁎⁎ Significant at 5%.⁎ Significant at 10%.

Please cite this article as: Abdulai, A., et al., Land tenure differencesempirical analyses, J. Dev. Econ. (2010), doi:10.1016/j.jdeveco.2010.08.

7. Conclusion

In this article, we developed a framework to examine therelationship between different land tenure arrangements and house-holds' investment in land-improving and conservation measures inthe Brong Ahafo region of Ghana. The land tenure arrangementsconsidered include owner-operated with full property rights, owner-operated with restricted rights, fixed-rent and sharecropping con-tracts. We employed variations in tenure arrangements betweendifferent plots to estimate plot-level regressions relating tenurearrangement to investment in tree planting, mulch, manure as well asmineral fertilizer application. We also examined the relationshipbetween land tenure arrangements and farm productivity.

The empirical results are consistent with our theoretical findingsand show that secured land rights tend to facilitate investment in soil-improving and natural resource management practices. In particular,farmers who owned land with secured tenure were more likely toinvest in tree planting, mulch, manure, but not in mineral fertilizer.Farmers on fixed-rent and sharecropping contracts were found to beless likely to attract investments in soil-improving measures such asmulch and organic manure, although fixed-rent farmers were morelikely to invest in yield increasing inputs such as mineral fertilizers.The positive impact of better land rights on investment remainedunchanged when we introduced household fixed effects into thespecification. As pointed out by Jacoby and Mansuri (2008), hold-upproblems in the sense of lack of full commitment on the part oflandlords could be driving the findings for fixed-rent farmers. Thepositive association of tenancy duration with investment in trees,mulch and organic manure also suggests that making temporaryrights longer would go a long way to enhance investments in soil-improving measures.

We also examined the impact of tenure arrangements on farmproductivity, using an instrumental variable approach. The resultsshowed a positive and significant effect of tenure security on farmproductivity, a finding that reinforces the significance of tenuresecurity in encouraging higher investment in soil-improving mea-sures. Access to credit was also found to be positively related to cropproductivity, suggesting that financial constraints may be a hindranceto investments in productivity-enhancing measures. The incorpora-tion of household fixed effects did not change the positive andsignificant impact of secured land rights on farm productivity. Themajor policy implication of these findings is that, ensuring tenurearrangements that confer permanent or sufficiently long temporaryrights to cultivators would enhance investment in both soil-improving and natural resource management practices. In addition,the results provide productivity-based arguments for enhancingfarmers' access to capital.

Acknowledgements

The authors have benefited significantly from the comments andsuggestions of three anonymous reviewers andMark Rosenzweig. Thethird author acknowledges financial support of the Ministerio deCiencia y Tecnología, Grant ECON 2010-17020, the Barcelona GSE andof the Government of Catalonia. The Netherlands Organization forScientific Research (NWO) and Amsterdam International Institute forDevelopment (AIID) also supported the collection of data for thisstudy. The usual disclaimer applies.

Appendix A

Demonstration of Observation 5

If the long-run soil capital of the owner-cultivated, Sl, is above theinitial value of the soil capital S0, we refer to it as case I, and if it isbelow the initial value of S0 we refer to it as case II. For case I, Fig. 1B

and investment in land improvement measures: Theoretical and002

12 A. Abdulai et al. / Journal of Development Economics xxx (2010) xxx–xxx

shows that owners initially apply more organic fertilizer thansharecroppers or fixed-rent tenants, if the soil improvement effectof organic fertilizer outweighs soil degradation. Consequently, soilcapital increases over time and the line pfXOr

evaluated at S(t)NS0,denoted by pfXOr

, with S(t)NS0, moves to the left since the cross de-rivative of fXOrS is negative. Hence, as time advances, owners reduce thelevel of organic fertilizer from X

OOr to

�XOrO. This reduction of the

organic fertilizer will be somehow less than graphed in Fig. 1B, be-cause an increase in soil capital displaces the graph of pOr+λS(−h′+δfXOr

)bpOr downwards. The reason for this change is the decrease inthe shadow price of soil capital λS resulting from an increase in soilcapital, and in the value of fXOr

due to the fact that fXOrS is negative.However, the impact of a change in fXOr

is likely to be far less for thegraph of pOr+λS(−h′+δfXOr

)bpOr than for the graph of pfXOr, since the

change in fXOris moderated in the former graph by the constant value

of pOr, whereas there are no constant values in the latter graph. Tofocus on the main issues this movement is not presented.

For case I, we also observe in Fig. 1B that sharecroppers and fixed-rent tenants apply less organic fertilizer. Therefore, these farmers mayeither build up the soil capital as well but to a lower degree thanowners (case Ia), or they deplete soil capital (case Ib). For now let usconsider the latter case. If sharecroppers and fixed-rent tenantsdeplete soil capital, the function pfXOr

evaluated at S(t)bS0, denoted bypfXOr

, with S(t)bS0, moves to the right and the optimal amount oforganic fertilizer applied by sharecroppers increases from XOr

Sh to�X ShOr

and from XOrTe to

�X TeOr for the case of fixed-rent tenants. The increase in

organic fertilizer applied by the sharecroppers is also somehow less

Table A1First-stage estimations of determinants of land rights.

Without householdfixed effects

With householdfixed effects

Fullrights

Fixed-rent

Sharecropping

Fullrights

Fixed-rent

Sharecropping

CONSTANT 0.056(1.39)

0.023(1.10)

0.133(1.29)

0.061(1.41)

0.028(1.26)

0.130(1.18)

PLTYEARS 0.046(2.18)

−0.116(1.61)

0.072(2.08)

0.049(2.25)

−0.108(1.69)

0.068(2.17)

PLOTFERT 0.212(2.26)

0.247(2.58)

0.077(1.23)

0.203(2.31)

0.236(2.66)

0.106(1.49)

PLOTLOC 0.074(1.93)

−0.038(2.13)

0.106(1.37)

0.041(2.28)

0.119(2.24)

0.108(1.44)

PLTYEARS 0.026(1.82)

0.012(0.693)

0.008(1.47)

0.032(1.88)

0.016(1.03)

0.012(1.56)

PLTSIZE 0.220(1.62)

0.286(1.39)

0.083(0.79)

0.218(1.73)

0.247(1.52)

0.079(1.06)

PLTDIST 0.031(1.85)

0.172(1.68)

0.088(2.16)

0.034(1.92)

0.184(1.76)

0.115(2.28)

LAND INHERITED 0.121(2.59)

0.128(2.66)

FEMALE 0.018(0.63)

−0.007(1.42)

0.016(1.17)

ETHNIC 0.172(2.82)

−0.043(1.98)

−0.133(1.67)

AGE −0.017(1.26)

0.012(0.78)

0.018(1.07)

EDUCN 0.038(1.53)

0.186(1.05)

0.075(1.28)

IMPLTS 0.220(1.62)

0.286(1.39)

0.083(0.79)

HHSIZE 0.031(1.85)

0.172(1.68)

0.088(2.16)

Village dummyvariables

Yes Yes Yes No No No

F-statistics[p-values]

26.7[0.00]

23.4[0.00]

21.8[0.00]

F-test of instruments[p-values]

14.72[0.00]

7.86[0.00]

9.13[0.00]

Observations 560 194

Please cite this article as: Abdulai, A., et al., Land tenure differencesempirical analyses, J. Dev. Econ. (2010), doi:10.1016/j.jdeveco.2010.08.

than graphed in Fig. 1B because the line pOr+γ1pfXOrmoves upward

due to the fact that fXOrS is negative. However, the impact of a change infXOr

is likely to be far less for pOr+γ1pfXOrthan for pfXOr

, since γ1 isbetween 0 and 1. To focus the discussion on the most likely cases wedo not discuss case Ia here, but it can be derived easily from Fig. 1B.

For case II, the owners' objective is to depreciate soil capital, sincethe soil improvement effect of organic fertilizer does not outweigh soildegradation. In this case, Fig. 1B shows that the line pfXOr

moves to theright and over time owners apply more organic fertilizer. The optimalapplication rate changes from XOr

O to�XOOr . Fig. 1B also shows that

sharecroppers and fixed-rent tenants apply more organic fertilizerthan XOr

O , the amount applied by owner-cultivated. Hence, share-croppers and fixed-rent tenant may either deplete soil capital,although to a lesser degree than owner-cultivated (case IIa), or theymay build up the soil capital (case IIb). To focus the discussion on themost likely cases, a detailed discussion of cases IIa and IIb is notpresented here. Following the same line of arguments these cases canbe derived easily from Fig. 1B.

With respect to optimal long-run application of mineral fertilizerFig. 1A shows for case I (increase of soil capital by the owner) thatowners apply less mineral fertilizer, i.e., XM

O declines to�XOM . If soil

capital deteriorates under sharecropping and fixed-rent tenancy (caseIb), Fig. 1A shows that the application rate of mineral fertilizerincreases from XM

Sh to�X ShM for sharecroppers, and from XM

Te to�X TeM for

fixed-rent tenants as time advances. For the sake of brevity, theremaining cases (Ia, IIa and IIb) are not presented here but can bederived directly from Fig. 1A.

Demonstration of Observation 6

If land is relatively scarce and fruit production is of minorimportance, the term μ2−pFϑ will be strictly positive. In this case,fixed-rent tenants tend to decreaseW, fromW* toWTe as it can be seenfrom Fig. 2. However, if fruit production plays an import role and theopportunity cost of land are small, (αμ2−pFϑb0), the opposite resultmay arise, i.e. WTe will be situated to the right of W*. Similarly, weobserve in the case of an owner-cultivated that the optimal Wdecreases from W* to WO or WO if land is relatively scarce and fruitproduction is of minor importance. The opposite result holds, if land isnot scarce and fruit production is quite important. However, it cannotbe determined whether the decrease of the owner is below or abovethe optimal W of the tenant WTe. A similar argument applies for asharecropper.

References

Angelsen, A., 1999. Agricultural expansion and deforestation: modelling the impact ofpopulation, market forces and property rights. Journal of Development Economics58, 185–218.

Banerjee, A., Ghatak, M., 2004. Eviction threats and investment incentives. Journal ofDevelopment Economics 74, 469–488.

Banerjee, A., Gertler, P., Ghatak, M., 2002. Empowerment and efficiency: tenancy reformin West Bengal. The Journal of Political Economy 113, 151–183.

Barbier, E.B., 2004. Explaining agricultural land expansion and deforestation indeveloping countries. American Journal of Agricultural Economics 86 (5),1347–1353.

Benneh, G., 1989. The dynamics of customary land tenure and agrarian systems inGhana. In: Ghana, P. (Ed.), WCARRD; Ten Years of Follow-Up, the Dynamics of LandTenure and Agrarian Systems in Africa, Case Studies from Ghana, Kenya,Madagascar and Togo. FAO, Rome, Italy, pp. 34–97.

Besley, T., 1995. Property rights and investment incentives: theory and evidence fromGhana. The Journal of Political Economy 103, 903–937.

Besley T. and M. Ghatak (2009). Property rights and economic development.Forthcoming in D. Rodrik and M. Rosenzweig (ed.s) Handbook of DevelopmentEconomics V, North Holland, 2009.

Binswanger, H.P., Rosenzweig, M.K., 1986. Behavioral and material determinants ofproduction relations in agriculture. Journal of Development Studies 22 (3),503–539.

Blundell, R.W., Smith, R.J., 1989. Estimation in a class of simultaneous equation limiteddependent variable models. Review of Economic Studies 56, 37–57.

and investment in land improvement measures: Theoretical and002

13A. Abdulai et al. / Journal of Development Economics xxx (2010) xxx–xxx

Brasselle, A., Frederic, G., Jean-Philippe, P., 2002. Land use security and investmentsincentives: puzzling evidence from Burkina Faso. Journal of DevelopmentEconomics 67, 373–418.

Carter, M.R., Olinto, P., 2003. Getting institutions right for whom? Credit constraints andthe impact of property rights on the quantity and composition of investment.American Journal of Agricultural Economics 85, 173–186.

Davidson, R., Mackinnon, J.G., 1993. Estimation and Inference in Econometrics. OxfordUniversity Press, New York.

Deininger, K., Ali, D.A., 2008. Do overlapping land rights reduce agriculturalinvestment? Evidence from Uganda. American Journal of Agricultural Economics90, 869–882.

Dorner, P., 1972. Land Reform and Economic Development. Harmondsworth; PenguinBooks.

Ehui, S.K., Hertel, T.W., Preckel, P.V., 1990. Forest resource depletion, soil dynamics, andagricultural productivity in the tropics. Journal of Environmental Economics andManagement 18 (3), 263–275.

Feder, G., Feeny, D., 1991. Land tenure and property rights: theory and implications fordevelopment policy. World Bank Economic Review 5, 135–153.

Gildea, R.Y., 1964. Culture and land tenure in Ghana. Land Economics 40, 102–104.Goldstein, M., Udry, C., 2008. The profits of power: land rights and agricultural

investment in Ghana. Journal of Political Economy 116 (6), 981–1022.Greene, W.H., 2008. Econometric Analysis, 6th Edition. Prentice Hall, Upper Saddle

River, NJ.Harrison, P., 1987. The Greening of Africa. Paladin Grafton Books, London.Jacoby, H.G., Mansuri, G., 2008. Land tenancy and non-contractible investment in rural

Pakistan. Review of Economic Studies 75, 763–788.Jacoby, H.G., Mansuri, G., 2009. Incentives, supervision, and sharecropper productivity.

Journal of Development Economics 88, 232–241.Jacoby, H.G., LI, G., Rozelle, S., 2002. Hazards of expropriation: tenure insecurity and

investment in rural China. The American Economic Review 92, 1420–1447.Lee, L., 1992. Amemiya's generalized least squares and tests of overidentification in

simultaneous equation models with qualitative or limited dependent variables.Econometric Reviews 11, 319–328.

Please cite this article as: Abdulai, A., et al., Land tenure differencesempirical analyses, J. Dev. Econ. (2010), doi:10.1016/j.jdeveco.2010.08.

Marenya, P.P., Barrett, C.B., 2007. Household-level determinants of adoption ofimproved natural resources management practices among smallholder farmersin Western Kenya. Food Policy 32, 515–536.

Migot-Adholla, S.E., Banneh, G., Place, F., Atsu, S., 1994. Land, security of tenure, andproductivity in Ghana. In: Bruce, J.W., Migot-Adholla, S.E. (Eds.), Searching for LandTenure Security in Africa. Kendall/Hunt Publishing Cy, Dubuque, pp. 97–118.

Noronha, R., 1985. A Review of Literature on Land Tenure Systems in Sub-SaharanAfrica. World Bank, Agriculture and Rural Development Department, Washington,D.C.

Otsuka, K., Quisumbing, A., Payongayong, E., Aidoo, J.B., 2003. Land tenure and themanagement of trees: the case of customary land tenure areas of Ghana.Environment and Development Economics 8, 77–104.

Pinkney, T., Kimuyu, P., 1994. Land tenure reforms in East Africa: good, bad orunimportant? Journal of African Economies 3.

Place, F., Hazell, P., 1993. Productivity effects of indigenous land tenure systems in Sub-Saharan Africa. American Journal of Agricultural Economics 75, 10–19.

Place, F., Otsuka, K., 2002. Land tenure systems and their impacts on agriculturalinvestments and productivity in Uganda. Journal of Development Studies 38 (6),105–128.

Quisumbing, A., Aidoo, J.B., Payongayong, E., Otsuka, K., 2001a. Agroforestrymanagement in Ghana. In: Otsuka, K., Place, F. (Eds.), Land Tenure and NaturalResource Management: a Comparative Study of Agrarian Communities in Asia andAfrica. John Hopkins University Press, Baltimore, Maryland, pp. 53–96.

Quisumbing, A., Payongayong, E., Aidoo, J.B., Otsuka, K., 2001b. Women's land rights inthe transition to individualized ownership: implications for the management oftree resources in Western Ghana. Economic Development and Cultural Change 50,157–182.

Rivers, D., Vuong, Q.H., 1988. Limited information estimators and exogeneity tests forsimultaneous probit models. Journal of Econometrics 39 (3), 347–366.

Wooldridge, J.M., 2002. Econometric Analysis of Cross Section and Panel Data. MITPress, Cambridge, Massachusetts.

and investment in land improvement measures: Theoretical and002