Factor Market Imperfectons and Rural Land Rental Markets n Northern Ethopan Hghlands By Hosa ena Ghebru and Sten Holden Department of Economcs and Resource Management Norwegan Unversty of Lfe Scences P.O. Box 5033, N_1432, Ås, Norway Emal: hosaenag@yahoo.com Abstract In response to the natonal land polcy that bans the formal land markets (sales market) and the nfrequent admnstratve land redstrbuton, the nformal land rental markets n Ethopa plays a pvotal role n smoothng the mperfectons n other farm factor markets. A three round household panel data set (1998, 2001 and 2003) from the northern hghlands of Ethopa (Tgra) has been used to emprcally explore the determnants of partcpaton and ntensty of transacton n the land rental market and assess the role of the recent regonal polcy reform that halts further land redstrbuton and executes certfcaton of rural land ownershp ttles. Results from the study ndcate that the land rental market (whch nvolves 53% of the sample respondents) serves as a crucal venue to correct the dsproportonal dstrbuton factor endowments and the mperfectons n the non-land factor markets. However, frctons n the land rental market tself cannot be dened whch was found to asymmetrcally (unequally) affect the demand and supply sdes of the tenancy market. Though the study does not provde evdences to support the barganng power of farm households, the large number of constraned tenant households as compared to landlords mples that, even f tenants are wth better endowments of farm asset and other durable asset endowments, the fragmented land holdng may cause the supply constraned tenancy market that favours landlord households. Thus, polcy measures to facltate the transferablty of land (legal enforcement of lease contracts) and functonng of other farm nput markets can sgnfcantly enhance the allocatve effcency role of the rural land rental markets. Key words: Land tenancy, market partcpaton, transacton cost, selecton bas, Ethopa
I. INTRODUCTION Wth markets beng perfect, rentng or buyng land would be dentcal. Tenants would pay rent to landlords and buyers would pay equvalent amounts to bank as nterest to loans. In a world, however, wth asymmetrc nformaton (costly or unavalable) and market mperfectons n captal and varous agrcultural nputs markets, there are good reasons for farm households (landlords and tenants) to prefer transferrng land through land lease markets than the formal land markets (buyng and sellng). Here tenancy contracts are servng as way of tacklng the market mperfectons n the land, labour, captal and other agrcultural nput markets. (Sadoulet et al: 1998). Besdes the acute market mperfectons for farm factors of producton and hgh transacton cost of non-land factor markets, the Ethopan consttutons (1995) gves the land rental markets an exceptonal mportance makng t the only venue of land transfer among farmers. In the current consttuton (Ethopan consttuton - artcle 40 / 1995) the rght of ownershp of rural and urban land, as well as natural resources, s exclusvely vested n the state and the peoples of Ethopa. It further states that land s a common property of the natons, natonaltes and peoples of Ethopa and shall not be subject to sale or any other means of exchange. In ths context of prohbted land sales, mountng land scarcty and reduced frequency of land reallocaton, land rental markets take a partcular mportance wth possble effcency and equty effects. Indeed, to the extent that land use rghts cannot be gven to all farmers, access to land for cultvaton or pasture can only be acheved through short-term rental agreements between farmers wth and wthout (enough) land. Had t not been for land rental markets, such legal ban on the land sale and purchase, and some reservatons wth regard to the admnstratve land redstrbuton would have put the country n a very pressng problem of hgh market gap concernng the demand and supply of rural arable land. Owng to ths fact, the functonng of such rural land rental markets affect, n one way or another, the functonng of the overall economy and the strve towards poverty reducton. As the consttuton permts, short-term leases, such as rental and sharecroppng, are practced all over the country [Ethopa] n response to land scarcty n all regons. 2
Moreover, the regonal land polcy reform n the study area (Tgray) that halts admnstratve land redstrbuton and certfcaton of land ownershp ttles magnfy the role the nformal land lease market could play n provdng access to land and facltate households farm resource adjustment decsons. The rural land tenancy market has receved ample attenton, both theoretcally and emprcally. Much of the attentons, however, has focussed on the effcency aspects of alternatve land tenure contracts (Otsuka, Chuma, and Hayam, 1992; Sngh, 1989; Otsuka and Hayam, 1988; Pender and Fafchamps, 2001; Stgltz, 1974; Sadoulet et al., 1997; Gavan and Ehu, 1999 and Ahmed et al., 2002) whle consderable studes have also been made on factors explanng farm households choce among varous tenure arrangements (Chaudhur and matra, 1996; and Allen and Lueck, 1992). Majorty of the studes, however, fal to address the prmary factors nfluencng the frst hand decson of households partcpaton n the tenancy market. Handful researches have been made wth partcular emphass on the role of factor market mperfectons as an mportant factor explanng the leasng behavour of households (Blss and Stern, 1982; Bell, 1977; Nab, 1985; and Skoufas, 1995). These emprcal lteratures on the leasng behavour of households, however, are domnated by studes conducted n South and Southeast Asa leavng the Sub-Saharan Afrca (SSA) case to be stll under explored. Wth partcular reference to Ethopa, despte the hghly debated ssue of land tenure system n the country that lasts for several decades, few studes have been conducted that address the ssue of land tenancy market (Pender and Fafchapmps, 2001; Kebede, 2003; Dennger, 2003; and Teklu, 2004). Smlar study by Tkabo and Holden (2003) has also been conducted usng data from hghlands of Ertrea, whch experence, more or less, smlar farmng and agro-ecologcal features wth the study area Tgray, Ethopa. To the best of our knowledge, no study has been conducted to emprcally explore the role of transacton cost (factor market mperfectons) n dctatng the land tenancy market n the country. The whole theme of ths paper, thus, s to help fllng ths research gap n the country and the SSA at large and, thereby, evaluates the mpact of the recent polcy reform that started to ndvdualze land rghts and elmnate the scope for further land redstrbuton. 3
In an attempt to explore the role of factor market mperfecton n dctatng the leasng behavour of households, the study focuses on the followng three broad objectves: ) to nvestgate factors that affect farm households probablty of partcpatng (market entry) n the nformal land rental markets, ) to assess the role of factor market mperfectons n the farm resource adjustment decsons (ntensty of partcpaton) of household, and ) to examne the barganng power status (market symmetry) of the tradng partners - landlord and the tenant. We address these ssues by analysng a household survey data from 400 sample households from 16 vllages of Tgray northern hghlands of Ethopa. 1 Conceptually, we buld upon the model wth factor market mperfectons (see Holden et al., 2001). The overall results of the study ndcate the unque structure of the tenancy market whch contradcts the predomnantly customary lterature of rch landlord and poor tenants. Supportng our basc hypothess, the farm asset endowment varables oxen and the tropcal lvestock unt other than oxen (TLUOXP) were found sgnfcantly (wth OXP at 1% level) affectng farm households lkelhood of partcpaton ndcatng households wth relatvely poor endowment of such non/sem-tradable farm nputs tend to rent-out ther land whle the vce versa holds true for the tenant households. The household relatve labour (male and female) endowments, however, were found to have weak effect on the lkelhood of partcpaton on both sde of the tenancy market whch can be asserted as a sgn of relatve abundance of farm labour or a relatvely better functonng labour market wth respect to oxen. Lke the cases of non-land factor markets, the regresson results show the persstence of frctons (mperfectons) n the tenancy market that cause an ncomplete factor rato adjustment process through the land lease market. The emprcal evdence shows 38% of the non-partcpant households are reported as constraned households who nvoluntarly assume the autarky poston of the tenancy market whle 33% of the partcpant (tenant and landlord) households stll clam to operate less than ther optmal level of land transacton. The smultaneous test for a balanced (equal) market response of two sdes of the market (the demand and supply sdes) shows the hypothess of asymmetrc 1 Due to ncompleteness of data and respondent dropouts, the entre analyss n ths paper consders 372 sample respondents only. 4
(unbalanced) land lease market could not be rejected at 1% level of sgnfcance. In support of ths argument, the emprcal evdences show 56 % of the tenant households operate (transact) less than ther desred level - they manage to obtan 64% of ther desred level of transacton. On the contrary, t s only 5% of the landlord households that are dentfed as constraned agents though, n aggregate terms, they manage to transact 98% of ther desred level of land transacton. The rest of the thess s organzed as follows. Charter 2 presents the conceptual framework of the paper wth some testable hypotheses. Chapter 3 presents the conceptual framework ncludng the testable hypotheses. Estmaton methods (econometrc models) utlzed to analyse the probablty and ntensty of land rental transactons are addressed n the fourth chapter. In addton, descrpton of the study area, the method of data collecton and statstcal descrpton of the data are dscussed n ths chapter. Chapter 5 s devoted for results and dscusson of the emprcal fndngs of the study. Summary and conclusons of the study wth mportant polcy mplcatons appear n chapter 6. II. Conceptual frame work Wth well functonng markets and perfect nformaton, no or lttle economc justfcaton can possbly be gven for the exstence of rural land lease market. In a world of perfect nformaton and complete markets wth zero transacton costs, the ownershp and dstrbuton of land only affects households welfare wthout havng any effcency outcomes. Ths s so snce resource adjustment for effcency purpose can be made n any of the factor markets: Land, labour, oxen, etc. Thus, the common exstence of the nformal land tenancy market n many developng socetes shows that the standard economc theory (wth market clearng prce) cannot explan such economes. A typcal feature of such agraran economes s the tenacous persstence of mperfectons n a number of factor markets. Thus, the absence of certan markets and the functonng of others s n response to the behavoural and materal factors that nfluences household producton and consumpton decsons. Varous lteratures (Bnswanger and Rosezweng, 1986; De Janvry et al., 1991) show that, n such rural 5
settngs, factor markets, especally non-land factor markets, wll ether be hghly mperfect or smply non-exstent. Thus, factor rato adjustments through land rental market partcpaton can only be justfable by comparng the relatve costs assocated wth transactng of such farm nputs (n ths case, only when resource adjustment costs through transactons n Nonland (NL) farm nput markets) out-weghs the land lease transactons. Gven ths stuaton, the development of rural land rental market plays a unquely mportant role n factor rato adjustment process. For nstance, havng an alternatve decson ether to take part n hrng-out excess NL nput or rentng (leasng)-n farm land, a ratonal cost effectve farm household opts for the later only under the condton of prohbtvely hgh mperfecton (transacton costs) n the NL factor markets. Land lease markets, therefore, facltate easy transfer of land to productve households wth dfferent ntal relatve endowment of farm nputs. Wth consderable varatons n methods and model specfcatons, recent emprcal works, from Asa, Latn Amerca and Afrca, wtnessed that the relatve ease of cost of transacton n the land rental market makes t more preferable alternatve venue of factor rato adjustment. Gvng more emphass on leasng as an adjustment of land area cultvated to farm nputs (Explctly mentoned as labour and bullock) fxed for the farm household, Blss and Stern (1982) developed a land leasng model usng data from the Indan vllage of Palanpur. Descrbng Desred Cultvated Area (DCA) as an area whch accords wth the avalablty of non-land farm nput (labour L and Bullock O ) supples for the household, the model, then, consdered leasng n or our as an adjustment of land owned ( A ) towards DCA whch they call the dfference as Net land leased-n (NLI). Showng the possble mperfectons n the land lease market, the model defned DCA and NLI as: (1) DCA = f ( L, O ), and (2) NLI = h(dca - A ) Lnear approxmaton by a frst-order Taylor seres expanson of equaton (2) yelds, 6
(3) NLI = K + φf L + φf O φa Where K s the constant term; ф = f / L and f / O L O h / DCA and f L and f O respectvely are Thus, the BS model was estmated usng a reduced form econometrc expresson of equaton 3 as NLI = β 0 + β1 L + β 2 O β 3 A + e, wth β 3 showng the mperfecton n the land rental market. Accordngly, they argued that as β 3 (coeffcent of the owned farm sze) converges to -1, t has a clear evdence of a perfectly functonng land rental market showng a complete adjustment process. Overall, ther study from Inda shows the mperfect functonng of the tenancy market wth a coeffcent of β 3 = -0.78 whch was found sgnfcantly dfferent from -1. However, the BS model may not be a plausble approach snce varables common to decson of leasng-n and leasng-out may have dfferent drectons and degree of nfluence n the two sdes (the demand and supply sdes) of the land rental market. (Tkabo and Holden: 2003) In addressng the problem of falure to capture zero values of the dependent varable, varous successve studes have utlzed a censored tobt model. Studes by Pender and Fafchamps (2000); Kebede (2003) and Dennger et al. (2003) all from Ethopa and studes from Skoufas (1995) were few among the recent studes that utlzed the tobt model n analysng factors explanng the leasng behavour of households. Usng data from four vllages of Ethopa, Pender and Fafchamps (2000) addressed the ssue of leasng decsons from the demand sde of the market (leasng-n decsons). Ther maxmum lkelhood estmates support the argument that mperfecton n the non-land factor markets (oxen market) to cause farm households leasng n decsons for the oxen endowment varable to have a postve sgnfcant nfluence. Consstent wth the fndngs of BS, the emprcal evdence from the study shows a pronounced mperfecton n the land tenancy market wtnessed by the sgnfcant postve correlaton between the area under cultvaton (operatonal land) and the amount of land owned. Contrary to studes by Kebede (2003) from Ethopa whch further dentfes lvestock holdng as an mportant factor dctatng the tenancy market partcpaton, labour endowment of the farm household was reported to have weak effect on the leasng behavour of households. 7
Usng a panel data from sx vllages of Inda, Skoufas (1995) separately analyse the demand and supply sde of the market for further test of symmetry between the two sdes of the market. The emprcal evdence shows that mperfectons n non-land factor markets are the key factors explanng the arrangements of the nformal tenancy market. The fndngs from the study descrbed leasng decsons as ways of poolng land to non/sem-tradable factors of producton lke famly labour and bullocks. Furthermore, the study wtnessed consderable mperfectons n the land lease market wth asymmetrc (unbalanced) mpacts on the demand and supply sdes of the market. Separately treatng the two sdes of the market, the tobt results of Dennger et al. (2003) are consstent wth fndngs of other fellow studes from Ethopa magnfyng the mportance of agrcultural ablty (relatve farm resource endowment) as key factor n dctatng the leasng behavour of households. Robust wth studes from Pender and Fafchamps (2000) but contrary to fndngs of Kebede (2003), the famly labour endowment was not found to have a sgnfcant effect on the leasng decson of households. Showng the persstence of mperfecton n the oxen market, however, the lvestock holdng (notably oxen) of farm households sgnfcantly affects household partcpaton decsons. Here, t s mportant to note that, n all the three studes from Ethopa, the number of bullocks owned by a farm household was found to play a leadng role n the tenancy market showng the hgh mperfecton n the sales and rental markets of such farm resource. Holden and Shferaw (2001) confrmed ths wth ther study from the central hghlands of Ethopa where they found lqudty constrants and moral hazard nvolved n the oxen rental market to cause the mperfecton n both sales and rental markets. Such studes wth the censored tobt model, however, were not free from crtcsm snce they mposes a smplfyng assumpton that the same set of factors to nfluence the decson to partcpate and decsons of how much to transact. Thus, the sequental nature of the leasng decsons wth dfferent sets of factors explanng the frst-hand decson of partcpaton and the extent of transacton may cause a selecton bas problem wth censored tobt results. 8
To escape the persstence of such sample selecton problem where dfferent factors nfluencng the partcpaton and ntensty of transacton, studes by Teklu et al. (2004) and Tkabo and Holden (2003) adopted the two stage Heckman (1979) selecton model n analysng factors explanng the leasng behavour of households wth the later study treatng the two sdes of the market separately. Usng data from three vllages of Ethopa, Teklu and Lem (2004) analyse the leasng decson of households as a sequental decson between partcpaton and ntensty decson focusng on the supply sde of the market (leasng-out decson). The Heckman estmaton results of the study found the relatve endowment of famly labour and allocated farm land sze sgnfcantly affect landlords leasng decson though t couldn t reject the weak effect of household lvestock holdng on the partcpaton as well as ntensty of transacton. Smlarly, the study by Tkabo and Holden (2003) usng data from the hghlands of Ertrea found the skew ness of dstrbuton and mperfectons n the non-land farm nputs (both labour and bullocks) play an mportant role n nfluencng the leasng decson of households. Moreover, the study also revealed the non-symmetrcal nature of the lease market wth potental tenants facng a prohbtvely hgh transacton cost that caused severe entry barrers. Unlke smlar studes from Inda and Ethopa, the results from the study wtnessed a relatvely well functonng tenancy market n Ertrea wth complete resource adjustment process n the nformal rural land lease market. III. The Model and hypotheses As mentoned before, the mssng formal land markets (sales market) and farly egaltaran land dstrbuton n the country (whch takes less, f not no, account of household farmng ablty) call for farm households to adjust farm nput combnatons ether through partcpaton n the non-land farm nput markets or the nformal land (land lease) markets. 9
The basc hypothess s that, n an attempt to adjust farm nput combnatons, the persstence of relatvely hgh mperfectons n non-land factor markets as compared to the nformal land markets determnes farm households partcpaton and level of transacton n the land rental market. These mperfectons can be attrbutable to endogenous factors related to each respectve household and/or other factors whch are exogenous to the household. Thus, gven the rural settng n the study area (where land sellng s legally banned) and revew of some related lteratures, a theoretcal model s developed that ncludes mperfectons n land and non-land factor markets. The model s developed under the assumpton that the non/sem-tradablty of nonland farm nputs affect households decsons to take part n land lease markets. Owng to the dosyncratc producton actvtes n the study area due to ran-fed nature of cultvaton system, market frcton s beleved to be relatvely hgher n the non-land factor markets as compared to the land rental market. Accordngly, t s postulated that household specfc farmng ablty and varaton n relatve endowment of land and non-land farm nputs across farm households partcpaton n the tenancy market. The model below uses Holden et al. (2001) as a sprng board. Let the representatve farm household be endowed wth cultvable land (Ā), labour supply ( L _ ), and oxen power (Ō). Wth smplfyng assumptons of: 1) perfectly functonng farm output market, 2) complementary relatonshp between farm land and non-land farm nputs; and 3) prohbtvely hgh transacton cost of exchangng oxen servces, the farm household total ncome can be derved from: 1. Farm actvty 2. Labour ncome (off-farm employment) 3. Land rental market partcpaton. 1. Farmng actvty Here, the representatve household crop producton actvty follows the standard producton functon whch s gven by: Q = f (A, L a, Ō ) Where: A s operatonal land (area under cultvaton) L a s labour power devoted for agrcultural actvty 10
Ō s oxen power whch s fxed 2 The producton functon Q satsfes the standard assumptons of producton functon: f La, f A and f o > 0 and f LaLa, f AA, and f oo < 0 as well as complementartes between farm nputs. 2. Labour ncome (farm and off-farm employment) A second possblty to generate ncome for the farm household s by devotng L H = L L a (L H s postve, zero and negatve for households that hres n labour, non partcpant and hres out labour, respectvely) amount of labour n the labour market at the endogenously determned shadow wage w Ψ. Smlar to the oxen exchange market, wth the dosyncratc nature of crop producton actvty and farly egaltaran land dstrbuton, t s assumed that farm labour employment and ncome generaton s lmted unlke the off-farm labour employment. 3. Land rental market As part of adjustment of farm nput combnaton through land lease market partcpaton, farm households can lease out part of ther land endowment or lease n addtonal land for farm producton (A R = Ā A ) ( A R s postve, zero and negatve for households that lease n land, don t partcpate and lease out land, respectvely) where A R s net land rented-n and A s the operatonal land holdng. Takng all these elements together, we obtan the representatve household s ncome y as: Y = P f(l a, A, Ō) + r Ψ (Ā A ) + w Ψ ( L L a ) Where: P s market prce of farm output r Ψ s land rental value w Ψ s the wage rate Ψ Є {s, *, b} s : Leased out land / Hres out labour * : Self suffcent n land / labour b : Leased n land / Hres n labour Assumng the farm household maxmzes utlty from ncome, the maxmzaton problem wll be: Max U = V [P f{(ā + A R ), ( L + L H ), Ō)} + r Ψ A R + w Ψ L H ] 2 It s our pror belef that oxen as farm nput s fxed n response to the hgh lqudty constrant and seasonalty of crop cultvaton system that lmts the oxen sales and rental markets, respectvely, accompaned by the moral hazard nvolved n the later case. 11
A R L H Where: A R, L H Є {s, *, b} The utlty maxmzaton problem above satsfes the standard assumpton of a utlty functon (twce dfferentable quas-concave utlty functon wth postve and negatve frst and second dervatves, respectvely) and yelds the followng frst order condtons for the land and labour factors: By dfferentatng the above utlty maxmzng problem wth respect to A R (A s, A *, and A b ), we may derve the followng three FOC s: FOC 1 : FOC 2 : FOC 3 : f P s A = r s.. the prce a household face as seller (landlord) f P = r * the shadow prce of land (endogenous prce) A f P = r b the prce a household face as buyer (tenant) b A The prce bands - (r s, and r b ) show the persstence of transacton cost n farm nput markets (the sem tradablty of the factors) wth some households sellng, some non partcpant (dffcult to say self suffcent) and some buyng households whch may be a more realstc representaton of factor markets n Ethopa where a rental market exsts for farm nputs. The prce band reflects possble transacton costs may accrue due to search cost, moral hazard and seasonalty (labour market), mmoblty of land and transportaton cost (land markets) (Holden et. al. 2001). The sze of the prce band depends on the relatve degree of mperfecton showng r s < r b. Generally, farm households partcpaton n the land lease market and ther respectve poston n the market (landlord or tenant households) depend on two major factors: 1) the magntude or level of the endogenous factor prce r* (shadow prce of land), and 2) the sze of the prce band whch depcts the persstence and degree of transacton cost that dctates market prces as a seller and as a buyer. Ths ndrectly shows households leasng behavour (to partcpate or not as well as the ntensty of partcpaton) depends, much, on the relatve cost of transactng the land wth respectve to non-land farm nputs. 12
Shadow prce of land (r*) When factor markets are mperfect and prces become unobservable, the shadow prce (household specfc endogenous factor prce) reflects the relatve endowment of factors supply versus demand (n ths partcular case, land to non-land). For nstance, wth a farm household endowed wth relatvely hgher land to non-land farm nputs wll have a relatvely lower shadow prce of land (r * = f/ Ā = MVP Ā ) than households wth lower relatve endowment of land. Thus, shadow prces, here, largely dependent on household specfc farm factor endowments though t s not a suffcent condton to trgger household partcpaton n the nformal land tenancy market. In the absence of well functonng markets, farm households face dfferent sets of prces dependng on ther respectve postons (net seller or net buyer) they take n such mperfectly functonng markets. Such prce varatons for a net-buyer and a netseller (the prce band) reflect the degree of market mperfecton (level of transacton cost) whch could possbly be household specfc (owng to varous soco-economc geographcal and polcy/nsttutonal factors) or common (fxed) for all partcpants n the market. See the fgure below. Fgure 2 : Effects of transacton costs (prce bands) on land transactons prce r 1 * case 3 r b A b r 2 * case 2 r s A s r 3 * case 1 Adopted from: Sadoulet et al. (1995) Quantty of land 13
Where: case 1 = Condton for leasng-out land Case 2 = Autarky poston due to hgh transacton cost (wde prce bands) Case 3 = Condton for leasng-n land r s, and r b = Market prce observed n the supply and demand sde of the market A s and A b = Amounts of land leased-out and leased-n, respectvely r 1 *, r 2 * and r 3 * landlord households. = The shadow prces perceved by the tenant, non- partcpant and Based on these argument, a farm household may assume three alternatve postons n the nformal land tenancy market dependng on the two nfluental foreces mentoned above. Case 1: Landlords (leasng out land) when: f [ P s A = r s f ] [ P = r * ] A Where: r s : Market prce n the supply sde of the market (Landlord prce) r * : The shadow prce of land, Note: We use the sgn delberately to nclude the possble cases of ncomplete resource adjustments Case 2: Non partcpants households (Autarky) @ f [ P s A = r s f ] < [ P = r * f ] < [ P = r b ] b A A Where: r b : Market prce n the demand sde of the market (tenant prce) Case 3: Tenant (leasng n land) when, f [ P b A = r b f ] [ P = r * ] A Testable hypotheses Wth zero or neglgble transacton cost n farm nput markets, the skew ness of endowments n non-land farm nputs would rather not trgger the farm households decson to partcpate n the nformal rural land tenancy market for they can even out ths resource endowment gap by drect partcpaton n labour or oxen markets. However, owng to the very nature of crop cultvaton whch demands contnuous and 14
dynamc adjustment decsons, hred labour s far from a perfect substtute of famly labour where the later s more nvolved n sharng output rsks. Such supervson constrants on the labour market accompaned by the lqudty constrant and moral hazard nvolved n the oxen sales and rental markets, respectvely, magnfes the persstence of market mperfectons n the non-land factor markets. Therefore, Poolng non/sem-tradable nputs play a domnant role n rural household decsons to partcpate n land lease markets as well as ther level of transacton. Landlords wth low endowments of sem/non tradable nputs (such as famly labour male and female, oxen and other draft anmals) relatve to ther land holdngs wll look for potental tenants wth hgh endowments of such farm nputs. (Hypothess 1) The sem/non-tradablty and skewed dstrbuton of the non-land farm nputs among farm households cause farmland (physcally mmoble factor and farly dstrbuted factor) to be more moble than non-land farm factors (contrary to ts ease of physcal moblty and hghly skewed dstrbuton). (Hypothess 2) Smlar to the non-land farm factor markets, however, the persstence of producton rsk, mssng nsurance markets, hgh searchng and supervson cost and weak property rghts wth lack enforcement tools, may result n mperfectons n the rural land tenancy market (Sadoulet, et al., 2001; Stgltz, 1974; Cheung, 1969). Moreover, the nformal (poorly developed) nature of the tenancy market tself accompaned by the past natonal land polcy falures that legally ban any form of land transfer threatens the functonng of the rural land tenancy market. Exposng the nformal nature of the tenancy market, almost all of the sample respondents from the 2003 survey were reported to have no legal contractual agreements regardng ther land lease transactons. Such mperfectons drectly assocated wth the rural land rental markets lmt the extent to whch farm households are able to adjust the sze of ther cultvated area (ncomplete resource adjustment process) to even out mperfectons n other resource markets. (Hypothess 3) Moreover, access to land through the land rental market s constraned n communtes wth weak property rghts and poor nsttutonal tools to enforce contractual agreements (Sadoulet, 2001). Knshp networks are thus partcularly effectve n 15
provdng mutual nsurance and reduce market frctons (Sadoulet, et al., 1997). Knshp tes among farm households, therefore, trgger more partcpaton, and hence the levels of transacton, n the nformal land lease market. (Hypothess 4) In response to the supply constraned nature of the land lease market wth a very hgh land scarcty and populaton densty, the mperfectons n the tenancy market causes non-symmetry and favours the landlord sde of the market. (Hypothess 5) 4. Estmaton methods In addressng the major research questons, four dependent varables have been utlzed to dentfy the determnants of partcpaton n the nformal land rental market as well as farm households ntensty of land transactons. Two dummy dependent varables TENT and LLORD (for decsons to partcpant as a tenant and landlord, respectvely) and two log transformed contnuous varables lnlli and lnllo (for decsons how much to lease-n and how much to lease-out, respectvely) have been used. Probablty of partcpaton (probt models) As mentoned above, we model the decson to partcpate n the land lease market as bnary decson model (wth left-hand sde varables: 1 f lease-n or lease-out land and 0 otherwse) treatng two sdes of the market separately. Wth major problems of non-normalty of the error term and heteroscedastcty (non-constant varance of the error terms) concernng the standard OLS estmaton of such models, we utlze the maxmum lkelhood bnary response model probt model (Woldrdge, 2000; Gujarat, 2003;Green, 2000) In makng the decson regardng partcpaton n the nformal land lease market, we assume there s unobserved factor (utlty ndex) * R that explan why farm households lease n or out land. We postulate ths varable * R (latent varable) s a functon of vector of exogenous varables wth the relatonshp specfed as: Where * R = γ Z + ε (1) * R s a latent varable (unobserved) that shows the perceved net benefts from partcpaton n the nformal land lease market; Z s a vector of exogenous varables 16
(household demographc varables, asset and physcal endowment varables and other vllage level factors such as dstance to major market and populaton densty) that nfluence leasng behavour of households; γ s a vector of parameters; and ε s ndependent of Z whch represents factors unobserved by the researcher but known to the farmers (sample respondents). We further assume the random dsturbance term ε s normally dstrbuted wth zero mean and varance equal to one ( ε ~ N (0,1)). Instead of observng the latent varable * R (perceved net-beneft of market partcpaton), we only observe a dchotomous varable (ndcator varable) R, whch follows the sgn of the latent varable R. * When the perceved net-beneft of partcpaton (leasng n or out of land) s postve ( R > 0), the farm household partcpates n the nformal land lease market wth * R = 1. Therefore, for farm household, the observed bnary response can be gven as: R = 1 R = 0 f f * R = γ Z + ε > 0. (2) * R = γ Z + ε 0. (3) Thus the dstrbuton of P ( R = 1 Z ) = P ( R > 0 Z ) R gven Z can be obtaned as: * = P(γ Z + ε > 0 Z ) = P ( ε > -γ Z Z ) (4) Denotng θ 3 as the standard normal cumulatve dstrbuton functon (cdf), equaton (4) yelds: P( R = 1 Z ) = 1 θ(-γ Z ) P( R = 1 Z ) = θ(γ Z ) (5) Thus, the probt model takes the basc form of equaton (5) wth 0 P 1. Intensty of partcpaton (level of land transacton) 3 ' θ ( R ) = θ ( γ Z * ) = 1 2π ' γ Z e 2 k 2 dk, where k s the standard normal varable.e., k ~ N(0, δ ) 17
Havng the above probt model been estmated for decsons to transact or not (probt model estmaton), the test for selecton bas nvolve two procedures: computng the selectvty bas varable (Inverse Mll s Raton IMR); and estmaton of the ntensty of leasng (the amount of land leased-n or out) usng Ordnary least square (OLS) methods by ncorporatng ths new varable (IMR) to the ntensty (second stage) regresson model. To analyse the level (ntensty) of transacton of farm households, let the land transacton functon be gven by: Where A β X + U (6) R = ' ' β s a vector of parameters to be estmated, X s a vector of regressors (household demographc varables, asset and physcal endowment varables and other vllage level factors such as dstance to major market and populaton densty), U s a normally dstrbuted random dsturbance term. Wth the normalty assumptons for the error terms U ~ N (0, σ 2 ), ε ~ N(0,1) and corr (U, ε ) 0, estmatng equaton 6 usng OLS method would yeld nconsstent estmates because the expected value of the error term condtonal on the sample selecton crteron s non-zero (Madalla, 1983). Thus, condtonal upon R = 1 (postve values of * R ), the second stage Heckman selecton model (1979) provdes consstent and asymptotcally effcent estmates wth the regresson functon gven by: = β + ηλ(γ Z ) + U....(7) R ' A X Where: λ s the selectvty bas varable (Inverse Mll s Rato) computed as λ(γ Z ) = φ (γ Z ) / Φ (γ Z ), where φ and Φ are the densty and cumulatve standard normal dstrbuton functon, respectvely. As shown n Equaton (7), the IMR s then ncluded as an addtonal regressor n the OLS lease ntensty regresson where the regressors (varables) n the frst stage model (Z ) are allowed to dffer from the regressors n the second regresson equaton (X ) to meet the dentfcaton condton. However, to avod dentfcaton problems of pckng the rght varable that sgnfcantly explan the frst equaton (probt estmaton) but not the later, we use dfferent functonal specfcaton (double log functon) n the second stage equaton 18
whle lnear specfcaton was adopted for the probt model estmaton. Ths procedure was requred snce dentfcaton of the ndcator varable s necessary to apply the two-stage Henchman selecton model. (Tkabo and Holden: 2003) Data and Sample Area Descrpton To analyse the probablty and ntensty of land lease transactons, a survey data consstng of 16 vllages (a sample 25 farm households from each) has been collected from fve admnstratve zones of the Tgray regon of Ethopa that covers 11 woredas or dstrcts of the regon (see fgure 3 for detaled descrpton of the study area). The samplng method was admnstered at two levels: stratfed random samplng (takng agrcultural potental, populaton pressure, access to rrgaton, and access to market nto consderaton) at vllage level and a smple random samplng at a household level. The survey was conducted n May-July 2003 that represents the 2002/2003 crop season. Descrptve summary of the vllage level and household level varables are shown n table 1 and 2, respectvely 5. Results And Dscussons As mentoned n the prevous secton, four maxmum lkelhood probt model specfcatons have been utlzed to analyse the frst stage estmaton (the lkelhood of farm household partcpaton n the nformal tenancy market). As shown n table 4, Prob(LI)-2 and Prob(LO)-2 models are the lkelhood estmates of leasng-n and leasng-out decsons usng varous household level varables and selected vllage level attrbutes (lke vllage access to major market, to rrgaton project, etc) whle the remanng two (Prob(LI)-1 and Prob(LO)-1) are vllage fxed effect models - probt models wth the same household level varables but usng 15 ste dummes nstead of usng the vllage level attrbutes. To assess the ntensty of partcpaton, we utlzed these two model specfcatons Prob(LI)-2 and Prob(LO)-2 for havng a relatvely hgh predctve power over the remanng two model specfcatons. The results usng the Heckman selecton model (see table 6 and 8 for the demand and supply sde of the market, respectvely) show that the problem of selecton bas could not be rejected at 5% level of sgnfcance n the leasng-n decson whle there s no sgnfcant selecton bas problem n the landlord sde of the market. Accordngly, we estmated the analyss of ntensty of 19
leasng-out usng alternatve models of ordnary least square (OLS) (Tkabo and Holden, 2003) and censored Tobt model (Skoufas, 1995; Pender and Fafchampe, 2001; Dennger, 2003; and Tkabo and Holden, 2003). On the other hand, two-stage Heckman selecton model has been utlzed n analyzng households ntensty of leasng-n decsons snce t captures and corrects the Selecton bas problem (Tkabo and Holden, 2003). Lkelhood of partcpaton Supportng our basc hypothess (hypothess 1), the relatve endowment of non-land farm nputs wth respect to owned farm sze were found to sgnfcantly nfluence the household decsons to partcpate n the land lease market. As shown n table 4, oxen per unt of owned land (Oxp) and the tropcal lvestock unt per unt of owned land (Tluoxp) were found hghly sgnfcant (wth Oxp at 1% level) affectng the leasng behavour of households n all the models utlzed. Ths result supports our basc hypothess of the non/sem-tradablty of non-land farm nputs trggers farm household partcpaton n the tenancy market. Ths result may ndcate that there are hgher transacton costs I the oxen rental market than n the tenancy market wth respect to access to those markets. Ths fndng has be supported by the fact that 43% of the sample respondents were reported to have no ox and yet only 11.5 % of them engaged n oxen rental market. The hghly sgnfcant and negatve coeffcent of ownershp of other lvestock per unt of own land (tluoxp) n the supply sde of the market ndcates that more wealthy households were less lkely to rent out ther land. Households that have anmals may need ther land themselves to produce fodder for ther anmals and landlords tend to be the poorest households that lack the resources to farm ther land themselves. Consstent wth our expectatons, the coeffcents showng household relatve endowments of labour (Totmalp and Totfemp) were found to have strong effect on the leasng-out and leasng-n decsons, respectvely. Ths fndng appears to be consstent across smlar studes from Ethopa (Dennger, 2001; Teklu et al, 2004; Pender and Fafchamps, 2001) whch can be an ndcaton of an abundance of labour or relatvely better functonng labour market as compared to the oxen markets. 20
Though t plays a less mportant role n the leasng-out decdon, the gender coeffcent (Hhhs) was found to be consstent and hghly sgnfcant n both models utlzed to assess the lkelyhood of rentng-n land. Accordngly, female-headed households were less lkely to rent n land after controllng for other dfferences n endowments. Ths may be because landlords have less trust n female-headed households as land managers. Potental female-headed tenant households may therefore have been totally ratoned out of the market even f they had the necessary oxen endowment for farmng. As a proxy varable for capturng farm experence and physcal ablty, the age of the household head (Hhha) was found to a strong sgnfcant effect on the land rentng out decson showng that the older the age of the household head, the hgher the probablty of rentng-out land. Prevous partcpaton n the market for tenances had a strong postve effect on current probablty of partcpaton n both sdes of the market, attestng our pror expectaton that households well acquanted wth the exstng tenancy market (havng nsders market nformaton) are more lkely to partcpate n the market than otherwse. Ths could be due to the hgh ntal cost (fxed transacton cost) an agent may face when frst attemptng to enter the market. All varables that capture vllage level attrbutes appeared to play less mportant n relaton to access to tenancy markets. Intensty of market partcpaton Intensty of leasng-out We wll now look at the determnants of the ntensty of partcpaton on the landlord sde of the market. As mentoned before, we found no sgns of selecton bas related to partcpaton on ths sde of the market (see table 7). We therefore dscuss the results from an ordnary least square and a truncated tobt model presented n Table 5. We used a vllage level knshp varable (vlkn), whch was constructed as the share of all contracts n the vllage that were all among kn partners. As the result shows, ownershp status of farm nputs (all labour and lvestock endowment varables), though havng a sgnfcant effect on the probablty of partcpaton, both the OLS and truncated Tobt regresson results shows a weak effect of such factors on the amount of land rented out. The facts that 50% of the landlords 21
leased out all the land they owned (zero operatonal holdng) and 83% of them wth zero oxen could be a possble explanaton for the absence of partal factor rato adjustment wth respect to oxen. Moreover, the fragmented land holdng of farm households (an average number of plots and farm sze of :::: and :::, respectvely), mght also explan the weak effect of these asset endowment varables on the ntensty decsons. We see that land leased out dd not adjust smoothly wth the farm sze of landlords as the coeffcent on farm sze (fs) s only 0.136 and 0.151 n the OLS and truncated Tobt models. Further attempt has been made to test f ths s sgnfcantly less than 1.0, whch was expected wth a perfect land tenancy market wth zero transacton costs and constant returns to scale. The result rejects the hypotheses that fs=1 that explans mperfect resource adjustment n the landlord sde of the tenancy market. In contrast wth our 4 th hypothess, knshp networks were found to have a negatve mpact n dctatng the leasng behavour of households. We hypothessed that trust may be hgher among kn and therefore adjustment would be smoother n kn contracts. Ths could mply that the land rental market functons better n vllages wth a hgh share of knshp contracts. The hghly sgnfcant but negatve coeffcent of vlkn varable, on both the OLS and Tobt regresson results, however, shows the land rental market functons sgnfcantly better n vllages wth low share of knshp contracts. A large share of knshp contracts could be a sgn of a low-trust communty (Vlkn to have postve effect) or only due to a hghly supply constraned (a very small land holdng) tenancy market where potental tenants solely rely on kn related landlords to get any access to arable land. Thus, the counter-ntutve result could be explaned f the second case domnates the frst one where socal captal plays less role than the mnature land holdng of farm households.??? T hs fndng contrasts wth the theoretcal foundaton that socal captal (mutual trust and mnmzed moral hazard among kn ted partners) matters n wnng the hearts of transactng agents - landlords (Sadoulet, 2001). More unequal land dstrbuton (vgn) stmulated sgnfcantly rentng out of land. Ths contrasts wth the partcpaton models. The probablty of land beng rented out was lower but those rentng out land rented out more land n vllages wth more 22
skewed land dstrbuton. Ths may ndcate that those landlords who trusted the market were wllng to rent out more land where the demand was hgher due to more n egaltaran dstrbuton. Another nterestng fndng n Table5 was that households wth off-farm ncome rent out more land. Ths may ndcate that there were mperfectons n the labour market snce extra ( hgh-pay ) off-farm ncome was not used to hre ( low-pay ) labour to substtute for lost household labour n farmng. There are addtonal transacton costs related to hrng and montorng hred labour and such costs may even be hgher for households engaged n off-farm work than for other households. Intensty of leasng-n We used a maxmum lkelhood Heckman (1979) selecton model for the analyss of tentants degree (ntensty) of partcpaton n the land rental market, and we found sgnfcant selecton bas. The results of the second stage of the model are presented n Table 6. We also run a Deaton regresson model for testng the robustness of results across these two models. The results of the Deaton regresson model are also presented n Table 6. Unlke ther profound effect on the lkelhood of partcpaton, both the Heckman and Deaton regresson results show oxen endowment to have a weak effect n dctatng the ntensty of leasng-n land. Ths could be a strengthenng force for our hypothess for havng a supply-constraned tenancy market n the study area. The fact that rental markets for such farm nputs (oxen) are hghly mperfect and yet the statstcal nsgnfcance of such parameters could lead to a strong argument that exogenous factors (possbly, screenng crtera of landlords) rather than the relatve endowment of farm nputs explan the level of transacton (degree of access to leased land). Lvestock wealth other than oxen (lntluox), however, was found to have a sgnfcant and postve effect on the amount of land to rent n. Households wth more endowment of other lvestocks tend to rent n more land possbly to provde fodder for ther lvestock. Wth both the endowments of famly male and female labour havng a weak effect on the lkelhood of partcpaton (frst-hand decson), the endowment of adult female 23
labour n a famly, however, sgnfcantly reduces the amount of land rented n. Ths result confrms the gender-specfc nature of an agrcultural task where the socal constrant on female labour market partcpaton s very hgh. Smlar to the case of the landlord sde of the market, the sgnfcant but negatve coeffcent of vlkn varable usng the Heckman model contrasts to our hypothess that tenants wth kn related landlords enjoy relatve ease of barrers n the process of factor rato adjustment through the nformal lease market. Ths counter-ntutve result could be the outcome of a possble nteracton between two contradctng factors (the share of knshp contracts n a vllage and the land holdng status of households n a vllage). Contrary to the predomnantly customary explanatons, we found a sgnfcant postve coeffcent on the owned farm sze varable. Surprsngly, the result s puzzlng as households wth a relatvely hgh allocated farm sze are expected to have less demand for rented plots and to lease-n less land. Spurous correlatons between plot sze, farm sze and area szes leased-n/out could possbly cause the stuaton. Locatons wth large farm szes could typcally be areas wth larger szes of land lease-n or out. In an effort to correct the possble spurous correlaton, an attempt has been made to estmate the lease-n ntensty regresson usng a relatve farm szes wth n vllages (ndvdual land holdng / total vllage farm sze). The result, however, remans robust across these alternatve models wth the farm sze varable stll havng a postve sgnfcant effect on the amount of land leased-n. There are good reasons, however, to treat the result cautously snce no further effort (test) has been exerted to address the ssue of scale of producton (effcency analyss) due to the drought nature of the data collecton perdor. Wth some degree of reservaton, the unexpected postve sgn could possbly mply that large farmers endowed wth large farm szes possess abundant complementary farm assets (labour, oxen and other farm mplements) and yet ther relatve factor rato (land to non-land) endowment reman to be low whch may demand further adjustment actons by leasng-n more land. Ths result, thus, could justfy our argument that relatve rather than absolute farm resource endowment matters n dctatng the leasng behavour of households. 24
As ndcated n table 6, tenants who resde n vllages wth rrgaton facltes (rrg), more skewed land dstrbuton (vgn), and less access to a major market (Mktd) were found to rent n more land than otherwse. Due to low opportunty cost of farm labour n remote areas (no or less off-farm engagements), landlords may opt for own cultvaton of ther plots leavng few amount of land to be tenant operated. Tenancy market frcton and resource adjustment process Our expectaton of ncomplete factor rato adjustment through the land lease market (Hypothess 3) was not rejected for the coeffcent of the allocated farm sze n the two ntensty models (OLS for leasng out and Heckman for leasng n decsons) were found to be sgnfcantly dfferent from 1 n both sdes of the tenancy market. The result s more pronounced as 56% of the actual tenants were reported to be constraned tenants who all want to lease n more land (see table 9). Ths shows, lke the mperfectons n the non-land factor markets, frctons n the tenancy market cause partcpant households to transact less than ther optmal level. Fragmented land holdng and hghly skewed dstrbuton of non-land farm nputs accompaned by the supply constraned nature of tenancy market, mght be possble explanatons behnd the observed fact n the demand sde of the market. Though t was only 5% of the landlord households that were dentfed as constraned landlords who want to lease out more land, the nformal nature of the tenancy market 4, the past polcy falure whch dscourages leasng out practces and other cultural and tradtonal factors 5 could explan the frctons n ths supply sde of the market. Market symmetry and ratonng out problem In an attempt to test the hypothess of asymmetrc (non-symmetrcal) nature of the nformal land lease market, a market symmetry test 6 has been made applyng the censored Tobt model that analyse the determnants of net land leased-n as adopted by Skoufas (1995). Usng the tobt regresson output (see table 10) from both the 4 Almost all of the land lease market partcpants have no formal contractual agreements whle transactng n the tenancy market. 5 Informal dscussons wth some of the elderly farmers show that farm households consder leasng out the entre land one owns as socally undesrable act. 6 Testng the equalty of two sdes of the market both the sgns and sgnfcance level of varables. 25