Land Markets in Transition: Theory and Evidence from Hungary. Liesbet Vranken and Johan F.M. Swinnen

Transcription

1 Land Markets n Transton: Theory and Evdence from Hungary Lesbet Vranken and Johan F.M. Swnnen Research Group on Food Polcy, Transton, and Development Katholeke Unverstet Leuven Workng Paper 2003/1* Abstract Ths paper analyses the determnants of household farms partcpaton n land markets n transton countres. We derve several hypotheses on the mpact of households management ablty, land endowment, land qualty and prces, transacton costs n the land market, rural credt and labour market constrants. We test the hypotheses combnng a representatve dataset on land rental actvtes of more than 1,400 Hungaran household farms wth data from the Hungaran Central Statstcal Offce. We fnd that land rental markets reallocate land to households wth better farm management capactes and less endowed wth land. Households combne buyng and rentng of land to extend ther farms. The contnued domnaton of large farm organzatons n some regons restrcts household s access to land. Rural credt and labour market mperfectons have an mportant mpact on land rental markets. JEL Classfcaton Q12, Q15, D10, P23 Ths research was supported by the Research Programme of the Fund for Scentfc Research Flanders, Belgum (F.W.O. G ). * The paper s also publshed as LICOS Dscusson paper 129/2003.

2 Land Markets n Transton: Theory and Evdence from Hungary 1 Introducton Land reform and the creaton of optmal land nsttutons has attracted renewed attenton because of ts mportance n transton processes such as n Chna, Vetnam, South Afrca, the former Sovet Unon, and Eastern Europe and because of new poltcal pressure for land reforms n countres wth hghly unequal land dstrbutons such as Zmbabwe and Brazl. New nsghts n the functonng of land markets and nsttutons have also nduced renewed attenton to land access as a poverty reducng tool (de Janvry et al., 2001). Much attenton has been pad to land sales markets or, more generally, the transfer of ownershp as an mportant nstrument to enhance effcency, and reduce poverty. Land ownershp transfers come wth a number of benefts, such as the potental to use land as collateral. However, n an envronment wth large uncertantes and hgh transacton costs, where credt markets and nsurance markets are mperfect, land sales markets are typcally thn and land sales may be lmted to dstress sales (Platteau, 2000). In such crcumstances, land rental markets can play an mportant role n mprovng effcency and possbly equty n land use and access (Sadoulet et al, 2001). As such, the role of land rental markets has recently been re-emphaszed as mportant for provdng access to land for the poor and as an effcency-enhancng nsttuton n envronments characterzed by large uncertantes, such as countres n transton (Dennger and Bnswanger, 2001; Swnnen, 2001). Transton countres provde a unque opportunty to study the development of land markets as land reforms have reallocated property rghts and lberalzed land exchange restrctons. Whle much has been wrtten on land reforms and farm restructurng n 2

3 transton countres (eg. Lerman et al. 2002; Swnnen et al. 1997), few studes have provded a formal conceptual and econometrc analyss of the development of land markets and ther determnants. The few studes have focused on Chna and Vetnam where transton started earler (e.g. Brandt et al, 2003; Dennger and Songqng, 2003). Ths paper s the frst to provde a formal analyss of the role of households n land market developments usng data from Central and Eastern Europe or the former Sovet Unon. The focus on households n analysng the development of markets n transton countres s mportant because n many transton countres household farms are usng a large part of the agrcultural land, although there s large varaton. The share of agrcultural land used by household farms vares from less than 20% n countres such as Belarus, Slovaka, Russa, and Ukrane to more than 80 % n Albana, Armena, Romana, and Poland (Lerman, 2001). But even n countres such as Russa where household farms use less than 20% of the land, they produce 60% of total output. Furthermore, the land used by large scale farms s often rented from households. The focus on households s also mportant to study the equty effects of the land market developments. An mportant queston s whether the land reforms and lberalzed land rental and sales markets wll contrbute to growng effcency n agrculture and to mproved access to land for small farms and poor rural households n transton countres. There s concern that land market lberalzaton wll lead to a re-concentraton of land. Whle the evdence on ths effect s mxed and mostly lmted to Latn Amerca, a contnent characterzed by hgh nequalty n access to land (see Dennger and Songqng (2003) for a revew), Lerman et al (2002) pont out that n an envronment characterzed by asymmetrc access to nformaton, captal, and legal means of enforcement, that s typcal of transton economes, reconcentraton may be a realstc outcome, wth undesrable socal and economc consequences. 3

4 The paper frst develops a theoretcal model to analyse the decson-makng of farmng households to partcpate n the land market, whch ncorporates transton characterstcs of land ownershp, land use, and rural market mperfectons. We derve a set of hypotheses on the land market partcpaton of rural households. The theoretcal model dffers from other models n the lterature n two ways. Frst, the tradtonal lterature on rental markets typcally focuses on sharecroppng or on the relatonshp between large landlords and small tenants (Bravermman and Stgltz, 1982; Bardhan, 1989). Whle these assumptons are relevant for parts of the developng world, they do not capture essental characterstcs of land rental markets n transton countres, whch are characterzed by dspersed landownershp among many rural and urban households and where small farms compete for access to land wth large-scale corporate farms (Lerman et al., 2002; Mathjs and Swnnen, 1998). One objectve of our study s to analyse how these specfc characterstcs affect the land market and small farmers access to land. Second, several studes have analysed the mpact of specfc factor market mperfectons on the development of land markets. For example, Carter and Salgado (2001) emphasze the mpact of credt constrants, Yao (2000) labour market mperfectons, and Skoufas (1995) the effect of land transacton costs. However, transton agrculture s characterzed by most, f not all, these mperfectons and transacton costs. Therefore, our model ncorporates multple factor market mperfectons. The second part of the paper s emprcal and uses a unque and representatve dataset on land rental actvtes of more than 1,400 Hungaran rural households. These data are combned wth county-level data collected by the Hungaran Central Statstcal Offce to estmate the determnants of household rental actvtes durng transton. We selected Hungary for the emprcal analyss for several reasons. Frst, Hungary has an nterestng mxture of household farms, farmng companes and cooperatve farms: all these farmng 4

5 organzatons use a sgnfcant share of the land, wth household farms usng slghtly more than 50%. Moreover, there are mportant regonal varatons n ther relatve mportance, whch allows to test for the mpact of land market domnaton of large farms on household farms access to land. Second, Hungary s, certanly n comparson wth other transton countres, well advanced n ts land reform process. Land ttles have been largely dstrbuted. By studyng land market developments and household access to land through land rental and sales markets n ths advanced transton stage we can analyze whether everythng wll be alrght when the land reform s fnshed. In other words, s t suffcent for polcy-makers n other countres to focus ther attenton on mplementng the land reform and ttlng process n order to get the land market gong, or are complementary reforms and polces needed? A related ssue whch can be addressed n Hungary, n contrast to many other transton countres where progress s less advanced, s the relatonshp between land sales and the land rental market. Whle restrctons on land sales stll exst, a sgnfcant amount of land sales occurred n Hungary n the years precedng the survey. The survey ncludes evdence on household land purchases n the past years and current land rental actvtes. By ncorporatng both sets of nformaton, we can derve mportant conclusons on the relatonshp between both. The emprcal part of the paper starts wth a dscusson of the data and general characterstcs of land use and ownershp n rural Hungary. Next, we present profles of households who rent n land and of those who rent out land and we provde evdence how the behavour of large farm enterprses affects small farmers access to land through rental. Afterwards, we estmate econometrcally the mpact of household and farmng characterstcs, such as physcal and human captal, as well as land market and regonal characterstcs on land rental actvtes. 5

6 2 Theory Consder a household wth endowments of land T, labour L and ntal wealth M. The household can derve ncome from agrcultural producton on ts own farm and from offfarm wage employment. Agrcultural output s produced accordng to the followng ncreasng, strctly quas-concave and twce contnuously dfferentable producton functon: Q = f(t, L, X, Z ) (1) where T s the land used by the farm; L s the effectve labour nput on the farm; X s the amount of purchased nputs wth prce p x ; and Z s the amount of non-tradable nputs and fxed productve assets, lke manageral or techncal sklls that are not rewarded by the labour market. We normalze the agrcultural output prce to one. The land used, T, can be larger or smaller than land owned by the household, T. If the household has more land than t wants to use t can rent t out or sell t. Addtonal land can be acqured through rentng n or buyng. Both the rental and sales markets are characterzed by mperfectons n transton countres. However, rental markets have developed much earler and much more wdely than land markets. Therefore, and to keep the analyss smpler, we wll ntally assume that land sales are not possble, and that the household can only rent land n or out. Later on we wll dscuss how the results are affected f buyng and sellng of land s possble. Hence, we defne the land used as T o = T + T T wth T the land ntally owned by the household, T the amount of land rented n and T o the amount of land rented out. The rental prce for land rented n (r ) may dffer from that of land rented out (r o ) due to mperfectons n the land rental market. Transacton costs, such as search costs and costs related to negotatng the terms of the tenure contract, result n the prce for land rented n (r ) 6

7 to be hgher than the prce for land rented out (r o o ) ceters parbus, and r - r > 0 s an ndcator of the sze of the transacton costs. Credt market mperfectons are mportant n rural areas of transton countres. Wth credt market mperfectons, a farmer may not borrow aganst future profts. In the development economcs lterature (e.g. Eswaran and Kotwall, 1986; Dasgupta, 1993) one typcally assumes that access to loans depends on the amount of owned land. However, n transton economes, especally durng the frst decade of transton, fnancal nsttutons often refused land as collateral or because of mperfectly defned property rghts, or because of thn land markets, or because of socal pressure preventng them from takng over land n case of default (Swnnen and Gow, 1999). We therefore assume that access to loans B depends on the amount of owned land and productve assets and wth B small and dependng on the country and phase of transton. In addton, we assume that households have some own lqudty M. Labour market mperfectons are due to off-farm employment constrants and moral hazard problems wth hred labour. Moral hazard wth hred labour requres supervson of workers. The effectve labour suppled by hred workers therefore depends on the amount of famly labour workng on the farm, whch s assumed to combne effectve nput and supervson, as well as on the area of land cultvated (Carter and Salgado, 2001; Feder, 1985). The effectve labour nput L s therefore: L f = L f + s(t, L ).L (2) where L s the nomnal amount of hred labour and L f s the famly labour devoted to the f farm. The supervson functon s(t, L ), wth 0 s (.) 1 reflects how nomnal labour nput s transformed nto labour effort. The effcency of supervson s a postve, but concave f 2 f functon of famly labour nput ( s / L 0, s / L 0 ), and dmnshes as the farm sze 2 2 grows and, for a gven level of famly labour nput: s / T 0 and s / T

8 We assume the labour market s cleared by quantty ratonng n order to fll the wage gap. Several theores explan why frms use quantty ratonng nstead of prce ratonng to clear the labour market (for example, the effcency wage theory proposed by Akerlof and Yellen, 1986; the moral hazard model proposed by Shapro and Stgltz, 1984; the tme ratonng model by Yao, 2000). Here, the assumpton of ratoned off-farm employment opportuntes s ncorporated n the model by allowng that the wage pad to hred labourers o ( w ) dffers from the wage that household members can gan off farm ( w ) and by settng an o o upper lmt ( L ) to the amount of labour employed off farm ( L ). Household utlty s an ncreasng functon of ncome (y) and lesure (l). Incorporatng all the characterstcs dscussed above, the household maxmsaton problem s then: max U ( y, l) (3) f o L, L, L, l o T, T, X wth y + o o o o = f(l,t,x,z) -px X-r T -w L + r T w L (4) s.t. p x X o o o o + w L + r T B + M + w L + r T (5) o o L L (6) L f o = L + L + l (7) Inequalty (5) reflects the lqudty constrant to whch the household s subjected. Constrants on off-farm employment opportuntes are ncorporated n (6), whle nequalty (7) captures the tme constrant of the household members. The frst order condtons for the amount of land rented n ( T ) and the amount of land o rented out ( T ), and amount of famly (L f ) and hred (L ) labour devoted to the farm δ T : ft (1 + ) r f LsT. U y L (8) δ o o T : ft (1 + ) L T. U y r f s L (9) 8

9 L f o δ µ : f L.(1 + s f. L ) w (1 + ) + 0 L U U y y (10) f δ L : f L. s( T, L ) w (1 + ) 0 U y (11) where subscrpts refer to frst dervatves and δ and µ are the Lagrange multplers for respectvely the lqudty constrant and off-farm employment constrant. The frst terms on the rght hand sde of equatons (8) and (9) capture the opportunty cost of land n the presence of credt constrants. The second term (whch s negatve wth s T < 0 ) reflects extra supervson costs wth growng farm sze, f hred labour s workng on the farm (L >0). Wth only famly labour employed (L = 0), ths term s zero. If there were no transacton costs n the rental market ( r = r o ), the demand for own land and rented land would be dentcal. Wth transacton costs n the rental market ( r>r o ) the household wll frst use ts own land 1. Combnng ths wth equatons (8) and (9) mples that a household rents n land f δ f T > ( 1+ ) r flst. U y L, (12) does not rent land f δ o ( 1+ ) r f LsT. L ft (1 + ) r U y U y δ f s. L, (13) L T rents out land f δ o f T < ( 1+ ) r flst. U y L. (14) We can use condtons (12)-(14) to derve several hypotheses on whch factors affect the partcpaton of rural households n land rental markets (all n ceters parbus terms). 1. A household s more lkely to rent n land (and less lkely to rent out) f the margnal product of land (f T ) s hgher. The margnal product of land s affected both by the 1 Ths also mples that there s no smultaneous rentng n and out. Ths result s based on the assumpton that there are no other dfferences between owned land and land rented, for example, qualty and locaton. If land plots have dfferent characterstcs, one may observe smultaneous rentng n and out of land by the same household. 9

10 ntrnsc qualty of the land and by the sklls of the household n managng the land and farmng t. 2. The land endowment of the household wll affect the decson to rent. Gven some fxed nputs, and market mperfectons whch constran extendng some other nputs, the margnal productvty of land wll decrease wth land use. If the margnal productvty of the land at the level of land owned by the household s stll larger than the margnal costs of rentng n addtonal land (.e. nequalty (12) holds) then the household wll rent n addtonal land. Ths wll depend on the amount of land owned by the household. The more land the household owns, ceters parbus, the less t s lkely to rent n and the more t s lkely to rent out. 3. The household s more lkely to rent n land f the land rental prce s lower, and vce versa for rentng out. Notce that, wth gven transacton costs, changes n the market rental prce wll affect both decsons to the same extent; or, n other words, wll equally affect r and r. 4. Transacton costs n the rental market wll cause a gap between r and r, and consequently wll reduce both rentng n and rentng out. Wth o r r larger the autarky nterval n equaton (11) ncreases. Such transacton costs can come from a varety of sources, such as search costs. In transton countres an mportant cause of the gap between r and r may also be obstructons or mperfect competton n the land market by large farm organzatons. The latter may complcate access to land for small farms and use ther scale advantages n admnstraton as well as n negotatng wth small and dspersed land owners to ncrease the land rental prce for small farms competng for land and decrease the rental prce for households rentng out. 5. Imperfectons n the credt market also affect land rental markets. Credt market constrants are reflected n equatons (12-14) n the value of δ, the shadow prce of the 10

11 lqudty constrant (5) n the household optmsaton problem. More credt market constrants mply a hgher value of δ and ths wll reduce the lkelhood that a household wll rent n land. It makes t more lkely that t rents out land. There s a secondary effect that renforces ths. Credt market constrants wll also reduce labour use on the farm. Ths can be seen from equatons (10-11) where an ncrease n δ wll result n less farm labour use. Ths wll, n turn, cause a declne n the margnal productvty of land, f T, and consequently, further reduce rentng n of land and ncrease rentng out of land. o 6. Constrants on off-farm employment ( L ) wll also affect land rental decsons. Such constrants are reflected n equatons (12-14) n the value of the µ, the shadow prce of the constrant (6) n the household optmsaton problem. If off-farm labour opportuntes are scarcer (.e. f µ s hgher), more famly labour wll be used on the farm (equaton (10)). Ths wll ncrease labour nput L and therefore rase the margnal product of land f T. It follows from equatons (12-14) that, through the ncreased margnal productvty of land, scarcer off-farm labour opportuntes wll nduce a farmng household to rent n more land (or rent out less land). The sze of ths effect depends on whether the household s usng only household labour or whether t s hrng labour (n addton to ts own household labour). The effect on land rentng wll be smaller when hred workers are employed on the farm because the supervson cost of montorng hred labour weakens the effect. Ths can be seen from the frst order condtons (10) and (11). 7. For the same reason, the household labour supply (L f ) wll affect the land rental decsons. Wth supervson costs makng hred labour more expensve than household labour, the household labour supply wll postvely affect the decson to rent n land, and rentng out of land. 11

12 8. Hgher wages, ether for off-farm employment (w ) or for hrng farm labour (w ), or both, reduces rentng n of land and ncreases rentng out of land as employng labour on the farm becomes more expensve ether n terms of actual wages or n terms of opportunty costs whch reduces the margnal productvty of land. 3 Land sales versus rental contracts So far we have assumed that buyng or sellng land was not possble. Ths s the case n several transton countres, and n most transton countres for at least some perod. For example, agrcultural land sales were forbdden durng the 1990s n Russa and most of the CIS countres. Hence, the hypotheses so far provde a suffcent theoretcal framework for analysng rental markets n several transton countres. However, n other countres sgnfcant sales of agrcultural land occurred n the past years. Moreover, land sales are lkely to become a more mportant form of land exchange n the future. Therefore t s mportant to consder how land rental actvtes are lkely to be affected when land sales are possble. Let us consder the case when a household wants to acqure more land for farmng. There are several reasons why a household may prefer buyng land over rentng n land, or vce versa. Factors that affect the trade-off between buyng versus rentng land nclude securty of operaton and nvestment returns, credt constrants, uncertanty regardng property rghts, prce and ncome rsks, and psychologcal and cultural values assocated wth land ownershp. In dong so, t s mportant to dstngush between early transton, characterzed by major economc and nsttutonal reforms and uncertantes, and nsecure property rghts, and the second phase of transton when some of the basc reforms have been mplemented, the economc stuaton has stablzed and property rghts are more secure 12

13 Key characterstcs of early transton, such as mperfectly defned property rghts and major prce and ncome uncertanty, are a major constrant on land sales. Unclear property rghts were promnent n all transton countres n early transton and contnue to be a major problem n several countres. They are obvously a major constrant on land sales. Economc and nsttutonal uncertantes also constran land sales. Both the demand and supply of land wll be constraned when households are uncertan about the future ncomes that land use wll yeld. Ths stuaton characterzed early transton n many countres: prces for agrcultural commodtes and nputs changed dramatcally and unexpectedly n the early 1990s. The result was few, f any, land sales and land exchange restrcted to rental agreements. In ths paper we do not analyse the type of rental agreements (long versus short-term, tenant rghts, etc.) However notce that ll-defned property rghts and major uncertantes wll also have an mportant mpact on the rental market. Rental agreements may be restrcted to short term and nformal agreements. Such agreements do not provde the necessary securty of operaton and guarantee for nvestment returns to tenants, whch they requre for makng optmal producton and land allocaton decsons. Moreover, n some cases rental agreements themselves may be constraned. For example, Macours (2001) shows how property rghts nsecurtes affect the choce of partners n rental agreements n Latn Amerca. Consder now a second phase of transton stuaton. In other words, we assume that basc reforms have been mplemented such that land rghts are suffcently well-defned for land sales to take place and that prces of nputs and outputs have become much more stable. Yet mportant transacton costs and mperfectons reman n land, credt and labour markets as captured by the theoretcal model n the prevous secton. Let us consder agan the case when a household wants to acqure more land for farmng. The key factors n the households decson are now the trade-off between securty of operaton and nvestment and credt 13

14 constrants. Buyng land (compared to rentng) ensures the household that t can capture the benefts of ts nvestment n the land; that t s certan to have suffcent land at hs dsposal for future cultvaton; and guarantees the locaton and qualty of ts land. Further, t allows better producton decsons as mult-year producton cycles (e.g. perennal crops) can be ncluded n ts producton plans. Other benefts are that land can be used as collateral for future nvestments and as an asset n the household s nvestment portfolo. Moreover, land ownershp may play an mportant role as hedge aganst nflaton for the household, and, n the absence of nsurance markets, as a bass for employment and food securty. Fnally, t may brng socal status and poltcal nfluence (Dennger and Feder, 2002; Platteau, 2000). The man advantage of rentng land over buyng s that t requres less lqudty or access to credt. 2 Wth credt market mperfectons, ths s a very mportant consderaton n the household s choce. Credt constrants wll reduce the demand for land by the household, as shown n the prevous secton, but wll also make t more lkely that addtonal land wll be rented nstead of bought by the household. Ths trade-off between securty of operaton and lqudty for the farmng household s not only mportant n transton countres. It also affects the decson between rentng and buyng of land n most western farms. For ths reason, farms often combne owned and rented land. A mnmum amount of owned land ensures securty of operatons whle extendng the farm by rented land prevents them from nvestng all ther captal n land and to use t for workng captal or other nvestment purposes (Sommer et al., 1995; Swnnen, 2002). 3 2 Rentng land may also be preferred when ncreases n the household s land demand s temporary, for example due to temporary fluctuatons n some of the other nputs. 3 Here we gnore ssues of securty and regulaton of rental contracts. In some countres n Western Europe, land tenure contract regulatons provde very extensve securty to the tenant. Ths shfts the preferences of farms to rental contracts as t ncreases the securty benefts wthout ncreasng ther credt requrements. For example, n Belgum extensve rental regulatons resulted n 70% of land use under rental contracts; n the Netherlands landowners n the 1980s refused further rental contracts wth farmers when regulatons mposed too strongly on ther property rghts (Swnnen, 2002). 14

15 4 Data The data used n the emprcal analyss are household level data collected n a 1998 rural household survey n Hungary and county-level data from the Hungaran Statstcal offce. The survey s a representatve country-wde survey of rural households wth some farmng actvtes. The dataset ncludes data on more than 1400 households. Household ownershp of land and household farmng has grown strongly snce the begnnng of transton (fgure 1). Under the communst regme only 10% of agrcultural land was used by households, mostly as garden plots. Around 66% of land was used by collectve farms, the rest by state farms. One-thrd of the land used by collectve farms was formally owned by ndvdual members of the collectve farms, but they had very lttle effectve rghts (Mathjs and Mészáros, 1997). These rghts were restored durng the land reform n the early 1990s. 4 In addton, the land reform process compensated former landowners, who had lost ther land n the collectvsaton process, through vouchers whch could be used for purchasng land n the prvatsaton process. 5 About 2.5 mllon hectares of collectve land and 0.2 mllon hectares of state owned land were prvatsed through voucher-based auctons. The remanng land from the collectve farms was allocated to ther members (European Commsson, 1998). 6 Legal restrctons constraned land ownershp and sales. Land receved through compensaton or as a share from the collectve farms could not be sold for three years after recept. There s an upper lmt of land ownershp of 300 hectares for ndvdual 4 There were major mplementaton problems snce the land had been consoldated and been subject to land mprovement actvtes under collectve farm management. 5 People elgble for compensaton were farmers whose land was sezed just after Second World War and farmers who were forced to sell ther land to the collectve farm for a low prce n the 1970s and 1980s. 6 The land cultvated by state farms was not subject to prvatsaton. The State Property Agency allocated the land as follows: 40 % was used for compensaton of prvate persons, 37 % s used by companes whch reman state property, 27 % was leased (manly to former state farms) and 6 % was allocated to employees of state farms (Mathjs and Mészáros, 1997). 15

16 ownershp and legal persons and non-resdent foregn ctzens cannot own agrcultural land n Hungary. By 1998, households owned 84% of all agrcultural land n Hungary, and used around 51% n household farms (or ndvdual or famly farms). The rest of the land s used by large scale cooperatve farms and farmng corporatons, who each use around a quarter of Hungaran land. Household farms are small on average and use mostly ther own land: on average they cultvate 5 hectares and also own 5 hectares (see table 1). They provde only a small part of total household ncome: on average less than 20% of household ncome comes from farmng (see table 2). Many of the farms are run by older (55 years on average) and low educated (9 years schoolng on average) heads of households. Income from pensons makes up around 40% of total household ncome. Three quarters (76%) of the households n the sample do not partcpate n the rental market. Sxteen percent of households rent out land, whle around 8 percent of the households rent n land. Land s rented out to other households and to collectve farms and farmng companes. The average amount of land rented n s 15 hectares, and that of land rented out s 5 hectares. There are mportant dfferences between households whch rent n land and those who do not partcpate n the rental market or rent out land (see table 2). On average, the heads of households rentng n land are sgnfcantly younger and slghtly better educated. The households cultvate much more land, and also own more land and machnery. More households n ths group have access to machnery servces and credt. Around 40% of ther household ncome comes from ther farmng actvtes on average, compared to less than 16% n the other categores, and pensons accounts for around 20% of household ncome, sgnfcantly less than n other groups. 16

17 These average numbers already suggest mportant conclusons. In the next secton we use an econometrc model to formally test whch characterstcs are mportant determnants of household partcpaton n the rental market, and to see to what extent external factors, such as regonal varatons n land qualty and n competton n the land market affect land rental actvtes. 5 The Emprcal Model The emprcal estmaton ncludes two models. One model uses the amount of land rented out as dependent varable, the other model uses the amount of land rented n as dependent varable. Both emprcal models have the followng structure: y = α 0 + x β + l γ + r δ + ε (15) where y represent the dependent varable, x s a vector of varables measurng household characterstcs, l a vector of county-level ndcator varables of land market characterstcs, r a matrx of regonal dummy varables to capture fxed effects not captured by the other explanatory varables, and ε refers to the error term. β, γ and δ are vectors of parameters related to respectvely the household characterstcs, the county level ndcators of land market characterstcs and to regonal varables. We use two ndcators for the land endowment of the household. LANDOWNED s the amount of land owned by the household when the survey was mplemented n We expect ths varable to be negatvely related wth the amount of land rented n, and postvely wth the amount of land rented out. Some of the households purchased part of ths land durng the prevous years. To test whether there s a dfference n whether the land was purchased n the past few years or whether the land was owned by the household before transton or gven to them n the land reform process, we splt up the land owned by the 17

18 household n ts ntal land endowment and land purchased by the household over the perod (LANDBOUGHT). As explaned n secton 3, the household faces a securtylqudty trade-off n the decson whether to purchase or to rent n land. Therefore we expect a postve relatonshp between LANDBOUGHT and the amount of land rented n. We estmate a model ncludng LANDOWNED and LANDBOUGHT as well as the squared form to capture non-lnear effects. There s a possble endogenety problem wth the LANDBOUGHT varable snce the decson to rent n land and buy land may be jontly decded. Therefore we use the Amemya Generalzed Least Squares (AGLS) estmator for Tobt wth endogenous regressors. The endogenous regressor, LANDBOUGHT, s treated as a lnear functon of the nstruments and the other exogenous varables. The second stage estmates are consstent despte the fact that the endogeneous varable that we are nstrumentng s censored (Madala, 1983). As an nstrument we use the amount of land cultvated by the households just after start of the reforms. Ths varable s exogenous snce households dd not have any decson power over the land they owned durng communsm. The land reform process resttuted land to former (pre-1947) owners so that the amount of land at the household s dsposal for cultvaton n 1991 s exogenous to the decson to partcpate n the land rental market n To capture transacton costs n the land rental market, we nclude three varables. Frst, DOMFCO reflects the extent of domnaton of the land market by farmng companes and cooperatves. Table 1 shows how n regons where only a very small share of the households (less than 5%) are rentng n land farmng corporatons and cooperatves stll cultvate on average 79% of the agrcultural land. Ths s consderably larger than n regons where the percentage of households who are rentng n land s larger than 10% (53%). Moreover, not only less households are rentng n land, they rent n much smaller amounts of land (1 hectare versus 17 hectares). Further, not only are they rentng less, they are usng less fertle land. 18

19 Comparng results from our survey wth land qualty ndcators of the Hungaran statstcal offce ndcates that n regons wth domnaton of large cooperatves and companes, land used by households s of sgnfcant lower qualty than the average land qualty of the county (17% less on average), whle n other regons we fnd no dfference between the average qualty of the land used by households and that of the county as a whole. All ths suggests that households face mportant transacton costs n accessng land n regons domnated by large farmng cooperatves and companes. To capture ths, DOMFCO s a dummy varable whch equals one f more than 85% of the agrcultural land n a county s cultvated by farmng cooperatves and companes. The two other transacton costs ndcators are MEMCOOP and PARTCOMP whch are dummy varables whch equal one f a member of the household s a member of a cooperatve farm or a partner n a farmng company, respectvely. These relatonshps are expected to reduce transacton costs ether n rentng land out to these large farms, or n accessng land for the household farm. They are expected to have a postve mpact on access to land (rentng n) and also on rentng out land. The sales prce of the land s also lkely to affect the decson whether to buy or rent land. We do not have data at the household plot level on land prces. Therefore we nclude as a proxy the average land sales prce at the county level, adjusted for qualty, SALESPRICE. The margnal productvty of the land s affected by ts qualty. The nformaton on the qualty of the land plots used by the households provded by the household surveys had many mssng observatons. Therefore we use an ndcator varable of the average land qualty at the county level, QUALITY, whch based on data from the Hungaran statstcal offce and whch s measured n Gold Crown. Households workng on better qualty land are expected to rent n more land and rent out less. 19

20 We nclude the human captal varables AGEHH and EDUCHH, measurng the age and the educaton level of the household head. Both are expected to affect the margnal productvty of the land, and hence rental actvtes; although the mpact may be non-lnear (Rzov et al. 2001). Age may have a negatve mpact on rentng n (and a postve mpact on rentng out) as younger household heads are expected to be more dynamc and entrepreneural. On the other hand, experence wll ncrease wth age, whch would lead to hgher margnal productvty and hence more rentng n of land. The trade-off between both effects may cause a non-lnear effect wth rentng n frst ncreasng wth age and later declnng. Educaton, whch s measured as years of schoolng, s expected to have a postve mpact on rentng n because t ncreases the management capacty of the household. However, beyond a certan educaton level, household heads may get access to better off-farm opportuntes, and hence reduce ther labour allocaton to farmng and shft to off-farm employment. We test for non-lnear effects of the age and educaton varables by ncludng the squared terms of both varables. Fnally, three regonal varables, EAST, WEST, and SOUTH, are ncluded to capture addtonal fxed effects. The reference regon s North-Central Hungary, whch ncludes Budapest, the captal cty. 6 Results Two models were estmated usng sngle censored tobt regresson wth the AGLS estmator to account for possble endogenety. Each model was estmated on dfferent subsamples. The rent-n estmaton used a subsample wth zero or postve values for the 20

21 amount of land rented n. The rent-out estmaton used a subsample wth zero or postve values for the amount of land rented out, and excludng landless households. Estmatng sngle tobt models on the two subsamples was preferred over poolng the data and estmatng one least squares model, because the frst procedure allows ntercept and slope coeffcents to be dfferent for the two subsamples (see Skoufas (1995) for a more detaled argumentaton). The results of the estmatons of the models are summarzed n tables 3. We tested for collnearty of the ndependent varables by usng the testng procedure proposed by Belsley et al. (1980) to test whether there exsts a lnear relatonshp among the explanatory varables that leads to unrelable regresson estmates. The condton number of the matrx of ndependent varables equals 17 whch ndcates that there are no collnearty problems. The estmated coeffcents on the landownershp varables are all sgnfcant and ndcate some nterestng relatonshps between landownershp and rentng. The mpact of the land varables on the decson to rent n land s mostly non-lnear, wth sgnfcant coeffcent estmates for the squared terms of the varables. However, over the relevant doman of the analyss the frst order effects domnate. The land varables dd not have a non-lnear affect on the decson to rent out. Therefore we dropped the squared term of LANDOWNED and LANDBOUGHT n the regresson wth the amount of land rented out as dependent varable. The coeffcents of LANDOWNED n table 3 confrm our hypotheses that households who own more land are more lkely to rent out land and less lkely to rent n land, ceters parbus. However, we fnd a hghly sgnfcant and postve relatonshp between buyng of land n the prevous years (LANDBOUGHT) and rentng n of land n the current perod. Ths suggests that households who want to extend ther cultvated area do so by a combnaton of buyng and rentng land. Whle they may prefer buyng land for property 21

22 rghts securty reasons, faced wth mportant lqudty and credt constrants, they opt for rentng of addtonal land. More land bought n the prevous perods s lkely to both ncreases the credt constrants n the current perod because of the nvestments n the land purchase, and to reduce the margnal benefts of securty, whch falls wth more land purchased already. Both forces explan the postve effect of the LANDBOUGHT coeffcent. Ths concluson s consstent wth fgure 2 whch shows how both rentng n and buyng of land by households ncrease wth the cultvated area for the household farm. Hence, wth credt market constrants, both buyng and rentng n of land go together n the household s decson to ncrease ts land use. The estmaton results are consstent wth our hypotheses on the mportance of transacton costs n the land market. DOMFCO, MEMCOOP and PARTCOMP all have a hghly sgnfcant effect on rentng n of land by households. The domnaton of large farm organsatons reduces access to land by households through the rental market. When households are partners of farmng companes or members of cooperatves t s easer for them to rent land. Hence, these large farm organsatons contnue to have an mportant mpact on the development of farmng by household through ther mpact on the land market, n partcular n regons where they contnue to use most of the land. The coeffcent of SALESPRICE s sgnfcantly postve. Land rentng s more mportant n regons where the sales prce of land, corrected for land qualty, s hgher. Where buyng land s more expensve, ceters parbus, households prefer rentng land. Notce that ths trade-off n the current perod s not nconsstent wth the complementary relatonshp between buyng and rentng of land n an ntertemporal perspectve, as explaned above. We fnd no sgnfcant effect of the average land qualty n the county on household decsons to rent out or rent n land. Ths may mply that the data (county averages based on 22

23 old ndcators) do not suffcently reflect household effects, or alternatvely that other factors, such as land transacton costs and mperfectons n labour and credt markets are much more mportant factors n determnng household land rental decsons. The human captal varables age and educaton do not affect the decson to rent n land, nor the decson to rent out land n a non-lnear way. Although when we drop the square terms, t becomes clear that educaton generally has a postve effect on the decson to rent n land and that there s a postve effect of age on rentng out. Fnally, the coeffcents of the regonal varables show that rentng n of land s consderably less n Eastern and Southern Hungary, and rentng out s consderably hgher n Western Hungary. Western Hungary borders Austra and consderably rentng n ths regon s gong on by Austran farmers, sometmes n collaboraton wth local farms. At the same tme, the closeness of ths regon to the Austran border and of the North-Central regon to the captal suggests that rentng n of land s more actve n regons n geographcal proxmty to places where hgh ncomes are concentrated. 7 Conclusons Ths study derves several theoretcal hypotheses on what determnes the partcpaton of household farms n land markets n transton countres. Households management ablty and land endowment, land qualty and prces, transacton costs n the land market, credt market mperfectons and constrants on off-farm employment were dentfed as mportant factors affectng land rental actvtes of rural households. Our emprcal analyss, usng data from a representatve survey of small Hungaran household farms, provdes emprcal support for several of these hypotheses. More specfcally, we draw the followng conclusons. Frst, households use the rental market to rent n more land f ther land endowment s small compared to ther optmal farm sze, and to rent out land n the other case. In 23

24 combnaton, the frst and second concluson support the fndngs of Dennger and Jn (2003) on land markets n rural Vetnam that rental markets allow poor (n terms of land endowment) but able producers to access land and extend ther farm. Second, households combne buyng and rentng of land to adjust ther land holdng to the optmal farm sze. Buyng of land provdes them wth a number of advantages over rentng of land, such as securty of operaton and mproved nvestment ncentves. However, lqudty constrants n the presence of mportant credt market mperfectons restrcts buyng as a strategy to enlarge the farm. Rentng n of land s used to complement buyng of land for enlargng the farm sze. We fnd strong evdence that households who buy more land also rent more land. Ths concluson s consstent wth observatons n Western Europe and the Unted States where many prvate farms also combne rentng and buyng of land to extend ther farm sze (Sommer et al., 2001; Swnnen, 2002). Thrd, even n transton countres where the land reform s largely mplemented and land ttles dstrbuted, mportant transacton costs may reman and can hnder effcent land transactons. In some regons of Hungary where large cooperatve farms and farmng companes use the vast majorty of the land, the effcency of the land market and postve equty effects are constraned by mperfect competton and unequal access to nformaton and uneven enforcement of land rghts and exchange. Moreover, n general, households wth connectons to these large organzatons, e.g. because household members are partners or members n them, have prvleged access to land. Fourth, we fnd that land rental markets allow households wth hgher farm management capactes to access more land. Better educaton of the household head s postvely correlated wth rentng n of land. When households grow older they rent out more. As such rental markets play an mportant role n reallocatng land between households wth dfferent needs and capactes n managng farms. 24

25 In summary, these fndngs mply that land rental markets are playng an mportant role n reallocatng land n transton economes to those most n need,.e. households wth relatvely better farm management capactes and relatvely less endowed wth land. Land rental markets wll contnue to play an mportant role even when the mportance of land sales transactons grow, and should not be seen as a temporary nsttuton that wll dsappear. Therefore t s mportant to focus polcy attenton on a set of ssues whch need to be addressed n order to allow the rental markets to contrbute to further effcency mprovements and poverty reducton n rural areas. These attenton areas are, frst, mperfect competton n the land market and transacton costs caused by large farm operators, and, second, constrants n other rural factor markets, n partcular markets for credt and labour. 25

26 References Akerlof, George and Janet Yellen, 1986, Effcency Wage Models of the Labor Market, Cambrdge Unversty Press, New York. Bardhan, Pranab K., 1989, The Economc Theory of Agraran Insttutons, Oxford Unversty Press, Oxford Belsley, Davd, Edwn Kuh and Roy Welsch, 1980, Regresson Dagnostcs, Identfyng. Influental Data and Sources of Collnearty, Wley, New York. Brandt, Loren, Guo L, and Scott Rozelle, 2003, Land Rghts, Farmer Investment Incentves, and Agrcultural Producton n Chna, Chna Journal, forthcomng. Braverman, Avshay and Joseph E. Stgltz, 1982, Sharecroppng and the Interlnkng of Agraran Markets, Amercan Economc Revew, 72(4), pp Carter, Mchael and Ramon Salgado, 2001, Land Market Lberalzaton and the Agraran Queston n Latn Amerca n: de Janvry, A., Gordllo, G., Platteau, J.-P. and E. Sadoulet (eds.), Access to Land, Rural Poverty, and Publc Acton, Oxford Unversty Press, Oxford, pp Dasgupta, Partha, 1993 An Inqury nto Well-beng and Desttuton, Oxford Unversty Press, Oxford. Dennger, Klaus and Hans Bnswanger, 2001, The Evoluton of the World Bank s Land Polcy n: de Janvry, A., Gordllo, G., Platteau, J.-P. and E. Sadoulet (eds.), Access to Land, Rural Poverty, and Publc Acton, Oxford, pp Dennger, Klaus and Gershon Feder, 2002, Land Insttutons and Land Markets n: Gardner, B. and G.C. Rausser (eds.), Handbook of Agrcultural Economcs, Elsever Scence, Amsterdam. Dennger, Klaus and Jn Songqng, 2003, Land Sales and Rental Markets n Transton: Evdence from Rural Vetnam, World Bank Polcy Research Workng, No de Janvry, Alan and Jean-Phlppe Plateau, Gustavo Gordllo and Elsabeth Sadoulet, 2001, Access to Land and Polcy Reforms n: de Janvry, A., Gordllo, G., Platteau, J.-P. and E. Sadoulet (eds.), Access to Land, Rural Poverty, and Publc Acton, Oxford Unversty Press, Oxford, pp European Commsson, 1998, Agrcultural Stuaton and Prospects n the Central European Countres: Hungary, Drectorate General for Agrculture, Brussels. Eswaran, Mukesh and Ashok Kotwal, 1996, Access to Captal and Agraran Producton Organsaton, The Economc Journal, pp Feder, Gershon, 1985, The Relaton between Farm Sze and Productvty, Journal of Development Economcs, 18 (2-3), pp Lerman, Zv, 2001, Agrculture n Transton Economes: From Common Hertage to Dvergence, Agrcultural Economcs, 26(2), pp Lerman Zv, Csaba Csak, and Gershon Feder, 2002, Land Polces and Evolvng Farm Structures n Transton Countres, Polcy Research Workng Paper, No. 2794, The World Bank, Washngton DC. Macours, Karen, Alan de Janvry and Elsabeth Sadoulet, 2001, Access to Land through the Land Rental Market: The Case of the Domncan Republc, Paper presented at the 26

27 Amercan Agrcultural Economcs Assocaton 2001 Annual Meetng, August 5-8, Chcago, Illnos. Maddala, Gangadharrao, 1983, Lmted-Dependent and Qualtatve Varables n Econometrcs, Cambrdge Unversty Press, New York. Mathjs, Erk and Sándor Mészáros, 1997, Prvatsaton and Restructurng of Hungaran Agrculture. n: Swnnen, J, Buckwell, A. and E. Mathjs (eds), Agrcultural Prvatsaton, Land Reform and Farm Restructurng n Central and Eastern Europe, Ashgate, Aldershot, pp Mathjs, Erk and Johan Swnnen, 1998, The Economcs of Agrcultural Decollectvaton n East Central Europe and the Former Sovet Unon, Economc Development and Cultural Change, 47 (1), pp Platteau, Jean-Phlppe, 2000, Insttutons, Socal Norms, and Economc Development, Harwood Academc, Amsterdam Rzov, Maran, Dnu Gavrlescu, Hamsh Gow, Erk Mathjs, and Johan Swnnen, 2001, "Transton and Enterprse Reorganzaton: Farm Restructurng n Romana", World Development, 29(7), pp Sadoulet, Elsabeth, Rnku Murga and Alan de Janvry, 2001, Access to Land va Land Rental Markets n: de Janvry, A., Gordllo, G., Platteau, J.-P. and E. Sadoulet (eds.), Access to Land, Rural Poverty, and Publc Acton, Oxford Unversty Press, Oxford, pp Shapro, Carl and Joseph Stltz, 1984, Equlbrum Unemployment as a Worker Dscplne Devce, Amercan Economc Revew, 74(3), pp Skoufas, Emmanuel, 1995, Household Resources, Transacton Costs, and Adjustment through Land Tenancy, Land Economcs, 71(1), pp Sommer, Judth, Robert Hoppe, Robert Green, and Penelope Korb, 1995, Structural and Fnancal Characterstcs of U.S. Farms, 1995: 20 th Annual Famly Farm Report to Congress, Agrculture Informaton Bulletn, No 746, Economc Research Servce, U.S. Department of Agrculture, Washngton DC Swnnen, F.M. Johan, 2001, Transton from Collectve Farms to Indvdual Tenures n Central and Eastern Europe n: de Janvry, A., Gordllo, G., Platteau, J.-P. and E. Sadoulet (eds.), Access to Land, Rural Poverty, and Publc Acton, Oxford Unversty Press, Oxford, pp Swnnen, Johan, 2002, "Poltcal Reforms, Rural Crses, and Land Tenure n Western Europe", Food Polcy, 27, pp Swnnen, Johan and Hamsh Gow, 1999, "Agrcultural Credt Problems and Polces durng the Transton to a Market Economy n Central and Eastern Europe", Food Polcy, 21(1), pp Swnnen, Johan, Buckwell Allan and Erk Mathjs (eds.), 1997, Agrcultural Prvatzaton, Land Reform and Farm Restructurng n Central and Eastern Europe, Ashgate Publshng, London. Yao, Yang, 2000, The Development of the Land Lease Market n Rural Chna, Land Economcs, 76(2), pp