CREDIT RATIONING IN RURAL INDIA

Transcription

1 JOURNAL OF ECONOMIC DEVELOPMENT Volume 7, Number, December 00 CREDIT RATIONING IN RURAL INDIA Uppsala Unversty The vew that households are credt ratoned by the formal sector, rests on the assumptons that all households have a postve demand for formal credt and t s a cheaper source for borrowng. To emprcally verfy formal credt ratonng three dfferent models are estmated n ths paper. The frst model s a conventonal credt-ratonng model. The second model assumes that the probablty to borrow from the formal sector s jontly determned by the demand for credt and the decson of the bank on access. Fnally, the thrd model relaxes both these assumptons and the household chooses between borrowng from the formal or the nformal sector. Emprcal results usng recently collected data from Pur, Inda, confrm that the access to the formal sector n the rural credt markets s lmted and there exsts a hgh demand for credt. Ths suggests a hgh degree of effectve credt ratonng by the formal sector n Pur. Keywords: Credt, Ratonng, Rural JEL classfcaton: G0, C5, O7, Q4. INTRODUCTION Most theoretcal and emprcal studes of rural credt markets assume that they are charactersed by hgh lendng costs and a hgh demand for credt, resultng n hgh nterest rates beng charged to the borrowers. It s further assumed that all households have a postve demand for formal credt and that the formal sector, whch s bound by the government regulatons to lend at a subsdsed fxed nterest rate, s the cheapest source of credt. Models based on such assumptons usually suggest wde scale credt ratonng by the formal sector. However, ths mght not always be true as the degree of effectve credt ratonng mght not be as hgh as t s generally assumed n lterature and farm households may have a low demand for credt. Ths paper s an emprcal nvestgaton nto the extent of effectve credt ratonng by the formal sector n the rural credt markets of Pur. It s based on the Kochar (99, 997) where the household s free to choose between the formal and the nformal sector. However, the emprcal model s further modfed to nclude the addtonal nformaton on the access to credt that s avalable n our survey data. Another mportant

2 contrbuton of the paper s to update the stuaton on the extent of formal credt ratonng exstng n the rural credt markets n Pur, Inda, by usng recently collected data. Three dfferent scenaros are consdered for the estmaton of the effectve degree of credt ratonng. The frst model assumes that the probablty of borrowng from the formal sector s determned by the bank s decson on access. Two assumptons are made here. Frstly, that the formal sector provdes the cheaper source of credt. Secondly, that all households have a postve demand for formal credt. The second model drops the latter assumpton and the probablty of borrowng from the formal sector s jontly determned by the demand for credt and the bank s decson on access. Fnally, the thrd model consders a two-sector model (formal and nformal credt sector) wth the household choosng to borrow from the cheaper source of credt. The emprcal analyss n ths paper s based on the Rural Credt Market Survey of Pur dstrct n Inda. The survey has nformaton on 05 households that were selected by a two stage stratfed sample, from 66 vllages, spread over the admnstratve blocks of the Pur dstrct, whch s the coastal dstrct of Orssa, n east Inda. After deletng the households wth mssng values on some of the varables we are left wth nformaton on 989 households. Of the 989 households 76 were dentfed as farm households. The analyss n ths paper s based on these farm households, as there mght be a basc dfference between the demand for credt and the cost of supplyng credt to a farm and a non-farm household. The followng secton presents a short revew of the government nterventons n the rural credt markets of Inda. The subsequent secton examnes some of the emergng patterns. Secton 4 descrbes the econometrc framework whch s adapted to the addtonal nformaton on access to credt n the survey data. It also gves a concse descrpton of the varables used n the analyss. The results of the estmaton of the three dfferent models of sectoral choce are explaned n Secton 5 of the paper. Fnally, the last secton presents the concludng remarks.. INTERVENTIONS IN THE CREDIT MARKETS Hstorcally, the nformal sector has played a domnant role n the rural credt markets of Inda. Accordng to the All-Inda Rural Investment Survey, n 95-5, almost 83 percent of the cash loans were provded by the professonal moneylenders whereas the formal nsttutons provded only 8.7 percent. The Government of Inda decded to actvely ntervene n the credt market wth the commencement of the Integrated Scheme of Rural Credt (95). The am was to set about a systematc expanson of the nsttutonal credt nfrastructure, wth the Reserve Bank of Inda (RBI) The survey was conducted and supervsed by the author n 997.

3 CREDIT RATIONING IN RURAL INDIA 3 n the pvotal role. The emphass of the actve government nterventon n the credt market has been to provde subsdsed loans to agrcultural and other prorty sectors for nvestment purposes for development of the weakest sectons of the rural populatons and modernsng agrculture. In the fftes and the sxtes the co-operatves were the man agency of dstrbuton of formal credt. To meet the fast expandng demands for credt a mult-pronged credt delvery system was ntroduced. In 969, major commercal banks were natonalsed 3 wth the specfc objectve of lendng a certan percentage of ther credt to the prorty sectors. In 98 the Natonal Bank for Agrculture and Rural Development (NABARD) was set up by the RBI for provdng all types of producton and nvestment credt for agrcultural and rural development. NABARD s the apex nsttuton accredted wth all matters concernng polcy, plannng and operatons n the feld of credt for agrculture and other economc actvtes n rural areas. 4 In addton to other functons, t co-ordnates the rural fnancng actvtes of all nsttutons engaged n developmental work at the feld level and mantans lason wth Government of Inda, State governments, RBI and other natonal level nsttutons. Furthermore, a scheme of Rural Infrastructure Development Fund (RIDF) and that of specalsed branches for agrculture and small-scale ndustres were ntroduced n wth a vew of augmentng the flow of credt to the rural sector. In , out of the target of 40 per cent of the net bank credt of the domestc commercal banks to the prorty sector, 8 per cent had to be dsbursed n the form of credt for agrculture (The Annual Report on the Workng of the Reserve Bank of Inda (997), Secton.9). As of the end of March 997, the publc sector banks had exceeded ther prorty sector Prorty sectors nclude agrculture, small scale ndustres, retal, trade and small busnesses, road and water transport operators, professonal and self employed and educaton, housng to weaker sectons and consumpton loans etc. 3 The All Inda Credt Revew Commttee (969) supported the vew that commercal banks should ncreasngly extend fnance n rural areas. Ths process was accelerated by the natonalsaton of 4 major commercal banks n July 969 and another 6 banks n 980. The Commttee to Revew Arrangements for Insttutonal Credt for Agrcultural and Rural Development (CRAFICARD), 98, whch assessed the role of commercal banks n rural credt endorsed the vew that commercal banks could play a sgnfcant role n the varous programmes of rural development and made a seres of recommendatons to mprove the qualty of lendng through commercal banks. The publc sector banks numberng 8 (0 natonalsed banks and State Bank of Inda and ts 7 assocates) now account for more than 90 percent of the total busness of all scheduled commercal banks. 4 Its prme role s to provde credt for the promoton of agrculture, small scale ndustres, cottage and vllage ndustres, handcrafts and other rural crafts and other alled economc actvtes n rural areas wth a vew to promotng ntegrated rural development and securng prosperty of rural areas and for matters connected wth t.

4 4 credt target wth a total lendng of Rs. 79,30 mllon consttutng 4.7 per cent of the net bank credt. Dsbursement to agrculture under the specal agrcultural credt plans, prepared on the advce of the RBI, was Rs. 7,60 mllon durng (the Annual Report on the Workng of the Reserve Bank of Inda (997)). These specfc efforts to ncrease the outreac h of formal credt n the rural credt markets have gven postve results. From nearly 83 percent n 95, the proporton of cash debt from moneylender has gone down to 36. per cent n 97. Accordng to the report on A Revew of the Agrcultural Credt System n Inda (990) the percentage of borrowngs of the rural households (cultvators) from the formal sector to the total debt has ncreased from 9 (3.7) percent n 97 to 6. (63.3) percent n 98. However, ths expanson of the nsttutonal lendng n the rural areas faled to reach a larger number of cultvator households and dd not lead to any major ncrease n the agrcultural captal stock per farm (Bnswanger and Khandker (995)). In 98-8 only 3 percent of the rural cultvator households n Inda had outstandng loans from the formal sector. 5 The unequal dstrbuton of the formal credt was reflected by the fact that cultvators who owned more than Rs. 00,000 n assets receved 7 percent of the formal loans whereas they consttuted only 0 percent of the total cultvator households. In the hghest asset group n rural areas (ownng more than Rs. 500,000) 85 percent of the cash dues were outstandng to the formal credt nsttutons whereas only 5 percent were outstandng to the nformal nsttutons. In sharp contrast for the lowest assets group (ownng asssts of up to Rs 000) nearly 94. percent of the cash dues were outstandng to the nformal credt nsttutons (RBI (98)). 3. EMERGING PATTERNS IN THE RURAL CREDIT MARKET The provson of subsdsed loans by the formal credt nsttutons has led to the general belef n lterature that there exsts excess demand for the formal credt and lmted access to the formal credt nsttutons. Moreover the credt ratonng s beleved to be n favour of the households n more productve regons and of larger farm households that are economcally and poltcally more powerful. Ths s beleved to have lead to a further wdenng of the ncome nequaltes (Adams et al. (984), Braverman and Stgltz (989)). However, t s dffcult to conclude that ths s observed due to the type of formal credt and ts lendng rules. The lower number of households demandng loans from the formal sector mght be a choce decson of the household. The farms, whch are small and fragmented, are n areas wth a low level of nfrastructure facltes or whch already have a hgh captal stock wll have a low return on captal. Such 5 Accordng to Bnswanger and Khandker (995) the data from the All-Inda Debt and Investment Surveys of appear to serously underestmate debt of cultvators.

5 CREDIT RATIONING IN RURAL INDIA 5 households restrct ther demand for producton loans even f they have access to them. The extent of the effectve formal sector ratonng cannot be nferred from the proporton of the non-borrower households alone, the demand sde also needs to be taken nto account. The dstrbuton of the borrowers n our survey data shows a smlar pattern. The formal nsttutons favour land ownng households, n partcular those wth large farms, as s evdent from the hgher share of formal borrowers n the category of households ownng more than 4 hectares of land (Table ). In terms of the total amount borrowed, almost 67 percent s borrowed from the formal sector. The nequaltes n the dsperson of credt n favour of the large and the medum farm households are reflected by the fact that 8. percent of the total borrowers take 46 percent of the total amount borrowed from the formal sector n loans. Table. Dstrbuton of Borrowers by Farm Sze ) Farm Sze (n hectares) more than 4 Total margnal small medum large All Households 380 (49.9) 96 (5.8) 5 (0) 33 (4.3) 76 (00) All Borrowers ) Share of borrowers out of all the households Formal Borrowers 3) Share of formal sector borrowers out of all the borrowers Informal Borrowers 3) Share of nformal sector borrowers out of all the borrowers 04 (53.8) (43.6) (67.9) (3.7) (8.) (7.6) (8.5) (3.) (.9) (4) 45.5 (5) (.5) (00) (00) 59 (00) Notes: ) The fgure n the parenthess s the percentage of the cell wth respect to ts respectve row total. Ths gves us the percentage of that partcular category of households wth respect to the farm sze category. ) The percentage of each cell n ths row s calculated wth respect to ther correspondng cell n the frst row and s the lower fgure n the cell. Ths gves us the percent of borrowers n each farm sze category. 3) The percentage of each cell n these rows s calculated wth respect to the frequency of ther correspondng cell n the all second row and s the lower fgure n the cell. Ths gves the percent of the formal or the nformal borrowers wth respect to the total borrowers n each farm sze category. Source: Survey Data. Never the less the nformal sector contnues to play an mportant role n the rural credt markets. Sources of nformal credt lke frends and relatves, traders and

6 6 commsson agents and landlords have shown an ncrease n the proporton of cash debt from 8.4 percent to 3.3 percent from 95 to 97. In 98, frends and relatves were the most mportant source of nformal credt provdng 4 percent of the total nformal loans n Inda (Bell (990)). In our survey data frends and relatves are the second largest lendng group wthn the nformal sector wth 7.6 per cent of all households borrowng from them. Informal lenders charge a hgher nterest rate as compared to the formal sector. Ths could reflect the hgh rsk, admnstratve or opportunty costs faced by them. However, the nformal sector based on ther socal and geographc proxmty to the lender also face a lower screenng, montorng and enforcement costs. Therefore, although some nformal lenders charge a hgher nterest rate than the formal lenders, there exst other nformal lenders that charge a lower nterest rate as compared to the formal sector s nterest rate. Ths s often the case for the category of frends and relatves. The formal nterest rates on the other hand are admnstered by the RBI. The followng structure was mposed on the nterest rates charged: for loans below Rs. 5,000 an nterest rate of per cent was charged, for loans between Rs. 5,000 and Rs. 00,000, nterest rate was 3.5 per cent, whereas for loans bgger than Rs. 00,000 the banks were free to fx ther own nterest rates. 6 The overall pcture that emerges s that of a formal sector n the rural market that s bound by the government regulatons to lend a certan fxed percent of ts net bank credt to the prorty sector on a fxed nterest rate structure. Whereas, the nformal sector s free to charge any nterest rate based on t s cost and proft consderatons. 4. ECONOMETRIC FRAMEWORK Gven the above consderatons of the rural credt markets ths secton seeks to set up three dfferent models. The frst model depcts the most prevalent scenaro n lterature. It assumes that the formal borrowng s determned by the bank s decson on access (Unvarate Probt Model). The second model s a further generalsaton where the probablty of borrowng from the formal sector s jontly determned by the bank s decson on access as well as the household s demand for loans (Bvarate Probt Model wth partal observablty). The thrd model estmates the complete theoretcal model where the households may have zero demand for formal credt and they are free to choose between the nformal sector and the formal (Two-Sector Model). A. The Unvarate Probt Model (Model I) In lne wth popular belef n lterature ths model also assumes that the formal sector 6 97 per cent of the loans n our sample were of amounts smaller than Rs. 5,000.

7 CREDIT RATIONING IN RURAL INDIA 7 determnes whether the household has access to ts loan or not. Such an assumpton mples that the formal credt s the cheapest source of credt for all households and that all households have a postve demand for formal credt at the exstng nterest rate. The probablty of borrowng s gven by a unvarate normal dstrbuton whch reflects that () the probablty of household s demand for formal credt, () the probablty of access to formal credt and the (3) probablty that the formal credt costs less than the nformal credt, s dependent on the formal sector nterest rate. Ths nterest rate s fxed by the government, but the stpulated nterest rate vares across households of dfferent geographcal regons and by the farm sze. The formal sector nterest rate s predcted for all the households by takng predctors that are exogenous to the decson of takng credt. The nterest rates on the loans taken from the formal sector are regressed on the dummy for the admnstratve blocks, reflectng the agro economc stratum to whch the household belongs and sze of the land owned by the household. 7 B. Bvarate Probt Model Wth Partal Observablty (Model II) The unvarate probt model s one partcular nterpretaton of the market based on unrealstc assumptons, therefore further models are estmated. The bvarate probt model wth partal observablty drops the assumpton that all households have a postve demand for formal credt. The probablty of borrowng from the formal sector s jontly determned by the demand for credt and the bank s decson on access. The formal structure of the model s as follows: z = â X + ε ; y = f z > 0, else 0 ; z = â X + ε; y = f z > 0, else 0 ; ε ε ~ BVN ( 0,0,,, ρ),, where X s the vector of varables determnng the access to the loan and X are the explanatory varables determnng the demand for loan. Instead of observng both y and y we observe the product, y = y y where y and y are smultaneously determned and ε and ε are correlated (Porer 980). The log-lkelhood for ths model s 7 For the data used n Kochar (997) ths regresson explans almost all the varablty n the formal sector nterest rate wth an R square of 0.95 whereas for our estmaton t s only around Ths mght be due to the fact that the sample data for ths paper was collected from a smaller geographcal regon (wthn a dstrct).

8 8 ln L = ln Φ [ â X, â X, ñ] + ln( Φ[ â X, â X, ñ]), y= y= 0 where Φ s the bvarate cumulatve normal dstrbuton functon. C. The Two-Sector Model (Model III) The econometrc specfcaton n ths secton s based on the theoretcal model derved n Kochar (99, 997). However, the econometrc model has been further adapted to the nformaton that s avalable n the sample data on the access of the household to credt. 8 Furthermore, some emprcal changes have been ntroduced as the data n the sample was collected at the dstrct level 9 and has consderably less varaton. A lnear approxmaton of the household s loan demand schedule and the sectoral supply schedules avalable to the household are specfed as: B d = X δ +, Loan demand. (a) ã r v B s f = X δ +, Formal sector supply. (b) ã r v B s = X δ +, Informal sector supply. (c) 3ã3 r 3 v3 These yeld the reservaton nterest rate equatons, whch are estmated as the nterest rate at whch the optmal loan s zero MR ( 0, X = X â + u, reservaton demand rate. (a), u ) MC f ( 0, X = X â + u, reservaton cost, formal sector. (b), u) MC ( 0, X = X â + u, reservaton cost, nformal sector. (c) 3, u3) where the random varables u, =,,3 are assumed to be..d across households, though such omtted varables can be expected to enter nto all three equatons, yeldng a non dagonal co-varance matrx of the error terms. The reservaton schedules generate 8 However, for reader s convenence we contnue to use the notaton used n Kochar (997) so that the modfed lkelhood functon mght be easly related to her model. 9 Inda s dvded nto a number of states, each state s further dvded nto dstrcts. Each dstrcts s splt nto varous admnstratve blocks. In Orssa, the Pur dstrct has admnstratve blocks.

9 CREDIT RATIONING IN RURAL INDIA 9 the followng four ndex functons or selecton rules: Pr(demands nformal sector loan) = [ MR (0,, u ) MC (0, X, )] Pr X > 3 u, 3 Pr(demands formal sector loan) = [ MR (0,, u ) > ] Pr(access to formal) = f [ MC (0,, u ) < r ] Pr X r f, (3) Pr X, f Pr(formal sector s lower cost) = [ MC (0,, u ) > r ] Pr X. 3 3 f These ndex functons underle the household s partcpaton decsons and the access decson of the formal sector, whch generate the market outcomes. The choce between the sectors s determned by the household s demand and by the supply decsons of the formal and the nformal sectors, whereas the data only report market outcomes. Therefore, the nformaton on ndvdual demand and supply schedules s unavalable. However, our survey has addtonal nformaton on whether a household has access to credt or was dened credt n our data. The households n the sample were asked the queston dd you try to take any loan durng ths year, but dd not get t?. 0 Reported answers to ths queston gve the nformaton on whether the household had access to credt or not. It s dffcult to know from the data f t was the formal or the nformal credt sector that dened the loan to the household. The lkelhood functon s adjusted to nclude ths nformaton for the complete model. Combnng the nformaton on the borrowng of the household from the formal and the nformal sector wth the nformaton on whether t had access to credt or was dened credt all the households are dvded nto fve categores. Those households, (I) whch borrow from the formal sector, (II) those who borrowed from the nformal sector but were not ratoned and (III) those who borrowed from the nformal sector but sad that they were ratoned. The two remanng categores nclude (IV) households that dd not borrow even though they had access and (V) households that were non-borrowers because they were refused credt. 0 We are aware that ths does not completely capture the defnton of ratonng. Addtonal nformaton s needed on: whether the household dd not borrow because t expected that t would be turned down. In addton, one should also have nformaton on whether the household was loan sze ratoned. Nevertheless, we have to work wth the restrcted amount of nformaton that we have. Kochar (997) dvded the households nto those who borrowed formal credt; those who borrowed nformal credt and those who dd not borrow credt at all. In Kochar (997) the households borrowng from both sectors were allocated to ether sector dependng on whch sector charged the lowest nterest rate. If the formal (nformal) sector nterest was lowest the household was classfed as a formal (nformal) sector borrower.

10 0 The sectoral outcomes are generated as follows: (I) Probablty that the household borrows from formal sector = Pr( MR (0, X, u) > rf, MCf (0, X, u) < rf, MC (0, X3, u3) > rf ). (4a) (II) Probablty that the household borrows from the nformal sector but s not ratoned = Pr( MC (0, X 3, u3) < MR(0, X, u ), MC (0, X3, u3) < rf ). (4b) (III) Probablty that the household borrows from the nformal sector and s ratoned f = Pr( MC (0, X 3, u3) < MR(0, X, u), MC (0, X3, u3 ) > rf, MC (0, X, u ) > rf ). (4c) (IV) Probablty that the household dd not borrow even when they had access = Pr( MR(0, X, u) < rf, MC (0, X3, u3) > MR(0, X, u)). (4d) (V) Probablty that the household dd not borrow because they were ratoned = f Pr( MR(0, X, u) > rf, MC (0, X, u) > rf, MC (0, X3, u3) > MR(0, X, u)). (4e) The above equatons reflect both the formal sector ratonng and the optmal choces by the households. The data reports observed market outcomes; therefore the parameters of the equatons have to be nferred from the jont occurrence of the ndvdual decsons rules. The dentfcaton crteron for such models s relatvely weak, requrng the dstncton of one equaton from the other. Ths may be acheved by varyng the set of the explanatory varables across equatons. Therefore, the sector-specfc lendng costs enter the set of X and X varables but not X 3. The u s are assumed to be normally dstrbuted wth varance-covarance matrx ó I. The log lkelhood of the sample for the model generated from the above sectoral outcomes s specfed as: ln L = yln( ΦaΦbΦc) + y ln( Φe) + y3 ln( Φ f Φg ) + y4 ln( Φh) + y5 ln( ΦΦg ) where y, = to 5, s defned as follows: y = f the household borrows from the formal sector, y = f the household

11 CREDIT RATIONING IN RURAL INDIA borrows from the nformal sector but s not ratoned, y f the household borrows from the nformal sector and s ratoned, y f household does not borrow and s not 4 = 3 = ratoned and y 5 = f the household s a non borrower and s ratoned, and: Φ a b =Φ X â r ), ( f Φ =Φ X â + r ), ( f Φ =Φ X â r ), c ( 3 3 f Xâ X3â3 Φe = Φ(, r f X3â3, ρ), Φ Φ Xâ X3â 3 (, X3â 3, ), f = Φ r f ρ g = Φ X â r ), ( f X3â 3 Xâ Φh = Φ (, r f Xâ, ρ), Φ X3â3 Xâ (, Xâ, ), = Φ r f ρ where Φ s the cumulatve dstrbuton functon of the normal dstrbuton and Φ s the cumulatve dstrbuton functon of the standardsed bvarate normal dstrbuton. D. Descrpton of Varables Table gves the detals on the means and the standard devaton of the varous explanatory varables used n the estmatons. The demand for loans wll be affected by the household s characterstcs and the regonal varaton n agrcultural productvty and nfrastructure. The amount of land owned by the household, the qualty of land and the number of earnng and workng members n the households all affect the demand for credt. The exogenous treatment of the amount of land owned s justfed on grounds of lmted transactons n the land sales market observed n Inda. 3 3 For example, n a study of a vllage by Blss and Stern (98) the dstrbuton of vllage land n

12 Table. Descrpton of Varables Used n the Regressons Varable Descrpton Mean (Std. devaton) Household Characterstcs Interest Rate predcted formal nterest rate 4.5 (.7) Land Owned amount of land owned by the.3 (.) household Land Qualty s a dummy that s for good land 0.7 (0.4) qualty ) Males Number of workng male members.9 (.) n household Fnancal Assets fnancal assets of the household (0359.9) Famly Sze Number of members n the famly 8.5 (4.7) Age age of the head of household 48.0 (.9) Vllage/Block Characterstcs Road the dstance of the vllage from the 3. (3.7) nearest concrete road Fyeld95 The yeld of paddy (dry paddy wth.9 (5.0) husks) has been taken at the block level for 995 Plancult Planned amount of the bank credt 0.0 (0.004) dspersed for all purposes Ran95 The amount of monsoon ranfall n (39.4) Note: ) The survey asked the queston What s the sol qualty of the land that you own? To ths the reples were average, salne, sandy and others. The average and others sol qualty has been classfed as good land qualty. Formal and Informal sector supply schedules are functons of the set of varables just descrbed as they affect the farm productvty and the demand for loans. The formal sector schedule s dentfed by the planned amount of credt dsbursed per cultvator for all purposes, agrcultural and non-agrcultural. Ths varable affects the cost of loans to cultvators, gven that bank lendng for agrculture s requred to be a fxed percentage of the total formal credt. Ths mples that f the planned amount of credt dsbursed ncreases t should mply an ncrease n access to formal sector. The nformaton on ths varable s avalable at the block level. The regonal dummes are also used to dentfy the formal sector n the two-sector model (Model III). The regonal dummes reflect the dfferences n operaton of the formal nsttutons that operate n certan areas and are a functon of exstence of busnesses, farm effcency, market opportuntes and other closely reflected that before Independence (947), and that land s rarely bought or sold.

13 CREDIT RATIONING IN RURAL INDIA 3 dfferences. The demand schedule s further dentfed from the formal sector supply schedule by the number of workng members n the household. The banks are offcally requred to evaluate the loan on the bass of the ncremental ncome expected from the project that s fnanced. Evaluaton studes of some banks show that famly labour was not ncluded whle makng these calculatons 4 and therefore ths varable should not affect the access to the formal sector. The nformal sector s dentfed on the bass of the fnancal assets of the households. A. The Unvarate Probt Model 5. RESULTS AND CONCLUSIONS It s wdely beleved n emprcal lterature that the households have a postve demand for the formal credt, whch s the cheaper source of credt for all households. Ths mples the exstence of wde spread credt ratonng by the formal sector. Model I assumes that the probablty of borrowng from the formal sector s determned completely by the bank whch decdes whether a household should get a loan or not. Thus the access to formal credt s a bank s decson as all households have a postve demand for formal credt. Accordng to the estmaton results n Table 3 the probablty of access s determned by the amount of land owned by the households, the land qualty, famly sze, the amount of planned bank credt dspersed for all purposes per cultvator, the average monsoon ran and the food gran yeld per hectare. The larger the sze of land owned by the household, the greater s the probablty of ts access to the formal sector. A better land qualty also mproves the household s prospects of access to loan and hence borrowng. Food gran yeld per hectare s used as a proxy for the level of nfrastructure development of the regon n whch the household s located. Households n more developed areas wth a hgher level of nfrastructure faclty have better access to a formal sector loan. Ths s especally true for households located n areas near the bank branches, whch are nearly always located n the area, near the block s headquarters. Proxmty to the bank branches, government offces and pestcdes and fertlser shops also decreases the level of transacton costs for such households. Expectaton of good weather and envronmental factors proxed by the average monsoon ranfall mples expectaton of a hgher demand and hgher producton of output and hence has a postve effect on access to credt. The planned amount of bank credt to be dspersed for all purposes per cultvator however s negatvely related to the formal sector access. One reason for ths 4 For example, UCO Bank (999).

14 4 dscrepancy mght be arsng from the fact that the banks do not follow the offcal targets for lendng to the agrcultural and the prorty sector at the block level. Another reason could be that ths model s ms-specfed as compared to the bvarate model, where the coeffcent for the planned amount of bank credt to be dspersed for all purposes per cultvator has the expected sgn (though t s nsgnfcant). The model predcts a hgh degree of ratonng by the formal sector. The probablty of access to the formal sector yelds the value 0.9. Ths value s evaluated at the mean levels of the explanatory varables. Therefore, under the assumptons of model I 7 percent of the households are credt ratoned by the formal sector. 5 Table 3. Unvarate Probt Model of Access to Formal Credt Varable Parameter Estmates Standard Error Constant.00 a Predcted nterest rate a Amount of land owned (hectares.) 0.30 *** Land qualty 0.40 * 0.50 Amount of planned bank credt dspersed *** for all purposes per cultvator Age of the household head 0.00 a Sze of the household 0.09 * 0.0 Food gran yeld per hectare n ** 0.00 Monsoon ran n * 0.00 Pur Sadar 0.80 a 0.50 Kakatpur 0.70 a 0.60 Astarang 0.30 a 0.90 Kanas a 0.40 Pr (access) 0.9 a Log Lkelhood 40.6 a Sample Sze 76 a Note: * ( **, *** ) ndcates sgnfcance at 0% (5%, %) level. B. The Bvarate Probt Model Model II specfes the probablty of access to the formal sector as a bvarate normal dstrbuton, jontly determned by the demand for credt and the decson of the bank on 5 From data coverng a much larger part of Inda, Kochar (997) fnds that 8% of the households were ratoned under smlar assumptons as Model I. However, these results are based on data that was collected n 98 whereas our data s much more recent (996).

15 CREDIT RATIONING IN RURAL INDIA 5 whether t wants to lend to the household or not. Although t no longer assumes that all households have a postve demand for the formal sector, t s stll consdered to be the cheaper source of credt. Ths yelds the dfferences n the estmates of the Model I and Model II. Interest rates have a margnally negatve effect on the probablty of demand for loans whereas ts effect on access s weakly postve but statstcally nsgnfcant. It has often been clamed n lterature that t s the large land ownng farmers that beneft from the access to the subsdsed formal credt n the developng countres (Braverman and Guasch (986)). The results from Model II confrm ths to a certan extent (Table 4). The amount of land owned and the qualty of land both have a postve and sgnfcant effect on the access to formal credt, however, ther effect on the demand for loans s negatve and nsgnfcant. Table 4. Bvarate Probt Model of Demand and Access to Formal Credt Varable Access Demand Parameter est. Std. error Parameter est. Std. error Constant 3.00 a a.83 Predcted nterest rate 0.00 a ** 0.57 Amount of land owned (hectares.) 0.88 ** a 0.08 Land qualty *** *** 0.77 Age of the household head 0.00 a a 0.00 Sze of the household ** a Amount of planned bank credt a dspersed for all purposes per cultvator Food gran yeld per hectare n 0.05 a ** Monsoon ran n a a Dstance from the road 0.04 a * 0.05 Number of workng male members a 0.4 n the household Pur Sadar a Astarang ** Kakatpur a Kanas a Rho 0.98 Pr (access and demand) 0.40 Log Lkelhood Sample Sze 76 Note: * ( **, *** ) ndcates sgnfcance at 0% (5%, %) level.

16 6 In contrast to the results from Model I, access to formal sector s not affected by nfrastructural and weather factors lke the food gran yeld per hectare and the average monsoon ran. Dstngushng between the demand and the access s responsble for the dfferences n the results of Model I and II. Food gran yeld and the road varable, both of whch reflect the level of nfrastructural development of the regon postvely affect demand. Households lvng n such regons have easer access to facltes, lower transacton costs and better accessblty to ther markets and as such have a hgher demand for credt. The factors determnng the access to the formal sector and the demand for credt are dstnct, wth formal credt access manly determned by the land varables (sze of land owned and the qualty of land). The probablty of access to the formal sector gven the demand and access s 40 percent mplyng that nearly 60 percent of the households are credt ratoned. Consderng that all households n the rural credt markets of Pur mght not have a postve demand for formal credt mples that the households are not as constraned as Model I suggests and ths s confrmed by the hgher probablty of access to the formal sector n Model II. C. The Two-Sector Model The rural credt markets have a dverse set of lenders and credt from formal nsttutons mght not always be the cheapest. Credt from frends and relatves, landlords and employer mght be offered at lower nterest rates than the formal credt. Model III s a general model where the household wth a postve demand for credt can make the choce of borrowng from ether of the two sectors, the formal or the nformal. It therefore relaxes the assumpton that the formal sector s the cheapest source of credt. The results of the estmaton are presented n Table 5. Model III confrms the results from Model II that access to formal credt ncreases wth the sze of land owned. The larger the amount of land owned the smaller s the reservaton cost of the formal sector and leads to an ncrease n access to the formal sector. Dstance of the household from the nearest road reflectng the nfrastructural development of the regon where the household s located shows that access to formal credt for households located near roads ncreases. The regonal block dummes are the catch all for the dfferences n the regonal varatons for the formal sector and are all statstcally sgnfcant n determnng the access to the formal sector. The amount of planned bank credt dspersed for all purposes per cultvator s also a sgnf cant but wth a postve sgn, whch s contrary to expectaton. Ths may be explaned by the fact that the plannng for the amount of the credt dspersed s done n terms of the needs of the dfferent actvtes at the block level. However, targets for lendng to the prorty sector are set and followed at the dstrct and not the block levels. Ths dscrepancy between the plannng and the mplementaton at the block level mght be responsble for the unexpected sgn.

17 CREDIT RATIONING IN RURAL INDIA 7 Table 5. Model of Demand and Access to Formal and Informal Credt ) Varable Demand Reservaton cost for the formal sector Reservaton cost for the nformal sector Constant.907 *** (0.40).0 *** (0.49) *** (0.35) Amount of land owned (hectares.) *** (0.06) 0.6 *** (0.080) *** (0.074) Land qualty 0.56 *** (0.053) *** (0.084) 0.08 *** (0.065) Famly sze *** (0.076) *** (0.068) 0.09 *** (0.070) Age of the household head 0.7 *** (0.053) ** (0.065) 0. *** (0.070) Dstance from the road *** (0.060) 0.85 *** (0.083) *** (0.064) Number of workng members n 0.60 *** (0.07) - - the household Amount of planned bank credt *** (0.0) - dspersed for all purposes per cultvator Pur *** (0.098) - Kakatpur *** (0.065) - Astarang *** (0.098) - Kanas *** (0.080) - Fnancal assets *** (0.048) Pr (access to formal sector) 0.53 Pr (demand) 0.76 Pr (formal sector s cheaper) 0.7 Pr (access/demand, formal sector 0.8 s cheaper) Mean Log Lkelhood. Sample Sze 76 Notes: ) The model was estmated allowng ρ and ρ to take any value. The parameter estmates (standard errors) for ρ and ρ are (0.058) and (0.058). ) * ( **, *** ) ndcates sgnfcance at 0% (5%, %) level. Fgures n the bracket are standard errors. Although land ownershp ncreases access to formal loans, the estmates of the reservaton cost for the nformal sector shows that t has no effect on the access to nformal loans. The varables that determne access to nformal loans are the age of the head of household, the dstance from the road and the fnancal assets owned by the household. Dstance from the road s a proxy for the nfrastructure development of the regon n whch the household s located. Proxmty to the road mples a lower reservaton cost for the nformal sector and mples greater access to nformal credt. Smlarly, households wth larger fnancal assets face a lower reservaton cost for nformal sector and hence have greater access to nformal credt. The factors that determne the demand for the households are amount of land owned,

18 8 land qualty, famly sze, age of the household s head and the number of workng male members n the household. The negatve relaton between the amount of loan owned and the credt demanded results from the fact that very large landowners and households wth small and fragmented farms have a low demand for credt. The number of males earnng n the household mght be engaged n prmary actvtes other than cultvaton for example, workng as a salared personal etc., whch would mply a fxed ncome and lower demand for credt for agrcultural nvestment and consumpton smoothng purposes. The estmates of the model suggest that a majorty of farm households (7 percent) face a lower reservaton cost of credt n the formal sector. The demand for credt s farly hgh n the regon, at 76 percent. The access to the formal sector s 53 percent among all households, but condtonal on demandng credt from the formal sector, t s 8 percent. Ths suggests a hgh degree of effectve credt ratonng of 7 percent of the households seekng credt from the formal sector. 6. CONCLUDING REMARKS The estmaton of the three models confrms the general belef n the lterature that a consderable number of households are credt ratoned by the formal sector. The degree of the effectve ratonng suggested by the three models s 7 percent (Model I), 60 percent (Model II) and 7 percent (Model III). The estmated parameters of the two-sector model were used to predct probabltes that were compared to the actual dstrbuton n the fve groups of households (see Equaton 4) and reflected a smlar pattern. 6 Ths suggests good predctve performance by Model III. By constructon, Model III s the closest to realty and presents the most relable results of all the three models. Gven that the two-sector model s the most general model, the relatvely lower level of credt ratonng n Model II could result from a ms-specfcaton. Ths would mply that the parameter estmates and thus also ther predctons cannot be expected to be good. Some recent lterature (Kochar (997), Bell (990)) has suggested that the demand for credt s low and the role that credt can play n enhancng agrcultural development s lmted. However, most of ths emprcal evdence s based on data from the 80s whch was collected from relatvely more productve areas. The evdence for hgher degree of ratonng should not be surprsng, gven that Orssa s one of the poorest states n Inda 7 6 The estmated parameters of Model 3, were used to predct the probablty for every group n equatons 4(a to e). The mean of the estmated probabltes for each group was then compared to the actual probablty n the data. 7 Report of the Expert Revew Commttee, submtted to the Plannng Commsson n Inda n 993.

19 CREDIT RATIONING IN RURAL INDIA 9 wth a hgh degree of dependence on weather and tradtonal methods of agrculture. 8 The estmates n Model III suggest a relatvely hgh demand for credt wth farly large credt ratonng n Pur, suggestng a need for further development of the credt programs. Ths requres not just ncreased outreach of credt to the cultvators but also well-desgned credt facltes that beneft the dsadvantaged and not just the rch and large landowners. The estmates of Model III reflect the land ownershp bas of the formal sector as the amount of land owned s a sgnfcant determnant of the acces s to formal credt. Recognsng that the project lendng approach generally followed by the banks mght not be sutable to the rural poor whose need were small loans and loans for short perod, credt facltes have been modfed by the formal sector n Pur. Some steps n ths drecton have been taken by the promoton of SHG 9 (Self Help Groups) by NABARD. The results n ths paper therefore support the lterature that states that credt polces stll have an mportant role to play n agrcultural development. Gven the hgh demand for credt and the lmted access to formal credt n Pur, the degree of effectve credt ratonng n Pur s very hgh. Ths result holds even when we relax the assumptons that all households have a postve demand for formal credt and that the formal sector s the cheaper source of credt. REFERENCES Adams, D.W., H.G., J.D. von Pschke (eds.) (984), Undermnng Rural Development wth Cheap Credt, Westvew Press, Boulder. Bell, C. (990), Interactons between Insttutonal and Informal Credt Agences n Rural Inda, The World Bank Economc Revew, Vol. 4, No. 3, Bnswanger and Khandker (995), Formal Fnance and the Indan Rural Economy, Journal of Development Studes. Blss, C.J., and N.H. Stern (98), Palanpur: The Economy of an Indan Vllage, Oxford Unversty Press, New Delh. Braverman, A., and J.L. Guasch (986), Rural Credt Markets and Insttutons n Developng Countres: Lessons for Polcy Analyss from Practce and Modern 8 PLP, Natonal Bank for Agrculture and Rural Development, Realsng the need for lendng to the poor n the rural areas, NABARD formulated a plot project n 99-9 for brngng out some nnovatons n moblsaton of the rural savngs and dspensaton of credt by way of lnkage of Self Help Groups wth banks. As per the nstructons contaned n RBI crcular R.P.C. D No. PL.BC. 0/04.09./95-96 dated nd Aprl 996 the SHG lnkage programme has become the normal actvty for the banks. Under ths, a small, homogenous and affnty group of rural poor, voluntarly formed to save and mutually agree to contrbute to a common fund to be lent to ts members as per group decsons (Potental Lnked Credt Plan ).

20 0 Theory, World Development, Vol. 4, No. 0, Braverman, A., and J.E. Stgltz (989), Credt Ratonng, Tenancy, Productvty, and the Dynamcs of Inequalty, n The Economc Theory of Agraran Insttutons, ed. by P. Bardhan, Oxford Unv. Press, New York. Government of Inda (987), All Inda Report on Agrcultural Census, 980-8, Mnstry of Agrculture, Department of Agrculture and Co-operaton, Agrcultural Census Dvson. Kochar, A. (99), An Emprcal Analyss of Ratonng Constrants n Rural Credt Markets of Inda, Ph.D. dssertaton, Unversty of Chcago. (997), An Emprcal Investgaton of Ratonng Constrants n Rural Credt Markets n Inda, Journal of Development Economcs, Vol. 53, Natonal Bank for Agrculture and Rural Development (997), Potental Lnked Credt Plan (PLP) for , Bhubaneswar, Orssa. RBI (Reserve Bank of Inda) (954), All-Inda Rural Credt Survey, Vol., The Survey Report, Vol., The General Report, Vol. 3, The Techncal Report, Bombay. (98), All-Inda Rural Debt and Investment Survey. (990), A Revew of the Agrcultural Credt System n Inda, Report of the Agrcultural Credt Revew Commttee, Bombay. (997), The Annual Report on the Workng of the Reserve Bank of Inda, for the year July, 996 to June 30, 997. UCO Bank (999), Manager s Handbook, 4 th edton. Malng Address: Assstant Professor, Department of Economcs, Uppsala Unversty, Box 53, S-750, Uppsala, Sweden. E-mal: Ranjula.Bal@nek.uu.se