Do Agricultural Subsidies Crowd-out or Stimulate Rural Credit Market Institutions?: The Case of CAP Payments

Transcription

1 Do Agriultural Subidie Crowd-out or Stimulate Rural Credit Market Intitution?: The Cae of CAP Payment PAVEL CIAIAN, JAN PORIVCA AND ATARINA SZEGENYOVA European Commiion (DG Joint Reearh Centre IPTS); Eonomi and Eonometri Reearh Intitute (EERI), Catholi Univerity of Leuven (LICOS); Slovak Agriultural Univerity Paper prepared for preentation at the EAAE 2011 Congre Change and Unertainty Challenge for Agriulture, Food and Natural Reoure Augut 30 to September 2, 2011 ETH Zurih, Zurih, Switzerland Copyright 2011 by Pavel Ciaian, Jan Pokrivak and atarina Szegenyova. All right reerved. Reader may make verbatim opie of thi doument for non-ommerial purpoe by any mean, provided that thi opyright notie appear on all uh opie. 1

2 Do Agriultural Subidie Crowd-out or Stimulate Rural Credit Market Intitution: The ae of CAP Payment 1 Pavel Ciaian 2, Jan Pokrivak 3 and atarina Szegenyova 4 Abtrat In thi paper we etimate the impat of CAP ubidie on farm bank loan. Aording to the theoretial reult, if ubidie are paid at the beginning of the growing eaon they may redue bank loan, wherea if they are paid at the end of the eaon they inreae bank loan, but thee reult are onditional on redit ontraint and on the relative ot of internal and external finaning. In empirial analyi we ue the FADN farm level panel data for period We employ the fixed effet and the GMM model to etimate the impat of ubidie on farm loan. The etimated reult ugget that (i) ubidie influene farm loan and the effet tend to be non-linear and indiret; (ii) both oupled and deoupled ubidie timulate long-term farm loan, but the long-term loan of big farm inreae more than thoe of mall farm due to deoupled ubidie; (iii) the hort-term loan are affeted only by deoupled ubidie, and they are altered by deoupled ubidie more for mall farm than for large farm; however (v) when ontrolling for the endogeneity, only the deoupled payment appear to affet loan and the relationhip i non-linear. Introdution Agriultural ubidie have important impat on agriultural market. Beide affeting farmer inome, tudie have hown that agriultural ubidie ditort input and output market and thu alter rent of other agent ative in the agriultural etor (for example onumer or input upplier). The impat of agriultural ubidie on inome ditributional effet depend on their type, truture of market and the exitene of market imperfetion (Alton and Jame 2002; de Gorter and Meilke 1989; Gardner 1983; Guyomard, Mouël, and Gohin 2004; Salhofer 1996; Ciaian and Swinnen 2009). Studie alo evaluate, among other, impat of ubidie on the environment and agriultural publi good (e.g. Beer Van Cee and Van Den Bergh 2001; hanna, Iik and Zilberman 2002) or produtivity and market ditortion (e.g. Chau and de Gorter 2005; Goodwin and Mihra 2006; Skokai and Moro 2006). With few exeption (e.g. Ciaian and Swinnen 2009), mot of thee tudie invetigate the diret impat of ubidie (on prie, quantitie, inome, environment, et.) by auming that ubidie do not alter the truture of agriultural market and do not interat with market intitution. In reality government poliie may have variou unintended effet. They an hange the truture of the market organization or rowd out ome market intitution. An analyi of uh effet goe beyond the fou of the urrent agriultural poliy analyi literature. 5 The objetive of thi paper i to ae the impat of the European Union Common Agriultural Poliy (CAP) on farm bank loan. Firt, extending the model of Feder (1985), Carter and Wiebe (1990) and Ciaian and Swinnen (2009) we theoretially analyze how ubidie may affet farm loan. Then, employing a unique farm level Farm Aountany Data Network (FADN) panel data for the period we empirially etimate the interation between CAP ubidie and farm loan. To our knowledge, thi paper i one of the firt attempt to tudy empirially how agriultural ubidie affet rural redit intitution. 1 The author are grateful to Unit L.3 Miroeonomi analyi of EU agriultural holding from the European Commiion for granting ae to the FADN data. The author aknowledge finanial upport from the European Commiion FP7 projet Rural Fator Market. The author are olely reponible for the ontent of the paper. The view expreed are purely thoe of the author and may not in any irumtane be regarded a tating an offiial poition of the European Commiion. 2 European Commiion (DG Joint Reearh Centre IPTS); Eonomi and Eonometri Reearh Intitute (EERI); and Catholi Univerity of Leuven (LICOS). pavel.iaian@e.europa.eu. 3 Slovak Agriultural Univerity; and Eonomi and Eonometri Reearh Intitute (EERI). 4 Slovak Agriultural Univerity 5 Thee iue are related to rowding out effet of other type of government program extenively analyzed in the literature. For example, the interation between private tranfer and publi welfare program attrated oniderable attention from aademi tudie (e.g. Barro 1974; Lampman and Smeeding 1983; Robert 1984; Maitra and Ray 2003; Cox, Hanen and Jimenez 2004). 2

3 The Model We build a theoretial framework of the preent tudy on the model of Feder (1985), Carter and Wiebe (1990), and Ciaian and Swinnen (2009). In thi tudy we extend the three model by analyzing how ubidie affet farm demand for bank loan. We onider a repreentative profit-maximiing farm. The farm output i a funtion of the fixed amount of land ( A ), fixed quantity of family labour (F) 6 and non-land input ( ), whih we refer to a fertilizer but whih apture alo other apital input ued by the farm. The prodution funtion i repreented by a ontant return to ale prodution tehnology f ( A,, F) with f > 0, f < 0, f > 0, for i, j = A,, F. We aume that all land i owned by the farm. End-of-eaon profit i: ii ij (1) = pf ( A,, F) k where k = ( 1+ i ) k, p i the prie of the final produt, k i the per unit prie of fertilizer and i i the interet rate. We aume that the eonomy i mall and open, whih implie that the fertilizer prie, the interet rate, and the output prie are fixed. An important iue i the timing of variou ativitie and payment. We aume that fertilizer i paid for at the beginning of the prodution eaon, wherea the revenue from the ale of prodution i olleted after harvet at the end of the eaon. Beaue of the time lag between the payment for fertilizer (variable input) and obtaining revenue from ale of prodution the farm ha a demand for the hort-term redit. The demand for redit an be atified either internally (ah flow, aving, ubidy) or externally (bank loan, or trade redit). For the ake of impliity we onider only external finaning through the bank loan and later on in the paper alo ubidy 7. The demand for redit might not be fully atified, whih mean that the farm an be redit ontrained. Perfet Credit Market To etablih a point of omparion, we firt identify the equilibrium without redit ontraint. With perfet redit market, the farm i not ontrained on the quantity of input it ue. The farm hooe the quantity of fertilizer that maximize it profit given by equation (1). Thi implie the equilibrium ondition: (2) pf = k The ondition (2) determine the farm fertilizer demand funtion. Total fertilizer demand i repreented by funtion D in Figure 1. Beaue we aume that both fertilizer prie and the interet rate are fixed, the upply of fertilizer i horizontal urve, S. The equilibrium quantity of fertilizer with no redit ontraint i. Imperfet Credit Market It i aumed that the maximum amount of money that the farm an borrow from the bank (C ) for fertilizer purhae depend on the farm ollateral (W ). For the ake of impliity we onider that bank aept only aet a ollator. 8 That i C = C( W ) where dc dw > 0. The redit ontraint i given by: (3) k C(W ) With the redit ontraint the farm maximize the end-eaon profit given by equation (1), ubjet to the redit ontraint (3). Thi amount to olving the LaGrangean funtion: i 6 The aumption of fixed amount of land and family labour i not tritly needed to obtain the reult. We introdue thi aumption in order to implify the expoition of the model reult. 7 Thi aumption i not tritly needed to obtain the reult. 8 Thi aumption i not tritly needed to obtain the reult. In reality, the level of farm redit may depend on farm harateriti (e.g. reputation, owned aet, profitability). In general, the evidene from the literature how that thee fator are important determinant of farm redit (e.g. Benjamin and Phimiter, 2002; Petrik and Latruffe, 2003; Latruffe, 2005; Briggeman, Towe and Morehart; 2009). For example, Latruffe (2005) find in the ae of Poland that farmer with more tangible aet and with more owned land were le redit ontrained than other. Briggeman Towe and Morehart (2009) find for farm and non-farm ole proprietorhip in US that the probability of being denied redit i redued, among other, by net worth, inome, prie of aet, and ubidie. 3

4 (4) Ψ = pf ( A,, F) k λ ( k C) where λ i the hadow prie of the redit ontraint. When the redit ontraint i binding, λ > 0, the farm annot ue the unontrained optimal level of fertilizer and the quantity demanded of fertilizer i determined by = C( W ) k.the optimality ondition are: (5) pf ( 1 + i + λ ) k = 0 (6) k C 0. From equation (5) it follow that the marginal value produt of fertilizer i higher than the marginal ot of fertilizer k : pf > k. By inreaing fertilizer ue the farm ould inreae it profit but the redit ontraint doe not allow it to buy additional fertilizer. In Figure 1 the redit ontraint urve (i.e. fertilizer upply), repreented in term of fertilizer unit, i given by the bold line k A S, where S = C k. With the redit ontraint, the equilibrium ue of fertilizer i equal to. Under redit ontraint the farm ue le fertilizer than under the perfet ompetition, <. Subidie and Credit Contraint We define DS a a deoupled ubidy whih the farm reeive irrepetive of it level of prodution. With ubidy the objetive funtion of the farm beome: (7) = pf ( A,, F) k k DS + where k = ( 1+ i ) k, i the fertilizer finaned through the bank loan C, i the fertilizer finaned with ubidie DS, i repreent farm opportunity ot of ubidy (i.e. the return on the mot profitable alternative invetment opportunity), and = +. We aume that the ot of bank loan i higher than the ot of ubidy, i > i. Thi aumption i reaonable given the information and inentive problem involved in providing a loan to the farm. Subidie affet not only the profit funtion but alo the redit ontraint. If the farm reeive ubidy at the beginning of the eaon, it an ue it for paying for the fertilizer. Reeiving the ubidie at the end of the eaon improve the farm ae to redit too. Future guaranteed payment of ubidy may erve a ollateral for obtaining loan from the bank (Ciaian and Swinnen 2009). Therefore ubidy may alleviate the redit ontraint of the farm irrepetive of the timing of the ubidy. With ubidy redit ontraint i given by the following inequalitie: (8) k C[ W + ( 1 α ) DS] + αds (9) k αds where α i a dummy variable taking value zero when farm ue ubidy to purhae fertilizer diretly or one when ubidy improve ae to fertilizer indiretly through enhaned value of ollateral. Equation (8) therefore implie that the farm an ue two oure to finane the purhae of fertilizer: ubidy, α DS, and/or the bank loan, C [ W + ( 1 α) DS]. If ubidy i paid at the beginning of the eaon, α =1, the farm an ue it to alleviate it redit ontraint. On the other hand when ubidy i paid at the end of the eaon, α = 0, the farm may ue it a ollateral to obtain a bank loan. In other word, ubidy inreae farm aet, improve it reditworthine and thu inreae ae to bank loan. Equation (9) tate that the ue of ubidy for fertilizer purhae annot exeed the total value of ubidie DS. With ubidie and redit ontraint the objetive funtion of the farm i repreented by the following LaGrangean funtion: (10) Ψ = pf ( A,, F) k k λ{ k C[ W (1 α) DS] αds} λ ( k αds) where λ i the hadow prie of the ubidy ontraint (9). The optimality ondition are given by: (11) pf ( 1 + i + ) k = 0 λ 4

5 (12) pf ( 1 + i + ) k = 0 λ (13) k C[ W ( 1 α ) DS] αds 0 (14) k α DS 0. The Impat of Deoupled Subidy Firt, we onider the impat of deoupled ubidy on the bank loan under perfet redit market. Then we analye the redit ontraint ae. We ummarie our reult in three hypothee. Hypothei 1: If farm are not redit ontrained and if finaning via bank loan i more expenive than finaning through ubidie, ubidie redue the amount of farm bank loan if they are paid at the beginning of the eaon; otherwie ubidie do not affet the farm loan. If finaning via the bank loan i more expenive than finaning through the ubidy, i > i (i.e. k > k ), ubidy an redue amount of bank loan but only in the ae when the ubidy i paid at the beginning of the eaon. In uh a ae the farm an ue the ubidy intead of the bank loan to buy fertilizer. The ituation i illutrated graphially in Figure 1. With no redit ontraint and with no ubidie, the equilibrium fertilizer ue i and all fertilizer i finaned through the bank loan. Availability of heaper finaning through ubidy DS allow the farm to redue the amount of bank loan. The farm will ue le loan and part of the fertilizer will be finaned with ubidy, equal to ( = DS k ). The remaining fertilizer,, will be bought with the bank loan. In welfare term the farm gain area a in Figure 1. Note that with ubidy DS, the equilibrium fertilizer ue i not affeted and remain at. Only if ubidie rowd out all bank loan, whih our for uffiiently high ubidie (if DS > k ), the equilibrium fertilizer ue inreae. If the ubidy i paid at the end of the eaon, the farm annot ue it diretly to purhae fertilizer. However, ubidy an till be ued a ollateral. We aume that the type of ollateral doe not affet bank loan interet rate; hene the ubidy doe not alter the equilibrium quantity of loan. 9 Next we analye the ae when farm i redit ontrained and ubidy i paid at the beginning of the eaon, α = 1. Hypothei 2: If farm are redit ontrained and if ubidie are paid at the beginning of the eaon, (a) farm will ue the ame amount of loan with and without ubidie if ubidie are uffiiently mall and (b) the farm will redue bank loan if ubidie are uffiiently high. If the ubidy i paid at the beginning of the eaon the farm an ue it diretly to finane the purhae of fertilizer. The impat of ubidy on the bank loan under redit ontraint i illutrated in Figure 2. The equilibrium quantity of fertilizer with the redit ontraint and with no ubidy i. Firt onider ubidy DS 1. The ubidy DS 1 hift the upply of fertilizer from k D S to k AE S 1, where S 1 = ( C + DS1) k. The equilibrium quantity of fertilizer i 1 ( = ( C + DS1) k ). Some fertilizer i finaned diretly from the ubidy, 1 ( 1 = DS 1 k ), and the ret i finaned through the bank loan, ( = 1 1 = C k ). The farm gain area ad when ubidy i ued to purhae fertilizer under the redit ontraint. Subidy DS 1 doe not hange the quantity of the bank loan: with and without the ubidy the farm purhae the ame amount of fertilizer through the bank loan,. With ubidy DS 1 the farm i till redit ontrained ( λ > 0 ) the amount of fertilizer 1 i lower than the amount of fertilizer ued under 9 In reality, the type of ollateral may affet the ot of the loan. For example, if bank pereive ubidie to be more eure and/or have lower tranation ot to adminiter them than other type of farm ollateral, then the interet rate may be lower for ubidy baked loan than for the loan baked by the other type of ollateral. 5

6 perfet market, 1 < and thu it i profitable for the farm to exploit fully all available finaning opportunitie (ubidie and loan). However, if ubidy i uffiiently high, it an redue the amount of bank loan. For example, with ubidy DS 2, where DS2 > > DS1, the upply of fertilizer hift to k BF S 2, where S 2 = ( C + DS2) k (Figure 2). The equilibrium fertilizer ue hange to : 2 ( = DS 2 k ) i finaned from ubidy and 2 i finaned with the bank loan. Now, the farm ue maller amount of loan. The amount of fertilizer finaned with the bank loan i 2, whih i le than the total amount of fertilizer finaned with bank loan without ubidy. Intuitively ubidy DS 2 eliminate the redit ontraint (i.e. the redit ontraint (13) i not binding with DS 2, λ = 0 ) and farm ubtitute part of expenive bank loan with heaper ubidie. With ubidy DS 2 the farm i not redit ontrained and it ue the ame level of fertilizer a under the perfet market ituation,. Finally, we onider the ituation with binding redit ontraint when ubidy i paid at the end of the eaon, α = 0. Hypothei 3: If farm are redit ontrained and if ubidie are paid at the end the eaon, the farm inreae bank loan. The graphial analyi i in Figure 3. The fertilizer upply without the ubidy and with the redit ontraint i k A S and the equilibrium fertilizer ue i. If the redit ontraint (8) i binding ( λ > 0 ), it i profitable for the farm to ue the ubidy DS paid at the end of the eaon ( α = 0 ) a ollateral for obtaining a bank loan for purhae of fertilizer at the beginning of the eaon. Hene, beaue of higher ollateral, ubidie inreae bank loan from C (W ) to C ( W + DS), where C ( W ) < C( W + DS). The availability of more loan hift the fertilizer upply to k B S 1 and the equilibrium fertilizer ue to 1, where 1 >. Relative to the ituation with no ubidy, the farm benefit from more loan; the gain i equal to area a. Note that for uffiiently high ubidy, the farm may beome redit unontrained (i.e. λ = 0 ). For example, thi i the ae when ubidy hift the fertilizer upply to k D S 2 whih inreae the equilibrium ue to and generate a gain for the farm equal to area ab. Eonometri peifiation Theoretially the impat of deoupled ubidy on agriultural loan i ambiguou. Agriultural ubidie paid at the end of the eaon have no impat on bank loan under perfet market while they redue bank loan when paid at the beginning of the eaon. Under redit ontraint ubidie paid at the beginning of the eaon have no impat on bank loan if they are uffiiently mall but they redue bank loan if they are uffiiently high. Furthermore, under redit ontraint when ubidie are paid at the end of the eaon they redue bank loan. The relationhip between ubidie and bank loan i therefore an empirial quetion. Solving the maximiation problem (equation (11)-(14)), the amount of farm loan depend on farm ubidy, profitability, and aet. We therefore etimate the following eonometri model: (15) loan = β 0 + βdds + βaaet + βπ Π + β x X + ε where ubript j and t repreent farm and time, repetively; loan tand for farm bank loan, aet are farm aet and X i a vetor of obervable ovariate uh a farm harateriti, regional, and time variable. A uual, ε i the reidual term. 10 We are epeially intereted in etimating the parameter β d whih meaure the impat of ubidie on bank loan. Statitially ignifiant negative value of the oeffiient onfirm either hypothei 1 (ubidie paid at the beginning of the eaon redue bank loan beaue farm are not redit ontrained) or 10 The definition of the ret of the variable i the ame a in the theoretial etion. 6

7 hypothei 2b (uffiiently high ubidie paid at the beginning of the eaon eliminate bank loan when farm are redit ontrained). Statitially ignifiant and poitive oeffiient onfirm hypothei 3 (farm are redit ontrained and ubidie are paid at the end of the eaon). Finally, if the oeffiient i not tatitially ignifiant, then the hypothei 2a hold (farm with ubidie remain redit ontrained and ubidie have no effet on farm loan). We expet that data will onfirm either hypothei 2 or 3 beaue there i overwhelming evidene that farm are redit ontrained (Carter 1988; Blanard et al. 2006; Lee and Chamber 1986; Fare, Grokopf and Lee 1990). Further, anedotal evidene indiate that ubidie tend to be paid at the end of the eaon 11 whih implie that the hypothei 3 hould hold. Thi i in partiular the ae of the long-term loan whih tend to finane invetment with higher value than hort-run loan. 12 Hene, the annual value of farm ubidie may not over the full ot of the long-term invetment even if they are reeived at the beginning of the eaon. The hypothei 3 i more likely to hold in the ae of the long-term loan. The etimation of equation (15) i ubjet to the omitted variable bia and partiularly to the endogeneity. There are unobervable harateriti like farmer ability that affet bank loan and may be orrelated with explanatory variable. Ignoring thi unoberved farm heterogeneity bia the reult. We ue panel data and etimate fixed effet model whih help u to ontrol for the unoberved heterogeneity omponent that remain fixed over time thu reduing oniderably the omitted variable bia problem. In order to ontrol for the endogeneity we etimate the generalied method of moment (GMM) model. Fixed effet model The following fixed effet model etimation implie the following peifiation: (16) loan = β 0 + b j + β dds + β aaet + βπ Π + β x X + ε where b j i the fixed effet for farm j, whih apture time-unvarying farm-peifi harateriti. Thee fixed effet repreent farm heterogeneity. For example, they ould reflet different tehnologie for different farm, or they ould reflet different managerial kill or other unobervable fixed farm peifi harateriti. Endogeneity Three oure of endogeneity might bia our etimate. If ubidie were aigned to farm randomly, then parameter β d would meaure the impat of ubidie on bank loan. In reality, however, ubidie are not aigned randomly to farm. For example, the oupled animal and rop ubidie depend on regional and farm level produtivitie. To addre thi oure of endogeneity, we employ the Arellano and Bond (1991) robut two-tep GMM etimator. Arellano and Bond (1991) have hown that lagged endogenou variable are valid intrument in panel data etting. Thi allow u to ue lagged level of the endogenou variable a intrument (additionally to exogenou variable), after the equation ha been firt-differened to eliminate the farm peifi effet. The GMM etimator i partiularly uitable for dataet with a large number of ro-etion and few period and it require that there i no autoorrelation. The FADN dataet mathe thee requirement, beaue it i a panel data and ontain a very large number of farm-level obervation relative to the period overed. Given that the robut two-tep GMM tandard error an be everely downward biaed, we ue the Windmeijer (2005) bia-orreted robut variane. Data and variable ontrution The main oure of the data ued in the empirial analyi i the Farm Aountany Data Network (FADN), whih i ompiled and maintained by the European Commiion. The FADN i a European ytem of ample urvey that take plae eah year and ollet trutural and aountany data on the farm. To our knowledge, the FADN i the only oure of miro-eonomi data that i harmonied (the bookkeeping priniple are the ame aro all EU Member State) and it i repreentative of the ommerial 11 There i not available onitent data on the timing of CAP ubidie. 12 In perfet market ituation, the prie of an invetment good i the preent value of the future return from the invetment good whih tend to be higher than the prie of a variable input (e.g. fertilizer). The prie of variable input i determined by it annual marginal ontribution to the farm profitability. 7

8 agriultural holding in the whole EU. Holding are eleted to take part in the urvey on the bai of ampling plan etablihed at the level of eah region in the EU. The urvey doe not, however, over all the agriultural holding in the Union, but only thoe whih are of a ize allowing them to rank a ommerial holding. The FADN data i a panel dataet, whih mean that farm that tay in the panel in oneutive year an be traed over time uing a unique identifier. In thi tudy we ue panel data for overing all EU MS exept Romania and Bulgaria. Romania and Bulgaria were exluded from the ample, beaue for thee ountrie the data were available only for one year (2007). All variable exept for ratio are alulated per hetare of utilied agriultural area (UAA) in order to redue the potential problem of heterokedaitiy. The dependent variable in equation (16) total loan, long-term loan, hort-term loan are ontruted from the FADN data by dividing total, long-mediumterm and hort-term loan, repetively, with the total utilied agriultural area. Similarly, all ubidy variable (ub_total_ha, ub_deoupled_ha, ub_oupled_ha) are ontruted from the FADN data and are alulated on per-hetare bai. Every agriultural produer in the FADN urvey i aked to report both the total ubidie reeived a well a to peify the amount by major ubidy type. Deoupled ubidie, ub_deoupled_ha, inlude SPS and SAPS payment. Coupled ubidie, ub_oupled_ha, inlude payment linked to farm input or output uh rop area payment, animal payment and RDP. The total ubidie, ub_total_ha, variable i the um of oupled and deoupled CAP ubidie. The independent variable aet and inome_ha repreent the value of farm aet and farm ah flow alulated on per-hetare bai. The ovariate matrix X inlude other explanatory variable whih affet farm loan. The following ovariate were ued: land rented ratio, labour own ratio, farm ize, irrigated land, area under gla, fallow land, and etoral, regional and time dummie. Reult The reult are reported in Table 1 for total farm loan (olumn 1-3), for long-term farm loan (olumn 4-6) and for hort-term farm loan (olumn 7-9). 13 Additional to the omplete equation peifiation (16), we add an interation variable between ubidie and farm ize (model 2, 3, 5, 6, 8 and 9) and the quare value of ubidie (model 3, 6 and 9) to aount for indiret and nonlinear relationhip between ubidie and loan. The model-adjuted R 2 range from 0.10 to The mot onitently ignifiant variable (prob(t) < 0.10) aro all model are aet (aet_ha), trend variable (year), own labour ratio (labor_own_ratio), and rented land ratio (land_rented_ratio). The etimated reult ugget that ubidie influene farm loan but the effet are indiret and nonlinear. The oeffiient for ubidie in model 1, 4 and 7, where only a linear ubidy term i ued, are tatitially not ignifiant for all type of loan. However, when interating ubidie with farm ize (model 2 and 5) it oeffiient i tatitially ignifiant but only for total loan and for long-term loan. At the ame time, the oeffiient aoiated with the linear ubidy term ub_total_ha (the diret effet) i tatitially ignifiant and take a negative value. Thi indiate that ubidie timulate farm loan but only for larger farm, wherea the diret impat of ubidie ha a reduing effet on total and long-term loan (model 2 and 5). For the hort-term loan (model 8) both oeffiient (i.e. for the interation variable and the linear term ub_total_ha) are not ignifiant. Further, the reult indiate that the relationhip between ubidie and loan i non-linear. A mall value of ubidie per hetare redue bank loan (the oeffiient for ub_total_ha i negative and ignifiant in model 3 and 6) and a the value of ubidie inreae farm ue more bank loan (the oeffiient for the quared value of ubidie ub_total_ha_q i poitive and ignifiant in model 3 and 6). Again thi hold only for total loan and for long-term loan. The hort-term loan are not affeted by ubidie alo when the non-linear relationhip i onidered (model 9). Thee reult indiate that the hypothei 3 hold for the total and the long-term loan whereby ubidie inreae farm ollateral and thu farm loan ue. Multi-annuality harater of the long-term invetment allow the ue ubidie by redit ontraint farm to finane invetment only through loan. For the hort-term loan the etimated reult ugget the validity of the hypothei 2a. However, thi doe not imply that farm are not redit ontraint with repet to hort-term loan. Farm may till be redit 13 We etimate fixed effet model with heteroedatiity-onitent tandard error. 8

9 ontrained and may ue ubidie to finane hort-term input beaue either reeiving them at the beginning of the growing eaon or beaue they may ue other informal oure whih may be ubidy ollateralied. On the other hand, the differene in the tatitial ignifiane between the long-term and the hort-term loan may indiate that farm may prefer to ue ubidie to finane the long-term invetment and not the hort-term one poibly beaue of tronger redit ontraint preent in the former type of invetment than in the latter one. In Table 2 we diaggregate ubidie in oupled (ub_oupled_ha) and deoupled (ub_deoupled_ha) payment and again etimate their impat on total loan (olumn 1-3), long-term loan (olumn 4-6) and hort-term loan (olumn 7-9). The reult indiate important differene the two type of payment have on the farm loan. For the long-tem loan (model 4-6) the effet are imilar to thoe hown in Table 1 where long-term loan (model 4-6) were regreed over aggregated ubidie. Both oupled and deoupled ubidie have an indiret (by timulating farm more loan of big farm than of mall one) and non-linear impat on long-term loan. For the hort-term loan, the effet of diaggregated ubidie (Table 2) differ with repet to the reult reported in Table 1. The hort-term loan are affeted only by deoupled payment. However, the diret effet (the oeffiient for ub_deoupled_ha in model 9) i poitive and ignifiant, wherea the interation term (the oeffiient for ub_deoupled_fize in model 9) i negative and ignifiant. Thee reult ugget that the hort-term loan are ued a ollateral to inreae farm loan but thi i more important for mall farm than for big farm. The oupled payment do not affet the hort-term loan: i.e. the oeffiient for variable ub_oupled_ha, ub_oupled_ha_q and ub_oupled_fize are tatitially not ignifiant in model 9. The reult in Table 2 onfirm that the hypothei 3 tend to hold for the long-term loan for both type of CAP payment. For the hort-term loan only the deoupled payment imply the validity of the hypothei 3, wherea the etimated effet for the oupled payment ugget that the hypothei 2a may better repreent the reality. The GMM etimate are hown in Table 3. Generally, the GMM reult indiate different reult a ompared to the one reported for the fixed effet model. When ontrolling for the endogeneity, the importane of ubidie in determining both the long-term and the hort-term loan redue ignifiantly. Only the deoupled payment affet loan and the relationhip i non-linear. A mall value of ubidie doe not affet the loan (the oeffiient for ub_deoupled_ha and ub_oupled_ha are not ignifiant in model 2, 3 and 4, 6) and a the value of ubidie inreae farm ue more bank loan (the oeffiient for the quared value of ubidie ub_oupled_ha_q i poitive and ignifiant in model 3 and 6). Thi hold for both type of loan. Conluion In thi paper we etimate the impat the CAP ubidie on farm bank loan. Firt, we theoretially analye the farmer' farm loan demand under perfet and imperfet redit market aumption. In empirial analyi we ue the FADN farm level panel data to tet the theoretial predition. Aording to the theoretial reult, ubidie may inreae bank loan, redue them or have no impat on bank loan depending on whether farm are redit ontrained, whether ubidie are alloated at the beginning or at the end of the growing eaon, and on the relative ot of internal and external finaning. We employ the fixed effet and GMM model to etimate the impat of ubidie on farm loan. The etimated reult ugget the following impat of ubidie on farm loan ue: (i) Subidie influene farm loan and the effet tend to be non-linear and indiret. (ii) Coupled ubidie affet hort and long term loan differently than deoupled ubidie. (iii) Both oupled and deoupled ubidie timulate long-term farm loan. But long-term loan of big farm inreae more than thoe of mall farm due to deoupled ubidie. (iv) Short-term loan are affeted only by deoupled ubidie. However, deoupled ubidie inreae hort-term loan of mall farm more than thoe of large farm. (v) When ontrolling for the endogeneity, the importane of ubidie in determining both the long-term and the hort-term loan redue ignifiantly. Only the deoupled payment affet loan and the relationhip i non-linear. (vi) In general our empirial reult indiate that the hypothei 3 hold for the deoupled payment, wherea oupled payment are found not to affet loan. Referene 9

10 Barro, R. (1974). Are Government Bond Net Wealth? Journal of Politial Eonomy 82(6): Blanard, S., J.P. Bouemart, W. Brie, and. erten Short- and Long-Run Credit Contraint in Frenh Agriulture: A Diretional Ditane Funtion Framework Uing Expenditure-Contrained Profit Funtion. Amerian Journal of Agriultural Eonomi 88: Beer Van Cee, J. and C.J.M. Van Den Bergh (2001). "Pereverane of Pervere Subidie and their Iimpat on Trade and Environment." Eologial Eonomi 36: Carter, M.R Equilibrium Credit Rationing of Small Farm Agriulture. Journal of Development Eonomi 28: Chau, N.H., and H. de Gorter Dientangling the Conequene of Diret Payment Sheme in Agriulture on Fixed Cot, Exit Deiion, and Output. Amerian Journal of Agriultural Eonomi. 87: Ciaian. P. and J.F.M. Swinnen (2009). "Credit Market Imperfetion and the Ditribution of Poliy Rent." Amerian Journal of Agriultural Eonomi 91(4): de Gorter, H., and.d. Meilke Effiieny of Alternative Poliie for the EC' Common Agriultural Poliy. Amerian Journal of Agriultural Eonomi 71: de Gorter, H Aeing the Tranfer Effiieny of Agriultural Support Poliie. OECD, Pari. Feder, G The Relation between Farm Size and Farm Produtivity - The Role of Family Labour, Superviion and Credit Contraint. Journal of Development Eonomi 18: hanna, M., M. Iik and D. Zilberman (2002). "Cot-Effetivene of Alternative Green Payment Poliie for Conervation Tehnology Adoption With Heterogeneou Land Quality." Agriultural Eonomi 27: Latruffe, L The Impat of Credit Market Imperfetion on Farm Invetment in Poland. Pot- Communit Eonomie 17: Maitra, P., and R. Ray (2003). The Effet of Tranfer on Houehold Expenditure Pattern and Poverty in South Afria. Journal of Development Eonomi 71(1): Figure 1. Farm fertilizer ue with and without redit ontraint k S DS k S k A a D k 10

11 Figure 2. Credit ontraint and ubidie with α = 1 k S S 1 S 2 k DS 1 DS 2 k k A B D E d e F k a b D Figure 3. Credit ontraint and ubidie with α = 0 k S S 1 S 2 k A a b B D D 1 11

12 Table 1. Fixed effet etimate of loan (total ubidie) Total loan Long-term loan Short-term loan Model 1 Model 2 Model 3 Model 4 Model 5 Model 6 Model 7 Model 8 Model 9 ub_total_ha ub_total_ha_q e-07 ub_total_fize aet_ha inome_net_ha inome_net_ha_l Contant -54,495-54,088-55,396-44,250-43,310-45,192 15,011 15,091 15,098 Obervation R-quared Number of idn Robut tandard error in parenthee; p<0.01, p<0.05, p<0.1; The etimated oeffiient for the ontrol variable X are not reported but are inluded in the regreion.. Table 2. Fixed effet etimate of loan (diaggregated ubidie) Total loan Long-term loan Short-term loan Model 1 Model 2 Model 3 Model 4 Model 5 Model 6 Model 7 Model 8 Model 9 ub_deoupled_ha ub_deoupled_ha_q e-05 ub_deoupled_fize ub_oupled_ha ub_oupled_ha_q e-06 ub_oupled_fize aet_ha Obervation R-quared Number of idn Robut tandard error in parenthee; p<0.01, p<0.05, p<0.1; The etimated oeffiient for the ontrol variable X are not reported but are inluded in the regreion. Table 3. Arellano and Bond etimate of loan (diaggregated ubidie) Long-term loan Model 1 Model 2 Model 3 Short-term loan Model 4 Model 5 Model 6 ub_deoupled_ha ub_oupled_ha ub_deoupled_fize ub_oupled_fize ub_deoupled_ha_q ub_oupled_ha_q aet_ha inome_net_ha invetment_ha L.invetment_ha Contant -1, Obervation Number of idn Standard error in parenthee; p<0.01, p<0.05, p<0.1; The etimated oeffiient for the ontrol variable X are not reported but are inluded in the regreion. 12