Farm investment, credit rationing, and governmentally promoted credit access in Poland: a cross-sectional analysis

Food Policy 29 (2004) 275–294 www.elsevier.com/locate/foodpol

Farm investment, credit rationing, and governmentally promoted credit access in Poland: a cross-sectional analysis Martin Petrick
Institute of Agricultural Development in Central and Eastern Europe (IAMO), Theodor-Lieser-Strasse 2, 06120 Halle (Saale), Germany

Abstract The aim of this paper is to empirically analyse the effects of governmentally promoted credit access on the investment behaviour of credit-rationed farmers in Poland. This is done by an econometric analysis of cross-sectional Polish farm household data. A Probit analysis revealed that the reputation of the borrower, but not the availability of land as collateral, had an effect of credit rationing. The estimates of an investment equation suggest that access to subsidised credit has a statistically significant role in determining investment behaviour of farmers. In various specifications of the credit-investment relationship, the average marginal effect of credit on investment was smaller than one, which implies that credit is partly used for purposes other than productive investment. Furthermore, investment volume is negatively related to farm size. A governmental policy which aims to promote productive investment should emphasise lending in larger amounts without discriminating against small farms. # 2004 Elsevier Ltd. All rights reserved. Keywords: Farm investment; Credit rationing; Credit policy; Microeconometrics; Poland

Introduction Poland’s accession to the European Union (EU) has recently stimulated public interest in the country’s agricultural sector, which is believed to pose substantial adjustment problems over the course of accession. Compared to farms in the former EU-15 member countries, per capita incomes from Polish agricultural production


Tel.: +49-345-2928-127; fax: +49-345-2928-199. E-mail address: [email protected] (M. Petrick).

0306-9192/$ - see front matter # 2004 Elsevier Ltd. All rights reserved. doi:10.1016/j.foodpol.2004.05.002

276

M. Petrick / Food Policy 29 (2004) 275–294

are quite low. Likewise, there is a noteworthy gap between rural and urban incomes within Poland (Petrick et al., 2002). A major reason for this is severe structural deficiencies in the farm sector, most notably an unfavourable man/land ratio and low labour productivity levels, which call into question the sector’s international competitiveness. To foster modernisation and structural change in agriculture, the Polish government launched a voluminous farm credit programme in 1994, which mainly encompasses interest subsidies granted on operational and investment loans extended by commercial banks (Christensen and Lacroix, 1997). In 1999, the last year under investigation in this study, subsidies on investment loans amounted to 771 million zloty (zl) (approximately 182 million euro; OECD 2000, pp. 106–107). Excluding expenses for the farmers’ social insurance fund, subsidy payments on all types of loans made up 38 percent of the budget of the Ministry of Agriculture and Rural Development (Ministerstwo Rolnictwa i Rozwoju Wsi, MRiRW; MRiRW 2000). Intervention on credit markets can thus be regarded as a major instrument of the Polish government to achieve its political objectives. The aim of this paper is to empirically analyse the effects of governmentally promoted credit access on the investment behaviour of credit-rationed farmers. More specifically, the following questions will be addressed: (a) whether there is any significant credit rationing despite government intervention and what are its determinants, and (b) whether subsidised funds are in fact used for productive investment in the farm sector or are diverted to other purposes. The methodological approach consists of a two-stage procedure. In the first stage, qualitative information on credit rationing of farmers based on a Probit equation is analysed. The latter was estimated on a cross-sectional sample of farm household data. In the second stage, an empirical investment equation for the credit-rationed (or credit-constrained) subgroup of farmers was estimated. The econometric analysis was used to estimate the marginal effect of credit on investment, which in turn provided the key information for policy evaluation. A marginal effect of credit on investment larger than one implies that subsidised credit is fully used for investment and even triggers the additional mobilisation of other (particularly own) financial sources, which is clearly desirable from the point of view of the government. On the other hand, a marginal effect smaller than one implies that the marginal unit of credit is only partly used for the supposed investment purpose. The paper proceeds in several steps. In the first section, the study is motivated by explaining the Polish policy background. The theoretical framework of the study is then briefly outlined and the data base used for the estimation is introduced. The fourth section describes the empirical approach in more detail and discusses the major methodological problems of the analysis. After that, the econometric results are presented, and the final section summarises the findings and concludes with some policy implications.

M. Petrick / Food Policy 29 (2004) 275–294

277

Government intervention in Polish agricultural credit markets Overview and economic significance The major form of state intervention in rural credit markets in Poland is the extension of preferential loans to agricultural producers. Borrowers pay only a part of the commercial interest rate, whereas the remainder is paid by the government. Hence, there is a subsidy on interest rates. A second form of intervention is by loan guarantees. However, the budgetary importance of the latter is small as compared with the subsidies (totalling less than one tenth). Focus was therefore placed only on the interest subsidy programme. Since 1994, preferential credits have been handed out by the Agency for Restructuring and Modernisation of Agriculture (Agencja Restrukturyzacji i Modernizacji Rolnictwa; ARiMR), which in the following years provided more than 30 different credit lines for various purposes (Czerwin´ska-Kayzer, 2000, p. 9). These credit lines comprise loans for the purchase of inputs, basic investment, land purchases, investments by young farmers, sector programs (milk, cattle, poultry, etc.), and others (Christensen and Lacroix, 1997, p. 18). The different credit lines are grouped into the two major categories ‘‘working capital’’ and ‘‘investment’’. Interest rates vary between credit lines. For each credit line, ARiMR establishes a maximum rate that a bank can charge, which is a multiple of the bank’s refinance rate. Borrowers then pay one quarter to one half of this maximum rate, according to the credit line, and ARiMR pays the rest (Christensen and Lacroix, 1997, p. 19). Fig. 1 depicts the outstanding amounts of total and preferential credits in the agricultural sector between 1993 and 2002. Monetary values are given in 1999 prices. The foundation of ARiMR marked the start of a phase of rapid credit expansion, with growth rates of the preferential credit volume of almost 60 percent in 1995 and 1996. In 1997, the volume of subsidised credits reached its peak and declined in the following years. This is consistent with the fact that the number of credit lines for agriculture and the volume of public funds earmarked for subsidising interest rates were considerably cut in 1998 (Czerwin´ska-Kayzer, 2000, p. 12). Since 2000, the volume of preferential credits has been almost stable in real terms. In the phase of credit expansion, the share of preferential credits in the total credit volume temporarily increased from 53.7 percent in 1994 to 85.9 percent in 1997, thereafter it decreased. Assuming there is a given amount of projects also viable under non-subsidised rates, this is evidence of a crowding-out effect, which means that borrowers turned to the cheaper government loans although they would have also borrowed under fully commercial terms. However, it seems that the total amount of outstanding credit was mainly driven by changes in the governmentally sponsored credit supply. In the first half of 1998, at the peak of intervention, preferential interest rates ranged between 6.13 and 15.31 percent per year. In the same period, the inflation rate was at 13.7 percent, and the difference between subsidised and non-subsidised interest rates ranged between 17 and 25 percentage points. Interest subsidies hence

278

M. Petrick / Food Policy 29 (2004) 275–294

Fig. 1. Outstanding volume of total and preferential loans in the agricultural sector 1993–2002 (in 1999 prices). Credit volume as outstanding on 31 December. Monetary values in 1999 prices, using the consumer price index. 1 zl ¼ 0:237 euro in 1999. Source: Author’s calculations based on Kulawik (2003, p. 77) and unpublished data of National Bank of Poland.

led to a substantial reduction of interest costs for farmers, even implying negative real interest rates (figures taken from Poganietz and Wildermuth, 1999, p. 537). Preferential loans under the government programme are extended through the existing network of banks. In Poland, there are two types of lending organisations which specialise in agriculture, namely the Bank for Food Economy (Bank Gospo˙ ywnos´ciowej, BGZ˙), and the system of cooperative banks (Klank, 1999). darki Z However, preferential credits can also be received via most of the commercial banks in Poland. The BGZ˙ was the primary channel for financing state-managed agriculture during the socialist period, which implied that the bank inherited quite a number of bad loans during the course of market reforms. Similar to other formerly state-owned banks in Poland, there were several attempts to comprehensively restructure or liquidate the BGZ˙ during the past decade. However, this was successfully blocked, inter alia by agricultural lobby groups. Local cooperative banks had often been founded prior to World War II, and existed under the umbrella of the BGZ˙ during socialism. In 1990, most of them left the BGZ˙ in order to form regionally-oriented cooperative banking structures. Even so, to date, their reconsolidation has remained incomplete. Furthermore, Khitarishvili (2000) provides evidence based on a stochastic frontier analysis that the efficiency of Polish cooperative banks lags behind international standards. Whereas the general privatisation and liberalisation activities in the Polish banking sector have proven largely successful, agricultural banking is still an exception. Milczarek (2003) argues that the banks which already existed under socialism are neither particularly innovative nor supportive to entrepreneurs, but adhere to their traditional role of simply channelling certain amounts of liquidity into the sector. These attitudes are

M. Petrick / Food Policy 29 (2004) 275–294

279

assumed to continue to impede the development of an efficient and professional rural banking sector in Poland. Prospective borrowers have to submit a loan application at a local bank branch, together with a business plan describing the envisaged use of the loan. The latter is usually evaluated by the public extension service prior to loan application. The bank then applies for subsidy payments at ARiMR. The bank bears the full default risk of the loan and therefore is also responsible for screening and monitoring borrowers, as well as possible enforcement of repayment or liquidation of collateral (Poganietz and Wildermuth, 1999, p. 539). In contrast to other transition countries, mortgaging loans are less of a problem because most of the land remained as private property during the period of socialism. Accordingly, mortgaging is currently a commonly used instrument to collateralise loans (Prosterman and Rolfes, 2000, pp. 128–129). However, as stressed by Karcz (1998, p. 96), the reliability or reputation of a borrower as indicated by previous punctual repayment of loans is at least as important for obtaining credit as is the sufficient availability of collateral. In general, default rates in rural Poland are quite small, about two percent according to Karcz (1998). Delayed payments are relevant in markedly less than ten percent of the cases (World Bank, 2001, p. 74). Policy questions Given this massive form of government intervention in agricultural credit markets in Poland, the following questions are of particular interest for an assessment of the programme: 1. Are all targeted farmers in fact able to borrow the subsidised funds? 2. Are the funds used in a way that tackles the structural deficits of the sector and therefore justifies the governmental subsidy? The first question addresses the problem of credit rationing, or the fact that borrowers cannot obtain as much credit as desired although they have appropriate investment projects available. The reason for this might be difficulties in overcoming asymmetric information between borrowers and lenders, known in the theoretical literature as adverse selection or moral hazard (see Swinnen and Gow, 1999, for an overview). Common devices to overcome these problems are the provision of collateral by the borrower and a careful screening procedure by the lender. Swinnen and Gow (1999, p. 39) cite the case of Bulgaria where credit subsidies had little effect on the flow of credit into the agricultural sector because of lacking collateral on the side of the borrowers. With regard to Poland, the issue should therefore be addressed to what extent farmers remain credit-rationed despite governmental subsidies, and which causes of rationing can be identified. The above-noted lack of efficiency and innovativeness of agricultural banks in Poland suggests that credit rationing might be a relevant problem. This was confirmed by a recent World Bank study (2001, p. 66), according to which almost half of the borrowers in a farm survey said they wanted a larger loan on the same terms as the

280

M. Petrick / Food Policy 29 (2004) 275–294

loan they received. On the other hand, the availability of mortgaging in Poland, as well as the substantial increase in credit flows depicted in Fig. 1, suggest the opposite. The second question highlights the problem of credit diversion, as introduced by Von Pischke and Adams (1980) in the context of developing countries. Due to the fungibility of credit and the difficulties in supervising thousands of rural borrowers, funds may frequently be used for purposes other than those intended by the government programme. In particular, credit may be used for consumptive rather than productive activities, thereby tending to follow the most attractive use available from the perspective of the loan recipient. Indeed, Poganietz and Wildermuth (1999, p. 540) provide some preliminary evidence that subsidised investment funds in Poland were in fact diverted to non-productive purposes, for example for renovating residential buildings and their technical infrastructure. On the other hand, the author demonstrates elsewhere that working capital loans could have been used quite productively on credit-rationed farms in Poland (Petrick, 2004a). The subsequent empirical analysis attempts to shed light on both questions by conducting an empirical analysis of farm level survey data, whereby the analysis concentrates on the use of investment loans with a repayment period of more than twelve months. In the following, some brief theoretical considerations should illuminate the analytical framework.

Theoretical considerations In line with the policy motivation of this paper is the assumption that financial constraints limit the capital accumulation of farms. A dynamic theory of the financially-constrained (farm) household has been formally elaborated by Steigum (1983) and Chambers and Lopez (1984), and is taken as a theoretical background for the present paper. These dynamic farm household models have the following implications: (a) limited access to credit causes a lagged adjustment of capital stocks to the steady state, (b) optimal investment is dependent on the equity formation of the household in terms of profit retention or savings, or, more general, on the availability of collateral, and (c) investment and credit demand are thus neither separable from consumption decisions nor independent of the equity position of the farm. More recent work on stochastic investment models arrives at similar implications (Hubbard and Kashyap, 1992). It should be added that providing equity or collateral is not the only way to improve credit access; similar effects can result from reputation acquisition (Diamond, 1989).

Data The analyses in this paper are based on data collected during the ‘IAMO Poland farm survey 2000’, which was a cross-sectional farm survey conducted in the boundaries of the former Szczecin, Tarno´w, and Rzeszo´w voivodships existing

M. Petrick / Food Policy 29 (2004) 275–294

281

prior to the recent administrative reform. The survey was carried out in 2000 in the form of face-to-face interviews and contains mainly data related to the economic outcomes of the year 1999. Investment and credit data is available for 1997–1999. The survey is based on a random sample of farms in the records of the official extension service. The records consist only of farms that show at least some degree of commercialisation and market integration; these account for the bulk of the traded agricultural produce in the research area. The sample consists of 464 farms; 120 from Szczecin, 108 from Tarno´w, and 236 from Rzeszo´w. Within the given geographic boundaries of the three voivodships, the sample is stratified in one stage. The strata are identical with administrative districts (powiat). Further details on sampling issues, the organisation of data collection and a reprint of the questionnaire can be found in Petrick (2001).

Methodological approach Probit model To analyse the rationing status of farmers, a qualitative approach based on directly asking the respondent about his borrowing experience (following Jappelli, 1990) was employed. Bank applicants were asked whether during their most recent loan application they would have liked to borrow more at the same interest rate. If so, this was taken as evidence for an excess demand, and the respondents were classified as being partially credit-rationed. Applicants who did not obtain a loan at all were classified as completely rejected. Non-applicants were asked whether they had the intention of applying for credit at a particular place in the past but did not do so because the application might have been turned down. Respondents who answered positively were classified as discouraged non-borrowers. All these groups of respondents were regarded as being credit-constrained, which implies that their access to credit was exogenously determined and hence not under the control of the decision-maker.1 A shortcoming of this approach is that it relies only on the individuals subjective assessment of their situation. It is a relatively crude measure that provides no objective information concerning the actual profitability of available on-farm uses of credit.2 However, it is relatively easy to collect but still lends itself to multivariate methods of analysis. Eq. (1) demonstrates how this qualitative information can be used to analyse the determinants of credit rationing. k
¼ c0 z þ u

ð1Þ

ki
is a latent variable denoting an excess credit demand. We observe ki, which is a dichotomous (1, 0) variable indicating whether the respondent falls into the 1

At this stage, the analysis did not distinguish between short-term and long-term loans. Among the first applications to credit rationing of farms was Feder et al. (1990). For a comparison with other methods for measuring credit rationing see Petrick (2004b). 2

282

M. Petrick / Food Policy 29 (2004) 275–294

credit-rationed subgroup or not. z represents a vector of explanatory variables (such as household and production characteristics). c is a vector of parameters, while u is a random error term. The choice of farm and household characteristics was based on the following considerations. In a recent case study on banks’ loan approval procedures in rural Poland, Latruffe (2003) found that collateral availability is crucial for obtaining credit. However, according to this study, banks in Poland rarely appropriate collateral in the case of loan default because of costly and uncertain legal enforcement. Instead, they prefer to sift out applicants in order to make default as unlikely as possible in advance. To test the importance of collateral, the value of land owned (expected sign: –) was included as an indicator of initial wealth of the household.3 The nominal value of land owned by the farm at the beginning of the investment period (in thousand zl) was calculated by subtracting land investment carried out in the period 1997–1999 from the stated value of owned land in 1999. Land quality was hence captured as well, at least as long as it is reflected in monetary land values. To test the empirical relevance of non-collateral factors, a dummy indicating a previously rescheduled loan (+) was taken to represent the credit history of the borrower, and a dummy indicating the expressed habit of regularly engaging in conversations with neighbours () was used as a measure of villageinternal information flow. Respondents were asked whether they had rescheduled the repayment of another loan taken earlier in the reporting period. This was regarded as evidence of a borrower’s relatively poor reputation. Intensive intra-village communication might reduce the probability of being credit-rationed if it increases the information available to the local bank. It was taken as a proxy for how well the respondent was known in the village. Liquidity shortages may also be due to the consumption behaviour of a farm household. The absolute number of adult males and females were therefore taken to reflect household characteristics. Under credit rationing, theory suggests that household characteristics do have an impact on credit demand, but the direction is an empirical question. The effect of the number of adults is indeterminate since more household members may both increase (via increased consumption) and decrease (via the generation of off-farm or unearned income) the liquidity shortage. The separate inclusion of males and females is motivated primarily by the fact that Polish women tend to benefit more from social transfer payments than men (World Bank, 1995, p. 117).4 In addition, the number of males or females may take on a signalling function for the bank. For example, a higher number of men in the households’ labour force may indicate that more resources are devoted to farm production as opposed to household work, and hence may imply a higher cred3

In the survey data, land was the most commonly used form of collateral, ranging before machinery and buildings. Furthermore, Petrick (2004a) presented evidence that land as collateral plays a role in short-term lending. Land titling has a long tradition in Poland and there was no land restitution during transition, two factors that considerably ease mortgaging. See Prosterman and Rolfes (2000) as cited above. 4 Note that women have an earlier retirement age than men and a higher life expectancy.

M. Petrick / Food Policy 29 (2004) 275–294

283

itworthiness. Finally, two dummies indicating the year in which the loan was approved by the bank were added to the model. Other intuitively plausible variables were tested as well but turned out to have no significant impact on the rationing outcome. These included the age and age squared of head of household and his educational level (on a scale from 0–5). Variables indicating whether the farm specialised in crop or livestock production did not show significant coefficients because most farms had a mixed production programme with a low degree of specialisation. Since the data covers a three-year period, farm size in terms of hectares under cultivation or livestock units was considered endogenous to the model and therefore not included in the estimating equation. All explanatory variables were assumed to be exogenous or predetermined at the time of loan application. Investment equation In the second stage, an investment equation was estimated on the subsample of credit-rationed respondents as identified above. The first-stage Probit results were used to test for selectivity, as will be explained shortly. The investment analysis was based on a potentially non-linear, reduced-form investment equation of the following type: I ¼ IðK; Z; fÞ þ e:

ð2Þ

In this equation, I denotes the investment volume, K the amount of long-term credit acquired, Z the existing capital stock or, more generally, the initial farm size, f a vector of dummies capturing regional and farm-specific effects, and e is a random error term. There are two important peculiarities compared with conventional neo-classical investment equations (Elhorst, 1993, p. 170). First, the equation contains a financial variable, K. This is due to the assumed relevance of the financial constraint, as explained earlier. Second, there are neither user costs of capital nor output or input prices included in the equation. This is due to the fact that the investment equation is estimated on a cross-sectional data set, so that prices are assumed to be equal for all farms and hence excluded (similar to Feder et al., 1992). The following signs of the parameters to be estimated are expected. Under the assumption that profitable investment projects are available, the relation between credit volume and investment is unambiguously positive. The effect of Z on I depends on the size of the desired capital stock or farm size. A negative sign implies that farm sizes converge over time, whereas a positive sign implies diverging farm sizes. f includes a dummy indicating whether the farm has permanent book-keeping, which is taken as a measure of the farmer’s management skills. It is likely that more skilled farmers invest more. A second dummy has the value of one if the farm is located in the north of the three regions under investigation. Since this region is supposed to be a more favourable economic environment for agricultural production, the effect on investment is also likely to be positive.

284

M. Petrick / Food Policy 29 (2004) 275–294

Eq. (2) can be used to determine the marginal effect of credit on farm investment, which is given as the first-order partial derivative with regard to K. There are three methodological problems related to the estimation of eq. (2), namely (a) endogeneity of regressors, particularly with regard to the credit access variable K, which will usually be regarded as a choice variable of the household, (b) censoring of the dependent variable, since approximately 20 percent of households considered in the estimation report a zero investment volume, and (c) choice of the functional form, to allow the marginal credit effect to vary with loan size. The problem of endogeneity was addressed by using specific qualitative information on credit access provided by the survey results as follows. The investment equation was estimated only on the subsample of credit-rationed respondents. The cost of this sample-splitting procedure was that selectivity might have introduced a bias into the estimates. This was tested by a two-stage procedure due to Heckman (1979). The estimating equation became: I ¼ IðK; Z; fÞ þ e iff c0 z þ u > 0

ð3Þ

In this system, e and u were supposed to have a bivariate normal distribution with zero means and a given correlation. To address the censoring problem, a Tobit model (Greene, 2000, pp. 905–926) was used in the estimation of the investment equation.5 In the standard Tobit formulation, the level of investment is still explained by a model that is linear in parameters. However, by including higherorder polynomials into the equation, the model can be made more flexible. In an additional specification, the Tobit equation was hence augmented by a quadratic and a cubic term for the credit variable. The virtue of this procedure was that a cubic function for the uncensored part of the investment equation was obtained, which was then used for the further analysis of the credit-investment relationship. The dependent variable, investment volume I in thousand zl, is the aggregate of all productive investment made from 1997–1999, including land, all types of agricultural machinery, farm buildings, livestock, and permanent crops, to name the most important. Only gross investment was considered here, to avoid the difficult choice of a depreciation rate. Credit access K was measured as the total volume of credit with a repayment term of more than 12 months borrowed by the farmer in the period between 1997–1999. There were 81 single loan contracts reported amongst the 156 farmers in the subsample under investigation. 78 percent of the borrowers obtained at least one loan under the government program. The capital 5 Conventional marginal effects in the Tobit model vary with different values for the regressors. They give the total effect of a change in explanatories on the observed, censored investment volume. The latter effect can be decomposed into an effect on the conditional mean (and thus the size) of investment plus an effect on the probability that the farm invests at all. The implicit assumption of the Tobit model is that the given regressors explain both effects. In the present study, the uncensored part of the model is the relationship of interest, whereas the qualitative choice whether to invest or not is not analysed more deeply. Regarding the zeros as implying unobserved disinvestment might provide a rationale for this approach. In this case, the marginal effects are given by the coefficients of the Tobit model (Johnston and DiNardo, 1997, pp. 436–439).

M. Petrick / Food Policy 29 (2004) 275–294

285

stock Z at the outset of the investment period was difficult to measure since detailed statements of farm assets were only available for 1999. Therefore, the nominal value of land owned by the farm in the beginning of the investment period was used, as in the Probit model. The advantage of using land as compared to other assets was that the problem of depreciation could reasonably be ignored.

Results An evaluation of the qualitative information on credit access showed that 45.2 percent of all farmers must be regarded as being exogenously credit-rationed.6 The pool of constrained respondents was used for the estimation of the investment equation. Only 79.5 percent of these respondents reported positive investment, and only 44.1 percent took long-term loans. Subsequently, the results of the Probit analysis and the empirical investment equation are presented. Probability of being credit-rationed Basic statistics for the variables used in the Probit model are described in Table 1, net of drop outs due to data cleaning and missing values. The regression results are shown in Table 2. Somewhat surprisingly, the coefficient of total land owned beginning of 1997 was not statistically significant. If the volume of available collateral was appropriately measured, it can be concluded that it is of less importance in the general observation of credit rationing. An explanation could be that the actual rationing decision of Polish banks is based on other types of collateral, such as machinery or buildings (Latruffe, 2003). This was confirmed by Petrick (2004a), who found that farmers with older machinery had more difficulties in obtaining short-term loans. However, the relevance of land as collateral is not supported by the present study. Unfortunately, due to lack of data for the time period covered in the present study, measures of available collateral other than land could not be included separately into the regression. On the other hand, the reputation effect as measured by the previous loan rescheduled dummy was statistically significant, with a t-value of almost three, which underlines the banks’ interest in extending credit only to reliable clients. The positive coefficient on the dummy denoting intra-village information flow is contrary to expectations. A possible explanation is that better information about the farming activities of a given borrower conveyed the impression to the bank that this borrower was in fact not creditworthy, so that he received less credit than he applied for. 6 The majority of farmers were partially credit-rationed, which means that they obtained some credit but not as much as desired. It is therefore unlikely that the depletion of government funds is the reason for credit rationing, since loans are extended according to the first-come-first-served principle. Rationing due to exhaustion of government funds would therefore have taken the form of complete rejection of borrowers (Poganietz and Wildermuth, 1999, p. 539).

286

M. Petrick / Food Policy 29 (2004) 275–294

Table 1 Description of variables used in the Probit model Mean Credit-rationed (dummy) Total land owned beginning of 1997 (thousand zl) Adult males in household (no.) Adult females in household (no.) Previous loan rescheduled (dummy) Conversation with neighbour (dummy) Year of loan approval = 1997 (dummy) Year of loan approval = 1998 (dummy)

0.5 65.2 1.7 1.7 0.1 0.7 0.3 0.2

Std. dev.

Minimum Maximum

88.6

0.0

600.0

0.9 0.9

0.0 0.0

5.0 5.0

Valid observations 345 345 345 345 345 345 345 345

Source: Author’s calculations.

Regarding household characteristics, the coefficient of the number of males is statistically significant at less than one percent. Apparently, more women in the farm household tend to tighten the credit constraint, which is in contrast to the conjecture that higher public transfer payments for females increase available liquidity. Alternatively, more women in the household could make the farm less creditworthy due to a lower share of labour devoted to farm production as opposed to household work. The reverse holds for men, but is less statistically significant. The available evidence therefore suggests that rationing by Polish rural banks is primarily based on the reputation of the borrower and individual household characteristics. Whereas other studies found that the availability of different types of Table 2 Probit estimates of the probability of being credit-rationed

Constant Total land owned beginning of 1997 (thousand zl) Adult males in household (no.) Adult females in household (no.) Previous loan rescheduled (dummy) Conversation with neighbour (dummy) Applied in 1997 (dummy) Applied in 1998 (dummy) LR test (v27 ) (P value) Predicted 1s that were actual 1s Predicted 0s that were actual 0s McFadden’s R2 Observations

Coefficient

t-value

0.272 <0.001 0.217 0.149 0.737 0.448 0.258 0.086

1.163 – 0.300 0.010 2.652 8.581 1.813 5.894 2.998 29.151 2.850 17.711 1.691 10.199 0.453 3.402 26.985 (<0.001) 57.4% 62.0% 0.057 345

Source: Author’s calculations. a Marginal effects in percentage points, calculated at sample means.

Marginal effecta

M. Petrick / Food Policy 29 (2004) 275–294

287

collateral—including land and farm machinery—had an effect on rationing outcomes, the present work implies that the amount of owned land is not a significant determinant of credit rationing. The regression also reveals that applying for a loan in 1997 significantly reduced the probability of being credit-rationed. This finding is quite in line with the fact that governmentally subsidised credit expansion in the farm sector showed a clear peak that year, as noted above. The null hypothesis that all slopes of the model are zero as represented by the chi-squared statistic is clearly rejected. The percentage of correctly predicted outcomes reveals a fairly satisfactory predictive power of the model. The McFadden statistic is low but still comparable to similar studies in the literature. The marginal effects display the slope of the probability function. At sample means, the subgroup of respondents who rescheduled a loan in the past were 30 percentage points more likely to be credit-rationed than the subgroup with a better record. Reputation thus plays a key role in determining farm households’ credit access. Investment equation The characteristics of the subsample used for estimating the investment analysis are illustrated by the descriptive statistics in Table 3. Values are displayed separately for all constrained respondents and for constrained non-zero investors. The results of the estimations are presented in Table 4. The results for three specifications are reported. Mainly for purposes of exploration and comparison, a linear investment equation (I) was estimated by Ordinary Least Squares (OLS). This equation was also used for testing selectivity and was therefore estimated in a two-stage procedure together with the above Probit equation (Heckman, 1979;

Table 3 Description of variables used in the investment model (constrained subsample)

All constrained respondents Investment volume 1997–1999 (thousand zl) Credit volume 1997–1999 (thousand zl) Land owned beginning of 1997 (thousand zl) Farm has permanent book-keeping (dummy) Farm is located in northern region (dummy) Respondents with positive investment Investment volume 1997–1999 (thousand zl) Credit volume 1997–1999 (thousand zl) Land owned beginning of 1997 (thousand zl) Farm has permanent book-keeping (dummy) Farm is located in northern region (dummy) Source: Author’s calculations.

Mean

Std. dev.

Minimum Maximum Valid observations

24.8 20.6 65.7 0.4 0.3

42.4 43.2 97.3

0.0 0.0 0.0

322.5 400.0 600.0

156 156 156 156 156

31.2 24.8 60.2 0.5 0.3

45.5 47.4 88.4

0.0 0.0 0.0

322.5 400.0 520.0

124 124 124 124 124

288

M. Petrick / Food Policy 29 (2004) 275–294

Table 4 Estimated investment equationsa

Constant Credit volume 1997–1999 (thousand zl) Credit ^ 2 Credit ^ 3 Land owned beginning of 1997 (thousand zl) Farm has permanent book-keeping (dummy) Farm is located in northern region (dummy) Inverse Mills Ratio Log-Likelihood Adjusted R2 N

Linear (I)

Standard Tobit (II)

Coefficient

Coefficient

3.478 (0.437) 0.739

0.782 (0.229) 0.783

(14.237) –

(13.161) –

Cubic Tobit (III)

Marginal effectb – 0.602

Coefficient 1.558 (0.442) 0.418

0.063

(2.031) 0.003 (1.392) <0.001 (1.120) 0.100

(2.678) 10.597

8.145

(3.131) 14.195

(1.853) 17.267

(2.087) 20.909

16.070

(2.684) 23.387

(3.393) 1.521 (0.162) 714.319 0.666 156

(3.422) –

–

–

0.051

0.082

(2.139) 7.868

– –

–

610.996 – 156

(3.813) – 608.950 – 156

Source: Author’s calculations. a t-values in parentheses. t-values of the linear model corrected for selectivity. b Marginal effects calculated at sample means of selected observations.

Johnston and DiNardo, 1997, pp. 447–450). The second specification is a conventional Tobit model (II), to account for the censoring of the investment variable. Marginal effects at sample means are reported separately. The third specification is the cubic Tobit model (III), which includes the credit variable in quadratic and cubic form. For this model, marginal effects are analysed separately below. A number of important conclusions can be derived from the linear model. First, in contrast to received neo-classical thinking, the financial variable does have an effect on investment outcomes. This is consistent with the self-classification of borrowers as being credit-rationed. Credit access is of overwhelming importance in the linear model (as measured by the t-value). The coefficients of all other regressors have the expected signs and are statistically significant, at least at the 10 percent level. The coefficient of the land variable is statistically significant at 5 percent, indicating that farms with fewer assets at the outset invest more. The coefficient of the Inverse Mills Ratio, which tests the impact of selectivity, fails to be statistically

M. Petrick / Food Policy 29 (2004) 275–294

289

significant: it can hence be assumed that there is no selectivity bias in the equation. For this reason, selectivity was ignored in all other estimations. A further conclusion concerns the marginal effect of credit on investment in the linear model. The effect (0.739) was smaller than one, which points to an under-utilisation of credit for investment purposes. This is in accordance with the farmers’ reported use of credit funds for ‘durable consumption goods’, for example renovating residential buildings or automobile purchases.7 For only 50 percent of borrowers does the amount of productive investment exceed the credit volume. However, the linear model imposes that the marginal effect is constant over the entire range of observations, as noted earlier. The two other specifications were primarily used to further examine the marginal credit effect. Since they introduced additional flexibility into the specification, they should have traced this effect more accurately. As can be seen from the values of the log-likelihood function, the fit of Models II and III continuously improved as compared with the linear model. Marginal effect of credit expansion The marginal effect given in Table 4 for the standard Tobit model is the partial derivative of expected investment with respect to credit, evaluated at sample means. Again this marginal effect is smaller than one, and even lower than in the linear model. If the coefficients of the Tobit model are taken as the correct marginal effects, they still indicate a slope smaller than one. The general result of a diversion of the marginal credit funds from investment is hence supported. The cubic Tobit model is the most flexible and therefore also the most accurate depiction of the true relationship. To check the precision of the marginal effect, the latter was calculated together with its standard error at sample means of non-zero investors (see Greene, 2000, p. 326). This resulted in a marginal effect of 0.559 with a t-value of 4.502, which is significantly different from zero at the one percent level. Therefore, using the coefficients of the cubic Tobit model, marginal effects were computed individually for all investors in the sample (Fig. 2). The shown histogram offers two interesting insights: First, almost all (98.4 percent, to be precise) observations fall below the threshold of one. Second, the mean of these individual marginal effects is substantially below the (constant) marginal effects obtained from the coefficients of Models I and II. A further analysis of the functional relationship between credit and investment based on the cubic Tobit coefficients revealed that, over the range of commonly observed credit volumes, the marginal effect increases with increasing credit volume. The function is hence convex, i.e. the second derivative is positive. At a credit 7

The survey data does not allow assessment of the return on investments carried out by farmers. Since complete defaults are of minor relevance in Poland irrespective of credit use, it must be assumed that many investment loans used for consumptive purposes are repaid out of current household income rather than direct investment returns.

290

M. Petrick / Food Policy 29 (2004) 275–294

Fig. 2. Distribution of marginal credit effects. Mean = 0.526; Std. dev. = 0.144; N = 124. Source: Author’s calculations based on cubic Tobit model.

volume of 200 thousand zl (about ten times the mean), there is an inflection point, indicating a decreasing marginal effect for larger credit volumes. In the range between 150 and 275 thousand zl credit volume, the slope is almost stable at about one or even slightly higher. In terms of the additional mobilisation of funds, this could thus be called an optimal credit volume range. However, only 1.6 percent of actual observations fall in this range. More than 95 percent of the observed credit volumes fall below 100 thousand zl and hence in a range of the function where its slope is clearly below one. There is a statistically significant positive correlation of 0.22 between farm size (measured as land owned) and investment volume, i.e. large farms seem to invest more. The correlation between credit volume and land owned is 0.30. Accordingly, particularly if non-borrowers are neglected, high farm-individual marginal credit effects are found in the group of relatively larger farms. However, net of the credit effect, large farms invest less, as can be seen from the regression results. The implication is that, out of a group of farms with equal credit volume, smaller farms devote a higher amount to productive investment. There are hence two opposing effects with regard to investment in absolute terms: larger credit volumes imply more, but larger farm sizes as such (i.e. controlling for credit volume) imply less investment. Overall, it is therefore incorrect to say that large farms invest more, but they obtain larger credit volumes and hence divert less to non-productive activities.

M. Petrick / Food Policy 29 (2004) 275–294

291

Conclusions Summary of the findings First, 45 percent of all respondents claim that they cannot obtain as much credit as desired. Major determinants of being credit-rationed are the reputation of the loan applicant as well as demographic household characteristics. Respondents with a good credit history have a 30 percentage points lower probability of being rationed than borrowers who rescheduled a loan in the past. On the other hand, an effect of land as collateral could not be detected. In addition, more adult males in the household decrease the probability of being credit-rationed, while more females increase it. This is assumed to be an effect of higher liquidity demand for consumption purposes by women or a signalling effect due to the higher share of male labour force. Second, access to subsidised credit plays a statistically significant role in determining investment behaviour of farmers who self-classified as being exogenously credit-constrained. However, in various specifications of the credit-investment relationship, the marginal effect of credit on investment was clearly smaller than one, implying that credit is partly used for purposes other than productive investment. Based on a cubic Tobit estimate of the investment function, the mean of the farm-individual marginal effects was at 0.53 on average. Every second borrower invests less in productive assets than borrowed. Only 1.6 percent of the selected respondents with positive investment displayed farm-individual credit effects larger than one. Over most of the observed range of credit volumes, the marginal effect increases with an increasing credit volume. Policy implications Despite continuing government intervention in rural credit markets, almost half of the interviewed farmers wish to borrow more at the going interest rate or are otherwise discouraged from borrowing. The subsidisation policy is clearly not successful in eliminating credit rationing, which is no surprise if lack of reputation is one of the key problems. Credit rationing by lenders in rural Poland might be a rational response to unresolved problems of asymmetric information, and the high repayment rates in agriculture imply that banks have succeeded in maintaining a low-risk lending portfolio. As supported by the regression results in this article, banks closely screen borrowers and sort out those whose lending history or personal characteristics suggest less than satisfactory loan repayment. This is in contrast to soft budget constraints and excessive default rates reported from other former socialist countries. However, the evidence presented here also implies that the strong risk aversion of banks might go too far, because almost half of the farmers would have preferred to borrow more than they actually received. The background literature cited in this article suggests that there is room for improving bank support for rural entrepreneurs and the efficiency of lending practices currently in use among Polish lenders. The only gradually restructured and still widely governmentally-controlled

292

M. Petrick / Food Policy 29 (2004) 275–294

rural banking system inherited from the socialist period seems to not yet fully conform with the requirements of a market economy. If traditional forms of collateral are of less relevance for the actual rationing outcome because they are difficult to enforce, innovative credit products such as leasing arrangements or cooperation with up- and downstream companies should be further developed (OECD, 1999; Gow and Swinnen, 2001). From the public policy point of view, the targeting of long-term loans in the agricultural sector is rather dubious, and there may be even a kind of saturation on the market for investment loans. Although farmers make their investment decisions conditional on the availability of subsidised credit, funds are only partly used for productive investment. The diversion of funds is more substantial the smaller the loan amount is. This might be taken as evidence for a band-wagon effect, i.e. small loan amounts are taken on favourable terms to finance consumption activities, whereas there is no actual investment project available. If diversion is regarded as undesirable, one solution could be to concentrate future lending on large loan amounts above 150 thousand zl, up to a size of 275 thousand zl. In contrast, lending in small amounts contributes less to foster productive investment. However, the results provide evidence against the view that investment is positively related to farm size. If high investment levels are the aspired policy goal, discrimination against small farms should be avoided. From the perspective of the banks it is most important that borrowers repay punctually, irrespective of how they use the loan. Reliable repayment is generally a given in rural Poland. In light of the high number of unsatisfied borrowers, banks even appear to be too conservative and risk-avoiding. As opposed to that, an assessment of government policy should scrutinise whether subsidies are used to foster structural change or simply raise the living standard of the rural population by providing them with cheap cash. The empirical results of this study provide evidence in favour of the latter. In line with this view, Poganietz and Wildermuth (1999, p. 540) suggest that the use of credit for non-productive purposes in rural Poland is not a result of inaccurately lax supervision by banks and other involved agencies, but a direct consequence of programme design, which allows an (excessively) unspecified spectrum of credit uses. The question which should therefore be addressed by the government is whether its credit policy is aiming at specifically targeted goals to modernise the agricultural sector or whether it is more understood as a part of social policy for rural areas. If the former applies, it might be better policy to foster progress in the restructuring of the rural banking sector than to pour huge amounts into interest subsidies.

Acknowledgements The author is grateful to Stephan Brosig, Catrin Schreiber, Peter Weingarten, and the referees and editor of the journal for helpful comments on earlier versions of this paper, and to Dominika Milczarek for providing access to statistical data.

M. Petrick / Food Policy 29 (2004) 275–294

293

James Curtiss kindly assisted in finally preparing the manuscript. The usual disclaimer applies.

References Chambers, R.G., Lopez, R.E., 1984. A general, dynamic, supply-response model. Northeastern Journal of Agricultural and Resource Economics 13, 142–154. Christensen, G., Lacroix, R., 1997. Competitiveness and Employment. A Framework for Rural Development in Poland. Discussion Paper 383. World Bank, Washington D.C. Czerwin´ska-Kayzer, D., 2000. Analyse von Investitionen und ihrer Finanzierung in privaten landwirtschaftlichen Betrieben in Polen wa¨hrend der Transformation. Paper presented at the KATO Symposium ‘Understanding Transition in Central and Eastern European Agriculture’ held in Berlin, Germany, November 2–4. Diamond, D.W., 1989. Reputation acquisition in debt markets. Journal of Political Economy 97, 828–862. Elhorst, P.J., 1993. The estimation of investment equations at the farm level. European Review of Agricultural Economics 20, 167–182. Feder, G., Lau, L.J., Lin, J.Y., Luo, X., 1990. The relationship between credit and productivity in Chinese agriculture: a microeconomic model of disequilibrium. American Journal of Agricultural Economics 72, 1151–1157. Feder, G., Lau, L.J., Lin, J.Y., Luo, X., 1992. The determinants of farm investment and residential construction in post-reform China. Economic Development and Cultural Change 41, 1–26. Gow, H.R., Swinnen, J.F.M., 2001. Private enforcement capital and contract enforcement in transition economies. American Journal of Agricultural Economics 83, 686–690. Greene, W.H., 2000. Econometric Analysis, 4th ed. Prentice Hall, London. Heckman, J.J., 1979. Sample selection bias as a specification error. Econometrica 47, 153–161. Hubbard, R.G., Kashyap, A.K., 1992. Internal net worth and the investment process: an application to U.S. agriculture. Journal of Political Economy 100, 506–534. Jappelli, T., 1990. Who is credit constrained in the U.S. economy? Quarterly Journal of Economics 105, 219–234. Johnston, J., DiNardo, J., 1997. Econometric Methods, 4th ed. McGraw-Hill, New York. Karcz, Z., 1998. Agricultural credit in Poland in the period of transformation. In: Frohberg, K., Poganietz, W.R. (Eds.), The Importance of Institutions for the Transition in Central and Eastern Europe. Studies on the Agricultural and Food Sector in Central and Eastern Europe 1. Wissenschaftsverlag Vauk, Kiel, pp. 94–97. Khitarishvili, T., 2000. Efficiency and Economies of Scale and Scope of Polish Cooperative Banks. Revised version of a paper presented at the IAAE conference ‘Tomorrow’s Agriculture: Incentives, Institutions, Infrastructure and Innovations’ held in Berlin, Germany, August 13–18. Klank, L., 1999. Changes in the financial system in rural areas. Wies´ i Rolnictwo 103 (suppl. 2), 33–41. Kulawik, J., 2003. Kredytowanie i opodatkowanie rolnictwa oraz ubezpieczenia rolnicze, In Analiza produkcyjno-ekonomicznej sytuacji rolnictwa i gospodarki z˙ywnos´ciowej w 2002 roku, Instytut Eko˙ ywnos´ciowej (IERiGZ ˙ ), Warsaw, 76–89. nomiki Rolnictwa i Gospodarki Z Latruffe, L., 2003. Farm credit and investment in Poland: A case study. Working Paper 03-08. Institut National de la Recherche Agronomique (INRA), Rennes. http://www.rennes.inra.fr/economie/pdf/ WP03-08.pdf Milczarek, D., 2003. Credit to Rural Areas in Poland. In: European Institute (ed.), Financial Intermediation in the European Union Candidate Countries: the Banking Sector Role. European Institute, Sofia, 201–220. Ministerstwo Rolnictwa i Rozwoju Wsi, 2000. Biuletyn informacyjny 2/2000, MRiRW, Warsaw. OECD, 1999. Agricultural finance and credit infrastructure in transition economies. OECD Centre for Co-operation with Non-Members, Paris. OECD, 2000. Agricultural Policies in OECD Countries, Monitoring and Evaluation 2000. OECD, Paris.

294

M. Petrick / Food Policy 29 (2004) 275–294

Petrick, M., 2001. Documentation of the Poland Farm Survey 2000. Discussion Paper 36. Institute of Agricultural Development in Central and Eastern Europe (IAMO), Halle (Saale). http://www.iamo.de/no36.pdf Petrick, M., 2004a. A microeconometric analysis of credit rationing in the Polish farm sector. European Review of Agricultural Economics 31, 77–101. Petrick, M., 2004b. Empirical measurement of credit rationing in agriculture: a methodological survey. Agricultural Economics, forthcoming. Petrick, M., Spychalski, G., S´witlyk, M., Tyran, E., 2002. Economic situation and development perspectives of farms in Poland. An analysis based on survey data from selected Polish voivodships and a comparison with German farms. Agrarwirtschaft 51, 203–214. Poganietz, W.R., Wildermuth, A., 1999. Kredit- und Finanzma¨rkte im Rahmen des Transformationsprozesses des Agrarsektors in Polen. In: Berg, E., Henrichsmeyer, W., Schiefer, G. (Eds.), Agrarwirtschaft in der Informationsgesellschaft. Schriften der Gesellschaft fu¨r Wirtschafts- und Sozialwissenschaften des Landbaues e.V. 35. Landwirtschaftsverlag, Mu¨nster-Hiltrup, pp. 536–540. Prosterman, R.L., Rolfes, L., 2000. Review of the legal basis for agricultural land markets in Lithuania, Poland, and Romania. In: Csaki, C., Lerman, Z. (Eds.), Structural Change in the Farming Sectors in Central and Eastern Europe. Technical Paper 465. World Bank, Washington D.C., pp. 110–139. Steigum, E., 1983. A financial theory of investment behavior. Econometrica 51, 637–645. Swinnen, J.F.M., Gow, H.R., 1999. Agricultural credit problems and policies during the transition to a market economy in Central and Eastern Europe. Food Policy 24, 21–47. Von Pischke, J.D., Adams, D.W., 1980. Fungibility and the design and evaluation of agricultural credit projects. American Journal of Agricultural Economics 62, 719–726. World Bank, 1995. Understanding Poverty in Poland. World Bank, Washington D.C. World Bank, 2001. Poland—The Functioning of the Labor, Land and Financial Markets: Opportunities and Constraints for Farming Sector Restructuring. World Bank, Washington D.C.