Empirical measurement of credit rationing in agriculture: a methodological survey

Empirical analysis of rural credit market failure has been of key scientific and political interest in recent years. The aim of this paper is to give an overview of various methods for measuring credit rationing of farms employed in the literature. Furthermore, based on a common analytical framework entailing a formal model of a credit rationed farm household, the methods are subjected to a comparative evaluation of their specific strengths or shortcomings. Six approaches are distinguished: measurement of loan transaction costs, analysis of qualitative information collected in interviews, analysis of quantitative information collected in interviews by using the credit limit concept, analysis of spill-over effects with regard to secondary credit sources, econometric household modelling, and the econometric analysis of dynamic investment decisions. The first approach defines credit rationing as the impossibility to take a loan due to prohibitively high, measurable transaction costs on loan markets, which is a price rationing mechanism. All other approaches at least implicitly define credit rationing as a persistent private excess demand in terms of a quantity restriction. The six approaches are more or less closely linked to the neo-classical efficiency concept. An explicit comparison with a first-best solution is impossible in the first three approaches, since they essentially rely on a subjective assessment of borrowers' access to credit, based on qualitative or quantitative indicators. The fifth and sixth approach allow a rigorous interpretation in the framework of neo-classical equilibrium theory. The fourth approach takes an intermediate position, since spill-over on segmented loan markets reveals a willingness to pay with regard to the supposedly less expensive but rationed primary source. Approaches are fairly data demanding in general, usually requiring specific data on loan transactions. Even so, most approaches are applicable to cross-sectional household data. Only dynamic modelling of investment decisions necessitates the availability of panel data, therefore restricting the applicability in low-income and transition countries. With the exception of the first, all methods surveyed might plausibly be used to empirically detect credit rationing.