Analytical Methods
Quantitative data was analyzed using descriptive statistics such as mean, standard deviation and percentage used to
investigate the relative importance of major variables hypothesized to influence loan repayment performance of
smallholder farmers. Moreover a two-limit tobit model was used to select variables which most significantly
distinguish between non-defaulters and defaulters of agricultural loan, from a set of personal and socio-economic
variables hypothesized to influence repayment behavior.
The various studies on loan repayment performance in different countries identified the most probable causes of
loan default. Moreover, the major independent variables such as age, gender, credit experience, loan diversion,
education level, weak supervision, among others, were analyzed using different models such as logit, probit, and
Ordinary Least Square multiple regression method. However most of the studies conducted in modeling the
determinants of loan repayment have used dichotomous discrete choice models (Logit and Probit) where the
dependent variable is a dummy that takes a value of zero or one depending on whether or not a farmer has
defaulted. However, Lynne et al. (1988) pointed out possible loss of information if a binary variable is used as the
dependent variable because of the dependent variable may have more than two outcome. In addition, binomial
models, explain only the probability that an individual made a certain choice (i.e. defaulted or has not defaulted)
and they fail to take into account the degree of loan recovery. The linear probability model (LPM), even though
computationally and conceptually simpler and easier than the binary choice models, it depends on the use of
ordinary least squares (OLS) approach. Application of OLS to censored model however, inherently produces
heteroscedastic disturbance term (ɛi) and as a result, the standard deviations of the estimates are biased. These
inadequacies are minimized with the use of the Tobit Model (Tobin, 1958). Therefore, the current study employed
two limit tobit regression model to determine causes of loan repayment performance in the study areas.In this study, the value of the dependent variable is repayment ratio that has been computed as the ratio of amount
of loan repaid to the total amount due from formal sources of credit. Thus, the value of the dependent variable
ranges between 0 and 1 and a two-limit Tobit model has been chosen as the appropriate econometric model. The
two-limit Tobit was originally presented by Rossett and Nelson (1975) and discussed in detail by Maddala (1992)
and Long (1997). The model derives from an underlying classical normal linear regression and can be represented
as: