Empirical Analysis of Farm Credit Risk Under the Structure Model

The study measures farm credit risk by using farm records collected by Farm Business Farm Management (FBFM) during the period 1995-2004. The study addresses the following questions: 1) whether farm’s financial position is fully described by the structure model, 2) what are the determinants of farm capital structure under the structure model, 3) how to estimate and test farm asset correlation, 4) what drives farm default, and 5) how to predict farm default and joint default.In the first part of the empirical study, a seemingly unrelated regression (SUR) model is proposed to investigate the predicting capability of the structure model and test applicability of theories of financial structure to farm business. The model considers dynamic property of the structure model and farm characteristics. A semi-parametric three-stage least squares (3SLS) estimation method is proposed to obtain the estimates and test the model. In the second part, a farm’s ability to meet its current and anticipated financial obligations in the next 12 months is predicted by the SUR model connected with credit rating models. In the third part of the study, copula approaches as an alternative are applied to measure farm credit risk under the structure model.Results indicate that the empirical dynamic model is stable, and the structure model is applicable in explaining most farms’ choice of financial structure. In addition, the farms adjust to long-run financial targets for asset-to-debt ratio with additional financing needs following both pecking order and agency theories that is stronger for farms with greater asymmetric information problems.Application of the SUR model for measuring credit risk indicates that some key financial ratios in credit risk assessment such as liquidity should enter the model; these variables have significant influence on a farm’s ability to meet its current and anticipated financial obligations in the next 12 months. The estimated average asset correlation is 20% while the average default correlation is around 1.2% across farms in the pool. The estimated average asset correlation is clearly higher than the reported average asset correlation of 16% by KMV’s risk classing (Lopez 2002). The result indicates that the systematic risk plays a more important role in agricultural production in contrast to other industries.Estimated average asset correlation from Gaussian and t copula is 11%, similar to that by using a single factor model (Katchova and Barry 2005). The estimated average default correlation from Gaussian copula is less than 1% while it is 3% from t copula. Test results indicate that Gaussian copula is more proper for asset distribution as implied from the FBFM data than t copula.Results indicate that asset correlation is on average much higher than default correlation, which is consistent with previous findings by Crouchy et al (2002) and Akhavein and Kocagil (2005).Results indicate that mean asset correlation from multi-factor model is clearly higher than that from Gaussian copula. This is also true for default correlation. Higher asset correlation and default correlation from the multi-factor model lead to relatively higher predicted probability of default and expected loss at portfolio level. Apart from difference in methodology for estimating asset correlation, the relatively lower estimated asset correlation under the copulas approach is more likely due to short time series observations for each involved farm.Overall, the predicted default rate and expected loss from the multi-factor model at one-year horizon are 0.77% and 0.19% respectively. These values are similar to those reported by FDIC for agricultural loans issued by commercial banks in Illinois for 1995-2004. Finally, the results illustrate that the methods used in the study can be also applied to agricultural lending using available farm records, which provides a solution to the two major issues in risk assessment for agricultural lending, i.e. lack of long-time loss data and limited information of macroeconomic factors on changes of farm assets.