Markov chain models for farm credit risk migration analysis

This study introduces two Markov chain time approaches, time‐homogeneous and nonhomogeneous models, for analyzing farm credit risk migration as alternatives to the traditional discrete‐time (cohort) method. The Markov chain models are found to produce more accurate, reliable transition probability rates using the 3 x 1 migration measurement method used by farm lenders. Compared to corporate bond ratings migration results, this study obtained larger mean differences in singular value decomposition between the cohort matrix and each of the Markov chain matrices. This finding suggests that the omission of transient, indirect migration activities under the cohort method is more costly when applied to farm credit analysis. This discrepancy could lead to understated transition probability estimates which, in turn, could produce misleading indicators of farm loan portfolio quality.