A quantification method for the collection effect on consumer term loans

Modeling state transitions of loan accounts as Markov transition matrixes, we propose a method for detecting the significance and quantifying the magnitude of collection effects on consumer term loan accounts. Quantification of the collection effect provides a theoretical basis for making optimal collection decisions with respect to loan accounts. A parameterization process is presented to reduce the number of parameters required to estimate. The quantification process consists of two steps. First, a Chi-square test detects whether the transition probability distributions with and without collection differ significantly. Second, a regression and a t-test are used to assess the magnitude of the collection effect. Application of this method to quantify collection effects in a Chinese automobile loan financing company shows that the method is able to recognize the magnitude and significance of collection effects. This paper further sets forth suggestions on how to design an experiment for necessary data collection.