A Markov Chain approach to determine the optimal performance period and bad definition for credit scorecard

TL;DR

This paper proposes a Markov Chain-based method to objectively determine the optimal performance period and bad definition in credit scorecard development, addressing subjective biases and improving model quality.

Contribution

It introduces a novel, simple, and effective approach using Markov Chains to resolve the subjective challenges in defining performance periods and bad definitions.

Findings

Provides a systematic method for performance period determination

Reduces subjectivity in bad definition selection

Improves credit scorecard development process

Abstract

Performance period determination and bad definition for credit scorecard has been a mix of fortune for the typical data modeler. The lack of literature on these matters led to a proliferation of approaches and techniques to solve the problems. However, the most commonly accepted approach involves subjective interpretations of the performance period and bad definition as well as being chicken and egg problem. These complications result in poorly developed credit scorecard with minimal benefits to the banks. In this paper, we will be recommending a simple and effective approach to resolve these issues.