A Method of Retail Mortgage Stress Testing: Based on Time‐Frame and Magnitude Analysis

Abstract

In this study, a non-stationary Markov chain model and a vector autoregressive moving average with exogenous variables coupled with a logistic function (VARMAX-L) are used to analyze and predict the stability of a retail mortgage portfolio, based on the stress test framework. The method introduced in this paper can be used to forecast the transition probabilities in a retail mortgage over pre-specified states, given a shock with a certain magnitude. Hence this method provides a dynamic picture of the portfolio transition process through which one can assess its behavior over time. While the paper concentrates on retail mortgages, the methodology of this study can be adapted also to analyze other credit products in banks.

A Markov chain model of crop conditions and intrayear crop yield forecasting

“…More specifically, we specify a model of the intrayear movement of reported crop conditions as a Markov chain. Markov chains have seen extensive use in forecasting by, for example, Liu et al (2015) for mortgage stress testing, Lo et al (2016) for latent volatility, Tang et al (2018) for scenario analysis, and Li and Andersson (2020) for density forecasting. Markov chains specific to forecasting crop yields have been employed by Matis et al (1985Matis et al ( , 1989, but those efforts predate the USDA crop condition data described above.…”