Uses and Abuses of ARIMA in PPNR Modeling and Risk Management: Why Not to Fear ARIMA

Abstract

Many forms of the ARIMA (auto-regressive integrated moving average) modeling method are used across risk management and specifically within PPNR (Pre-Provision Net Revenue) for CCAR (Comprehensive Capital Analysis and Review) and DFAST (Dodd-Frank Act Stress Testing). The ARIMA method allows for flexible modeling of PPNR and the inclusion of exogenous variables however model stability can be a concern. I argue that model instability is occurring because of improper ARIMA model development and the practice of forcing all data into the ARIMA framework. I apply a basic method of testing model stability over time and have chosen to model both Citigroup and the S&P 500 using Federal Reserve domestic data which is used in the actual CCAR and DFAST exercises. This paper aims to show common mistakes that occur throughout risk management from the perspective of model development, validation, implementation, and internal audit at major financial institutions.