Mathematical Sciences Department
Statistics Seminar
Speaker: Dr. Mai Dao
Thursday, September 21, 2023
11:00 am - 12:00 pm
Virtual, via Zoom: 99064789615
https://wpi.zoom.us/j/99064789615
Title: Bayesian inference in quantile regression models for macroeconomic stress testing.
Abstract: Identification of key risk factors is an important first step in designing an effective macro stress test for financial institutions. The key assumption is that the identified risk factors should be useful in predicting the tail risks of financial institutions under severe macro and financial conditions. We argue that high-dimensional Bayesian quantile regression models with shrinkage priors are ideal tools for that task. Our framework allows researchers to evaluate information from a large set of risk factors and their contributions to tail risks. We illustrate our methods by estimating several models with hundreds of macroeconomic and financial variables to identify the relevant risk factors for bank performance in both the short and long runs. We find that the best predictors for the banking system’s credit risks and profitability are their own lags, the NFCI indices, and the credit-to-GDP gap. Our approach reaffirms the importance of the credit cycle and having multiple scenarios with different economic environments in designing macroeconomic stress tests.
Organizing Committee: Frank Zou, Fangfang Wang, Tharindu De Alwis, and Nadeesha Jayaweera.