the web's biggest credit risk modeling resource.

Credit Jobs

Home Glossary Links FAQ / About Site Guide Search


Submit Your Paper

In Rememberance: World Trade Center (WTC)

Export citation to:
- Text (plain)
- BibTeX

Macro Stress Tests and History-Based Stressed PD: The case of Hong Kong

by Michael C.S. Wong of the City University of Hong Kong & CT Risk Solutions, and
Yat-fai Lam of Hong Kong Monetary Authority

March 2008

Abstract: This paper discusses the issues about the stress testing of banks' credit portfolios. Currently there is no standard methodology to perform stress tests on banks' credit portfolios and no standard to evaluate self-reported stress testing results from banks. Some banks and bank supervisors have attempted to build econometrics models for macro stress tests. These models may provide inconsistent conclusions because of insufficient data available, unstable patterns of association, nonlinear behavior of credit loss in stress conditions, and the relevance of the historical data in calibrating the model parameters. These issues on econometrics models are illustrated with data of Hong Kong in 1997-2007. This period is an unusual stress period for Hong Kong economy, having Asian financial crisis in 1997, burst of Internet bubble in 2001 and SARS outbreak in 2003. With the given data, we find that it is challenging to identify suitable models for forecasting. This paper proposes a methodology to estimate history-based stressed probability of default (PD) to complement the use of macro stress tests. By analyzing the default rates of the banking sector, bank supervisors can easily identify the stressed PD of individual banks' credit portfolios. These estimates are also very helpful for bank supervisors to verify those self-reported stressed PD and to compute the capital adequacy ratios of the banks under stress.

Keywords: Stress Test, Credit Risk, Probability of Default.

Published in: Journal of Financial Regulation and Compliance, Vol. 16, No. 3, (July 2008), pp. 251-260.

Download paper (149K PDF) 17 pages