Macro stress testing of the credit risk of banking book exposures has attracted an increasing interest from market participants in the last years due to three main reasons. First, the Basel II capital accord (Basel Committee on Banking Supervision, 2006), more specifically the internal models approach contained therein, has led commercial banks and supervisors to focus attention on credit risk stress testing exercises as a way to further test the reliability of internal models derived capital measures. Furthermore, commercial banks are likely to use stress testing of their banking exposures for a variety of other purposes, including economic capital management, planning of contingent measures and risk transfer transactions.
Second, the increasing role of financial stability as a policy goal of central banks has promoted increasing interest in system-wide exercises of credit risk macro stress testing, often using data aggregated at a higher level than the analysis performed in commercial banks. Cihák (2007) and Foglia (2009) discuss and review general methodologies for implementing stress tests in financial systems. Such tests may help central banks evaluate existing capital adequacy of commercial banks and foresee the consequences of unexpected macro shocks to the stability of the banking system. This paper focuses on this system-wide version of credit risk stress testing.
Third and finally, the outbreak of the recent financial turmoil, coupled with lack of more warning signals raised before the crisis, has, if anything, reinforced the two previous points and stimulated further research on the theme and its limitations (e.g. Alfaro and Drehman, 2009). As a response to the crisis, regulation has once more devoted new attention to the area of stress-testing (e.g. Basel Committee on Banking Supervision, 2009).
While the relation between the macroeconomy and the volume of credit is relatively well studied, e.g. the credit channels of monetary policy (Bernanke and Gertler, 1995), the economic theory is still incipient to explain the link between macro variables and credit risk. In the absence of well-established theoretical models to explain the macro-credit risk link, the majority of macro stress-testing approaches currently in use by central banks or supervisory agencies are non-structural. One reduced-form approach widely employed in the applied literature is Wilson (1997a, 1997b). This paper discusses and estimates Wilson model and uses it to perform macro stress testing of the credit risk of the Brazilian household sector.
Wilson model, originally conceived basically as a credit risk portfolio model, has the interesting builtin feature that macroeconomic surprises affect the macro-credit risk relationship, which is maybe a reason for its popularity in stress-testing applications. On the other hand, Sorge and Virolainen (2006) perform a critical review of stress testing methodologies, including approaches of Wilson type, pointing to the potential instability of reduced form parameter estimates, due to the break-down in historical patterns derived from extreme shocks (e.g. in default correlations). That motivates us to also consider an alternative model for the macro-credit risk link that incorporates stochastic macro sensitivity of the credit risk indicator. In estimating and applying these models on a system level, this paper situates itself amid a recent but fast growing literature on credit risk stress-testing applications by central banks and supervisory agencies (e.g. Kalirai and Scheicher (2002), Boss (2003), Lehman (2006), van den End et al. (2006), Jiménez and Mencía (2007), Breuer et al. (2009), Simons and Rolwes (2009)).
Download
Macro Stress Testing of Credit Risk Focused on the Tails
