Michael B. Gordy
Senior Economist, Federal Reserve Board, Washington DC, USA1 Visiting Scholar, Indian School of Business, Hyderabad, India
It is with great pleasure that I welcome you to this special issue of The Journal of Credit Risk. The issue is devoted to a selection of papers from a workshop on “Concentration risk in credit portfolios”, which was convened at the gracious Bundesbank Training Centre in Eltville am Rhein in November 2005. The workshop was organized jointly by the Basel Committee on Banking Supervision, the Deutsche Bundesbank and The Journal of Credit Risk, to promote research on the identification and modeling of single-name and sectoral concentrations in portfolio credit risk.2
In the portfolio risk-factor frameworks that underpin both industry credit VAR models and the internal ratings-based (IRB) risk weights of Basel II, credit risk in a portfolio arises from two sources: systematic and idiosyncratic. Systematic risk represents the effect of unexpected changes in macroeconomic and financial market conditions on the performance of borrowers. Borrowers may differ in their degree of sensitivity to systematic risk, but few firms are completely indifferent to the wider economic conditions in which they operate. Therefore, the systematic component of portfolio risk is unavoidable and is only partially diversifiable. Idiosyncratic risk represents the effects of risks that are peculiar to individual firms. As a portfolio becomes more and more fine-grained, in the sense that the largest individual exposures account for a smaller and smaller share of total portfolio exposure, idiosyncratic risk is diversified away at the portfolio level.
The model framework for the IRB approach assumes that (1) there is only a single source of systematic risk, and (2) bank portfolios are perfectly fine-grained. To the extent that either assumption is violated, IRB capital requirements may understate the true economic capital requirement. When there are material concentrations of exposure to individual names, there will be a residual of undiversified idiosyncratic risk in the portfolio. This form of credit risk concentration is known as name concentration or granularity, and it can be addressed via a granularity adjustment to portfolio capital. While well understood in principle, in practice there may remain challenges in implementation.3
Violations of the “single systematic factor” assumption may be more difficult to discern and also more difficult to address. Within a large single market such as the United States or the European Union the macroeconomic performance of different geographic regions may not be fully synchronized. Exposures in foreign jurisdictions are additionally subject to country-specific risks, including transfer risk and legal risk. Similarly, different industries within a market may experience different cycles. If so, then distinct geographic regions and industries ought to be represented by distinct (though certainly correlated) systematic risk factors. A bank portfolio may be overweight in exposure to some of these risk factors and underweight to others. This form of credit risk concentration is known as sectoral concentration. The development of analytical tools for adapting single-factor frameworks for sectoral concentrations is still at an early stage. Difficult practical issues arise as well, for example, on how the relevant sectoral factors can be identified empirically. Appropriate methodologies for the modeling and estimation of default dependence are essential ingredients for progress in this area.
The research articles in this issue develop new tools for the theoretical, numerical or empirical analysis of single-name and sectoral credit risk concentrations. We begin with a study by Erik Heitfield, Steve Burton and Souphala Chomsisengphet on "Systematic and idiosyncratic risk in syndicated loan portfolios". The authors assess the influence of name and sectoral concentrations on credit value-at-risk (VAR) in a sample of real bank portfolios of syndicated loans. They decompose VAR into systematic and idiosyncratic contributions, and then further allocate the systematic component by industrial sector. They find that larger portfolios tend to be better diversified across sectors, but also tend to be more heavily weighted towards high-risk sectors. The authors propose several simple indices for concentration in credit portfolios and evaluate the indices as predictors for
We turn next to two complementary methodological studies on measuring sectoral diversification. In Dirk Tasche’s article, "Measuring sectoral diversification in an asymptotic multifactor framework", the asymptotic single risk factor model of the IRB framework is extended to multiple factors. The emphasis in this paper is on rigorous derivation of analytical solutions for risk contributions, as well as on generality with respect to the choice of risk measure. Tasche introduces a new diversification factor, which measures the reduction in VAR that comes from recognizing imperfect correlation between sector risk factors, as well as its marginal counterpart at the sector level. Numerical exercises demonstrate how the marginal diversification factors can be applied to capital allocation in a two-factor setting. Juan Carlos Garcia Cespedes, Juan Antonio de Juan Herrero, Alex Kreinin and Dan Rosen pursue a more heuristic strategy in "A simple multifactor “factor adjustment” for the treatment of credit capital diversification". The emphasis in this paper is on tractability, intuitive appeal, robustness and flexibility across a variety of model specifications. A diversification factor is proposed which is similar in spirit to that of Tasche, but which is approximated as a simple function of two parameters that represent sector concentration and average cross-sector correlation. Once the function is calibrated to a full portfolio model, it lends itself to “real-time” sensitivity analysis, stress testing and capital allocation.
In the fourth paper, Alexander J. McNeil and Jonathan P. Wendin apply Bayesian techniques to the difficult econometric problem of estimating latent risk-factor models of "Dependent credit migrations". The approach allows for heterogeneity across industry sectors and serial dependence in the risk factors and yet remains computationally tractable. As an empirical application, they extract implied asset correlations from quarterly migration data from Standard & Poor’s for the years 1981–2000.
The final paper is devoted to the underexplored topic of stress testing of credit risk concentrations. In "Credit risk concentrations under stress", Gabriel Bonti, Michael Kalkbrener, Christopher Lotz and Gerhard Stahl provide a clear and compelling paradigm for stress tests of this sort. The central idea, that stress tests should be constructed via truncation of the distribution of a sector risk factor, is independent of model choice and robust to the dependence structure among the risk factors, yet also amenable to computationally efficient implementation. To formalize the approach, the authors introduce and analyze a measure of factor concentration.
1 The opinions expressed in this introduction are my own and do not reflect the views of the Board of Governors of the Federal Reserve System or its staff.
2 The programme committee was composed of Klaus Düllmann (Deutsche Bundesbank), William Perraudin (Imperial College), Michael Pykhtin (Bank of America), Tom Wilde (Credit Suisse First Boston) and myself.
3 The treatment of hedging positions in a granularity adjustment may raise special challenges, in particular when credit risk mitigation activities indirectly give rise to concentration of exposure to providers of credit protection.