Gunter Löffler

The effects of estimation error on measures of portfolio credit risk

Abstract This paper uses Monte Carlo simulations to assess the impact of noisy input parameters on the accuracy of estimated portfolio credit risk. Assumptions about input quality are derived from the distribution of historical sample statistics commonly used in default risk modelling. The resulting estimation error in the distribution of portfolio losses is considerable. Losses that are judged to occur with a probability of 0.3% may actually occur with a probability of 1%. The paper also shows that estimation error leads to biases in value at risk estimates and significance levels of backtests. The biases can be corrected by analysing predictive distributions which average over the unknown parameter values.

Refining the Carlson-Parkin method

Abstract The measurement error of the Carlson–Parkin estimator, which is often used to quantify qualitative survey data, can be derived through Monte Carlo simulations. Based on such simulations, it is possible to check the estimator’s sensitivity to mis-specifications, and to improve its performance.

Incorporating the Dynamics of Leverage into Default Prediction

A firm’s current leverage ratio is one of the core characteristics of credit quality used in statistical default prediction models. Based on the capital structure literature, which shows that leverage is mean-reverting to a target leverage, we forecast future leverage ratios and include them in the set of default risk drivers. The analysis is done with a discrete duration model. Out-of-sample analysis of default events reveals that the discriminating power of the duration model increases substantially when leverage forecasts are included. We further document that credit ratings contain information beyond the one contained in standard variables but that this information is unrelated to forecasts of leverage ratios.

The Complementary Nature of Ratings and Market-Based Measures of Default Risk

Agency ratings and market-based measures of default risk are useful complements. Combining the two improves the prediction of defaults over the use of a single measure. While in-sample analysis suggests that one should give more weight to ratings as the horizon increases, or issuers become less risky, a simple equal-weight combination of ratings and market-based measures is hard to beat out of sample. The results suggest that both ratings and market-based measures provide genuine information of their own.

Pitfalls in the Use of Systemic Risk Measures

We examine pitfalls in the use of return-based measures of systemic risk contributions (SRCs). For both linear and non-linear return frameworks, assuming normal and heavy-tailed distributions, we identify non-exotic cases in which a change in a banks systematic risk, idiosyncratic risk, size or contagiousness increases the risk of the system but lowers the measured SRC of the bank. Assessments based on estimated SRCs could thus produce false interpretations and incentives. We also identify potentially adverse side effects: A change in a banks risk structure can make the measured SRC of its competitors increase more strongly than its own one.

Caught in the Housing Crash: Model Failure or Management Failure?

I apply standard time series models to US housing prices. Forecasts made in 2005 or earlier would have produced stress scenarios that are worse than the subsequent actual change in housing prices. The probability of these scenarios is in the range that financial institutions should consider in their risk management. Results are robust across a wide range of specifications, and fundamental prediction models lead to the same conclusions. Hence, the fact that the crash caught many market participants by surprise should not be attributed to deficiencies in standard prediction models. Many market participants seem to have focused on the trend predictions, giving too little consideration to risks.

Wall Street’s Bailout Bet: Market Reactions to House Price Releases in the Presence of Bailout Expectations

The rise and subsequent collapse of US house prices was one of the factors underlying the recent financial crisis. One could expect that the crisis brought increased attention to the housing market and thus led to stronger market reactions to house price news. We find that reactions indeed change, but with a peculiar twist: from September 2008 on, good news from the housing market are associated with falling US stock prices, and vice versa. The likely explanation, for which we provide cross-sectional evidence, is that falling house prices increased the market’s trust in a government bail-out, thereby increasing market valuations of firms that were expected to benefit from government rescue measures.

Implied asset value distributions

In portfolio credit risk models, correlated credit events are often modelled by means of correlated latent variables. The latent variables are interpreted as the firms’ asset values, and assumed to follow a normal distribution. A procedure is described that uses the information embodied in rating transition matrices to infer the shape of the latent variable distribution. Applying the approach to transition matrices of different origin yields consistent results. Compared to the normal distribution, the implied asset value distributions are fat-tailed, leading to a substantial increase of portfolio credit risk relative to the one under normality. The results are thus highly relevant for the design of credit risk measurement systems.

Is Contagion Infecting Your Portfolio? A Study of the Euro Sovereign Debt Crisis

For the euro debt crisis, we assess the relevance of financial contagion from an investor perspective. We find that contagion, which we identify through the joint occurrence of extremely negative bond returns, has only small and transitory effects on broad government bond portfolios. For portfolios with concentrated country exposures, the frequency of contagious events does not add information to a standard analysis of risk. Hence, contagion in the euro debt crisis appears to bear little relevance for market timing and country selection alike.

How do Rating Agencies Score in Predicting Firm Performance

We use dynamic panel analysis to examine whether credit rating agencies achieve what they claim to achieve, namely, look into the future when assigning their ratings. We find that Moodys ratings help predict individual financial ratios over a horizon of up to five years. Ratings also predict a multivariate credit score, again over five years. The contribution of ratings appears to be economically significant and robust for different specifications.