Forecasting probabilities of default and loss rates given default in the presence of selection (original) (raw)

Predicting and Pricing the Probability of Default

In this paper we study how corporate bond defaults can be predicted using financial ratios and how the forecasted probability of default relates to the cross-section of expected stock returns. Using several performance measures we find that the duration model outperforms existing models in correctly classifying both Default and Non-Default firms. Using the default probabilities predicted by our model, we analyze the relation between default risk and the Fama-French distress factors, HML and SMB. We find evidence that supports the interpretation on HML as a distress related factor. Both portfolio and individual stock factor loadings are related to the estimated default probabilities. We find a negative and significant contemporaneous correlation between HML and shocks to the level of aggregate financial distress. Torous for valuable suggestions and helpful comments. All remaing errors are the author's responsability.

Performance of default-risk measures: the sample matters

Journal of Banking & Finance, 2020

This paper examines the predictive power of the main default-risk measures used by both academics and practitioners, including accounting measures, market-price-based measures and the credit rating. Given that some measures are unavailable for some firm types, pair wise comparisons are made between the various measures, using same-size samples in every case. The results show the superiority of market-based measures, although their accuracy depends on the prediction horizon and the type of default events considered. Furthermore, examination shows that the effect of withinsample firm characteristics varies across measures. The overall finding is of poorer goodness of fit for accurate default prediction in samples characterised by high book-to-market ratios and/or high asset intangibility, both of which suggest pricing difficulty. In the case of large-firm samples, goodness of fit is in general negatively related to size, possibly because of the "too-big-to-fail" effect.

Measuring the Risk of Default: A Modern Approach

The UniTed STaTeS is a nation of debtors. By the end of 2007, total debt outstanding by households, businesses, state and local governments, and the federal government added up to 31.2trillion.Thedomesticfinancialsectoraccountedforhalfofthistotal,or31.2 trillion. The domestic financial sector accounted for half of this total, or 31.2trillion.Thedomesticfinancialsectoraccountedforhalfofthistotal,or15.8 trillion. The size of the debt market is quite large. Indeed, it exceeds both the U.S. GDP in 2007 ($13.8 trillion) and the equity market value of all domestic corporations ($15.5 trillion). 1 The primary risk of all this debt is credit risk, or the risk of default. The current credit crisis demonstrates how shifts in credit spreads and market liquidity can also significantly impact debt values. Although these alternative factors are important for understanding debt markets, we will focus here only on default risk. Investors measure default risk in many different ways, and there have been important recent innovations in this regard. The state of the art in assessing corporate credit risk is based on one of three approaches: 1) the Merton distance-to-default measure, 2) the reduced-form approach, and 3) credit ratings. We will compare and contrast these three approaches, showing that the reducedform approach is preferred because of its generality, flexibility, and superior forecasting ability. Merton's Distance-to-Default For more than three decades, a common approach used to measure a firm' s default probability has been the so-called distance-to-default. This measure is based on the pioneer

Empirical performance of loss given default prediction models

The global financial crisis highlighted the fact that default and recovery rates of multiple borrowers generally deteriorate jointly during economic downturns. The vast majority of the literature, as well as many industry credit-portfolio risk models, ignore this and analyze default probabilities and recoveries in the event of default separately. As a result, the models project losses that are too low in economic downturns such as the recent financial crisis. Nevertheless, alternatives that incorporate the dependence between probabilities of default and recovery rates have been proposed. This paper is the first of its kind to assess the performance of these structurally different approaches. Four banks using different estimation procedures are compared. We use root mean square errors and relative absolute errors to measure the predictive accuracy of each procedure. The results show that models accounting for the correlation of default and recovery do indeed perform better than models ignoring it.

Estimating & Forecasting Default Risk: Evidence from Jamaica

2017

This paper employs the GMM estimation technique to evaluate the impact of macroeconomic factors on bank default risk for listed Jamaican banks and securities dealers (SDs) over the period December 2004 to June 2016. Default risk is captured by a distance-to-default measure which is computed using a Merton type, option-based model. This indicator accurately tracks the default experience of listed Jamaican banks and SDs over important dates throughout the sample period. The estimation results of the model revealed that GDP growth, inflation, the unemployment rate, growth in domestic private sector credit as well as the REER have a statistically significant impact on the performance of the distance to default measure. As such, the econometric findings validate the sensitivity of the fragility measure to the variability of key macroeconomic variables. The model was also utilized to forecast the distance to default measure six-quarters ahead, as this will aid in the formulation of policy...

Incorporating prediction and estimation risk in point-in-time credit portfolio models

In this paper we focus on the analysis of the effect of prediction and estimation risk on the loss distribution, risk measures and economic capital. When variables for the determination of probability of default and loss distribution have to be predicted because they are not available at the time the prediction is made, the prediction is prone to errors. The model parameters for the estimation of probability of default or asset correlation are not available, and usually have to be estimated using historical data. The incorporation of prediction and estimation risk generally leads to broader loss distributions and therefore to rising values of risk parameters such as Value at Risk or Expected Shortfall. The level of economic capital required may be strongly underestimated if prediction and estimation risk are ignored.

The Bayesian Approach to Default Risk Analysis and the Prediction of Default Rates

2011

A Bayesian approach to default rate estimation is used to predict default rates on the basis of information from data and experienced industry experts. The principle advantage of the Bayesian approach is the potential for coherent incorporation of expert information crucial when data are scarce or unreliable. A secondary advantage is access to efficient computational methods such as Markov Chain Monte Carlo. The power of this approach is illustrated using annual default rate data from Moody’s (1999-2009) for two risk buckets and priors elicited from industry experts. Three structural credit models in the asymptotic single risk factor (ASRF) class underlying the Basel II framework (Generalized Linear and Generalized Linear Mixed Models), are analyzed using a Markov Chain Monte Carlo technique. The predictive distributions for defaults are obtained.

Estimating Probabilities of Default

SSRN Electronic Journal, 2000

We conduct a systematic comparison of confidence intervals around estimated probabilities of default (PD), using several analytical approaches from large-sample theory and bootstrapped small-sample confidence intervals. We do so for two different PD estimation methods-cohort and duration (intensity)-using twenty-two years of credit ratings data. We find that the bootstrapped intervals for the duration-based estimates are surprisingly tight when compared with the more commonly used (asymptotic) Wald interval. We find that even with these relatively tight confidence intervals, it is impossible to distinguish notch-level PDs for investment grade ratings-for example, a PD AA-from a PD A+. However, once the speculative grade barrier is crossed, we are able to distinguish quite cleanly notch-level estimated default probabilities. Conditioning on the state of the business cycle helps; it is easier to distinguish adjacent PDs in recessions than in expansions.

Parameterizing Credit Risk Models

SSRN Electronic Journal, 2004

The present paper shows how the parameters of three popular portfolio credit risk models can be empirically estimated by banks using a Maximum Likelihood framework. We apply the method to a database of German firms provided by Deutsche Bundesbank and analyze the inclusion of macroeconomic and borrower specific rating factors. Given the uniform ML estimation methodology, we compare the parameter estimates and the forecasted loss distributions for the credit risk models and find that they perform in very similar ways, in contrast to the differences found in some previous studies. We also propose an approach for addressing estimation errors. Our findings suggest that for a financial institution "model risk", i.e. the risk of choosing the "wrong" credit model, may be considerably reduced.

The Evaluation of Model Risk for Probability of Default and Expected Loss

2019

The quanti�cation of model risk is still in its infancy. This paper provides an operational quanti�cation of this risk for credit portfolio, when the objective is to approximate the average loss. The methodology is easy to implement and does not require the construction of any worst-case model. The required capital computed to cover for model risk depends on three components, that are an estimated impact of the incorrect model, an evaluated risk of inaccurate estimation of model risk and the prediction error hedge factor. The approach is illustrated by an application to a portfolio of corporate loans segmented by grades.