Bank regulation urged an increasing use of credit ratings during last decade. For instance, we may cite the package of rules to assess the required capital as recommended by the Basel Committee on Banking Supervision. Moreover, the CreditMetrics model, developed in 1997 by JP Morgan, utilizes these credit ratings as a key input to measure credit risk. As a result, we notice an increase of the credit rating activity: credit rating agencies such as Moody's or Standard & Poor's, assign ratings for a higher number of issues. Thus, more valuable and accurate information could be obtained from these larger data samples.;In the first essay, we develop a new method for estimating the rating transition matrix based on Bayes theorem. We show that default probabilities are non-zero even for the highest rated classes and short maturities. Besides, the Bayesian technique allows us to derive confidence intervals for the transition probabilities. Such ability conforms the regulatory concerns about out of sample testing. Then, we use our Bayesian default probabilities to determine the corporate bond spreads explained by default risk. We adopt the same methodology as described in Dionne, Gauthier, Hammami, Maurice, and Simonato (2005) to compute the default spreads. Our results show that the default spreads are higher than those obtained by cohort technique for short maturities.;The second essay deals with migration correlation. An omission or an inaccurate estimation of that correlation will induce a misestimation of the regulatory capital and, consequently, will embrittle the company's financial stability. Moreover, Das, Duffle, Kapadia, and Saita (2007) document that default correlation analysis is needed for asset pricing such as basket default swaps. Thus, an efficient estimation of the migration correlation is necessary to assess the credit risk.;However, reporting all migration correlations seems confusing: 3136 migration correlations should be estimated if we consider 8 credit rating classes. Thus, we follow Jafry and Schuermann (2004) to summarize the rating transition matrix into a scalar (a mobility index). Once time series of theses indices are obtained for each U.S. business sector, we check if the crisis transmission phenomenon exists within each business sector. Then, we test for possibly crisis transmission phenomenon among sectors by estimating a Markov Switching Vector Autoregressions model. The results obtained provide evidence of high and low correlation regimes and prove default contagion among some sectors. For example, more downgrades in the U.S. industrial sector during the high correlation regime imply more downgrades in the U.S. banking sector during the next three months.;Usually, the credit risk dynamics are captured via transition matrices. A credit rating transition matrix corresponds to a summary of probabilities for a particular rating to migrate to other ratings within a period of time. Thus, an accurate estimation of these probabilities is needed to measure the Credit Value at Risk of bonds portfolio or to price defaultable securities and credit derivatives.;The third essay proposes an empirical test to the asset-allocation puzzle posed by Canner, Mankiw, and Weil (1997). These authors conclude that the recommendations of some financial advisors are inconsistent with optimal allocation as predicted by modern portfolio theory.;Our study considers individuals' portfolio choices instead of recommendations from financial advisors. From data on the portfolio composition of 470 clients of a brokerage firm, we have presented a careful verification of the reliability of the risk-tolerance measurement used by the authors. Then, we have obtained that the bonds/stocks ratio does decrease in relation to risk tolerance by using individuals' portfolios. This result complements the findings of Canner, Mankiw, and Weil (1997) and Elton, and Gruber (2000). Finally, we have verified the existence of the two-fund separation theorem in the assets data available to the investors in our sample.;Keywords: Credit rating transition matrices, Bayesian estimation, default spread, mobility index, credit contagion, investor rationality, asset allocation puzzle, risk tolerance, separation theorem, bonds/stocks ratio.
展开▼
