Analyzing Returns in the European Banking sector in the context of the Default of Credit Suisse

Abstract

The present study tests the predictive capabilities of the Capital Asset Pricing Model and the Fama French Three Factor Model, two of the most popular asset pricing theories. The CAPM is a single factor model that dictates the relationship between risk and return, the FF3FM is a three factor model with a component of market risk but also has two other factors, SMB (small minus big) and HML (high minus low), that together try to explain the returns of a certain security or portfolio. The forecasting power of both models will be tested in five stocks of major European Investment Banks, one of which is Credit Suisse that defaulted in March of 2023. In order to test the models, weekly historical returns from June 2017 to June 2022 will be used to run linear regressions which will allow for the estimation of the necessary coefficients to be applied in the empirical models and therefore test their prediction capacity. Results suggest that, although some approximation between the real returns on the stocks and the forecasted returns, the reality is that these models were not able to accurately predict the market movements of the stocks neither the default of Credit Suisse. In the CAPM, all the beta values were deemed necessary to explain a portion of the model, even though without the best accuracy. The perhaps more interesting result came with the FF3FM outputs. For all stocks, while the beta coefficient that was related to the market risk was considered statistically significant, the same can not be said for the other two beta coefficients related to SMB and HML. Not one coefficient of the SMB factor was proven to be statistically significant and regarding the factor HML, only BNP Paribas and Santander had significant beta coefficients for it. The difference between the forecasted stock returns values with the CAPM and the FF3FM was very little, almost insignificant. The market risk beta coefficients are very close between the two models, the R 2 value is slightly higher in the FF3FM and the impact of the two statistically significant betas related to the HML factor is residual.