This study is motivated by the development of credit-related instruments and signals of stock price movements of large banks during the recent financial crisis. What is common to most of the empirical studies in this field is that they concentrate on modeling the conditional mean. However, financial time series exhibit certain stylized features such as volatility clustering. But very few studies dealing with credit default swaps account for the characteristics of the variances. Our aim is to address this issue and to gain insights on the volatility patterns of CDS spreads, bond yield spreads and stock prices. A generalized autoregressive conditional heteroscedasticity (GARCH) model is applied to the data of four large US banks over the period ranging from January 01, 2006, to December 31, 2009. More specifically, a multivariate GARCH approach fits the data very well and also accounts for the dependency structure of the variables under consideration. With the commonly known shortcomings of credit ratings, the demand for market-based indicators has risen as they can help to assess the creditworthiness of debtors more reliably. The obtained findings suggest that volatility takes a significant higher level in times of crisis. This is particularly evident in the variances of stock returns and CDS spread changes. Furthermore, correlations and covariances are time-varying and also increased in absolute values after the outbreak of the crisis, indicating stronger dependency among the examined variables. Specific events which have a huge impact on the financial markets as a whole (e.g. the collapse of Lehman Brothers) are also visible in the (co)variances and correlations as strong movements in the respective series.