Presentation date: 2007-06-21Graduation date: 2008This study has two main objectives. First, we propose an alternative way for treating deposits in modeling a banking firm, which account for both their input and output features. Second, we contribute to modeling failures in the banking sector by distinguishing three groups of factors affecting failures: bank level, industry level and economy-wide level, recognizing the risks associated with these factors. We apply both models to a data set of Russian banks, spanning 1999-2004. Traditionally researchers assumed that deposits are either an input, used to generate loans (intermediation approach) or an output, a service that a bank provides, utilizing labor and capital (production approach). In Chapter 2 we propose to account for both input and output characteristics of deposits by introducing a substitution effect. In the framework of non-parametric Data Envelopment Analysis we maximize deposits, just like other outputs, while introducing the possibility of substitution between deposits and other borrowed funds, an input. Even though we did not find evidence that the results of our model are significantly different from the other two approaches, it is still preferred, since it provides a more general way of treating deposits: both production and intermediation models can be deduced from it. Chapter 3 extends existing literature on modeling bank failures. We model failures as a function of different risks that a banking firm faces. We argue that a bank fails if cumulative risks exceed an unobserved critical level and use a binary response model to carry out our empirical estimation for a sample of Russian banks. We add the efficiency metric from Chapter 2 to our data set and use it as a proxy for managerial quality. We also adjust for the fact that bank failures represent rare events as suggested by King and Zeng (2001). We found that higher deposit and liquid assets balances, as well as efficiency (banks-specific variables) were crucial in affecting failures, while macroeconomic and industry-level variables appeared to be not as important.