Frequency domain blind multiple-input multiple-output system identification

Since commercial banks play important roles in the financial markets, it is important to evaluate whether banks operate efficiently. Moreover, given increased competition from non-bank financial institutions, commercial banks should operate more efficiently than they did previously. Commercial banks might operate more efficiently if they have superior information. If this is true, bank size should not matter to the operation of the bank. Thus, as long as the bank has superior information, it will operate more efficiently. Therefore, is it necessary that banks should be big to be efficient? Will the numbers of small commercial banks decrease? This dissertation will investigate the survival value of small commercial banks. There are three essays related to X-efficiencies in U.S. commercial banks in this dissertation. First, the relation between commercial banks' x-efficiency and agricultural factors is examined. Two hypotheses are examined in this essay. First, pairwise comparison tests and regression analyses are used to test relations between e-efficiency and bank size, location, and specialization. Second, the relation between bank X-efficiency and agricultural factors at the country level is examined. Like previous studies, economies of scale are shown to exist in the banking industry. However, the degree of scale efficiency is found to be related to loan specialization. Larger is not always better for banks with loan specialties in agriculture. Furthermore, agricultural factors, regarded as a proxy for local economic activity, have a significant impact on small agricultural banks. The second essay examines the degree to which commercial bank X-efficiencies are affected through time by monetary policy and macro economic factors. We are able to document the X-efficiency does change through time in a predictable manner. Finally, in the third essay, an improvement in the methodology for calculating bank X-efficiency is examined. The improvement is designed to solve the problems associated with using cross-sectional and time-series panel data. The improved methodology is used to re-examine the empirical results found in the first two essays.