Estimate of the Capital Assets Pricing Models by means of Kalman filter

The Capital Assets Pricing Model (CAPM), which was published by W. F. Sharpe and J. Linter in the middle of the sixties, has since that time grown to one of the piers of foundation of the financial economics. During the time it used to be empirically tested for several times, but these tests in most of the cases contradicted its validity - especially (since as early as the seventies) rose the doubt about the time stability of the coefficient. Hence many economists have tried hard to find a new model, which could concisely express the progress of this coefficient. In have focused on three basic models in my thesis - they are the Model with Random Coefficients, the Random Walk and the Mean Reverting Model. I estimated these models for selected share issues from Prague Stock Exchange and New York Stock Exchange by Kalman filter and, finally, I tried to make a confrontation of all those models mentioned above. It is quite clear, that as for the sequel from empirical estimation, there always exists at the least one model with variable parameters, which better (in quite a concise way) describes to behaviour of the coefficient than the standard model CAPM with constant parameters