| Summary: | 碩士 === 國立暨南國際大學 === 經濟學系 === 95 === The conventional market timing tests are often used to evaluate the accuracy of the event forecasts. Under the IID assumption of event forecasts, these tests perform well. However, all of these tests have servere size distortions when the event forecasts are serially correlated. In this thesis, the method of heteroskedasticity-autocorrelation consistent (HAC) is employed to cope with this problem. After accounting for the serial correlation, the HAC robust tests including fixed-b approximation and naive block bootstrap can effectively reduce the over size problem. In the empirical study, we find that conventional tests over reject the null when the event forecasts and event realization of all countries have high serial correlation, which means the event forecasts have the timing ability. According to the evidence of sample correlation coefficient, however, it seems that the event forecasts have no timing ability by using these the HAC robust tests. Therefore, the HAC robust tests are strongly suggested to evaluate the event forecasts even when the event forecasts have serial correlation.
|
|---|
