dc.description.abstract | Unlike the earlier research, which used the conventional univariate regression to test the forward rate unbiased hypothesis, we adopt 4 panel unit root tests to examine forward premium series of 6 currencies over panels of 1-, 2-, 3-, and 6-month forward contract maturities. In the beginning of the thesis, we briefly present and review the LL, IPS, Fisher P_lambda and panel LM test, which consider the structure change of data, to realize the motivation of adducing these methods, and their original meanings, then we discuss and analyze our empirical findings. Not only compare the differences between panel unit root tests and the univariate ADF-GLS test, we also observe and compare the outcome of 4 panel unit root tests,
therefore we can understand how NT dollar vary in the foreign exchange market from 1992 to 2002,
by contrasting with other currencies.
From the test results we find not only the univariate ADF-GLS test is different to the 4 panel unit root tests, but also, the 4 panel unit root tests have different outcome, either. In the results of ADF-GLS test, no matter with or without the time trend term, British pound (BP), Euro, Canadian dollar (CD) and Japanese yen (JY) all can reject the null of unit root.
However, in the results of LL, IPS and Fisher P_lambda test, almost all currencies can reject the null except Euro. The reason of making the huge difference between the 2 kinds of tests might be by using the panel model, which can include more information, can effect the conclusion of statistical inference.
Furthermore, by applying panel LM test with allowing structure breaks of data, we find no matter with or without the time trend term, CD,
Singapore dollar, JY and NT dollar can reject the null hypothesis at 5% significance level.
However, BP, which can reject the null in LL,
IPS and Fisher P_lambda test, almost cannot reject when considering the structure change of data. | en_US |