Comparison in VAR, MARS, and PISPLINE Models for Testing Market Efficiency Hypothesis

[[abstract]]In testing MEH (Market Efficiency Hypothesis), VAR models are used in this area. For example, we could check the Granger (1969) causality result in a VAR model, and find if there is joint significance from a policy variable to a stock market. Besides, we also could check if there is any lag effect form a particular lag to the stock market. If we find any lag effect or Granger causality results, we might conclude a stock market are not consistent with market efficiency hypothesis. Since the Stock Market Efficiency (SME) hypothesis contends that there should be no significant lagged relationship between policy variables and stock markets, because current stock prices reflect all publicly available information on a policy moves, as mentioned in Darrat (1990). However, when we set a model by using past data, we seem set a general pattern for the data behavior. For example, we could set a VAR model includes how many lags in right hand side variables after choosing appropriate lags by criteria. However, it might be questioned to do this, since the behavior will change their pattern during the data period, as mentioned in Lucas's critique. Therefore, it might not appropriate to use VAR models in testing MEH[[sponsorship]]中國財務學會; 暨南國際大學國際企業系[[conferencedate]]19970628~19970628[[booktype]]紙本[[iscallforpapers]]Y[[conferencelocation]]南投縣, 臺