Simulation approach to evaluate the statistical power of different statistics tests and return-generating models in the Mexican stock market
Description
The propose of this dissertation is to test the efficacy of event study methods in the context of an important emerging market: Mexico The dissertation answers the following five questions: (a) Which is the most appropriate model to be used with the event study technique into the Mexican market for the following methodologies: Mean Adjusted Returns, Market Adjusted Returns, and the Market Model? (b) Which of the following tests---the T-test, Wilcoxon test, Sing test, and Corrado (1989) test---is the most appropriate test for each of the previously mentioned methodologies? (c) How should we measure the returns to obtain the best results with the event study technique? (d) How will our results be affected by the methodologies, tests, and returns used in the study? (e) What similarities and differences can we find as a result of the simulation process used between the U.S. market and Mexican market? One of the most frequent questions asked to managers or stockholders is: What is the effect of a specific decision on the price of a financial asset. The study of events is a relatively simple methodology used to answer a question like this In order to use such a methodology, we must have a model that describes the 'normal' behavior of the price of an asset through time. If exists difference between the model, and the market data, it says that a event exist. The analysis of the differences is what constitutes the study of events technique We need ex ante generators, normal expected return Rit of the security i in the time t and compare them with their actual ex post return For our work, the selected universe for E [Rit|Xt] are the returns produced by the prices of the 101 most negotiated stocks on the Mexican Stock Exchange from January 2, 1987 to March 2, 1998; such returns are considered in weekly and monthly time intervals In order to do comparisons we take random 101 stocks of the New York Stock Exchange from January 1987 to March 1998 in monthly intervals. The models we use to generate the Rit are: Mean Adjusted Returns, Market Adjusted Returns, Market and Risk Adjusted Returns The tests we considered are the t-test that is a parametric test and the non-parametric tests: sign-test, Wilcoxon test and Corrado test Once known the model with its parameters, a stock portfolio is made up and we compute epsilonpt and with the tests the null hypothesis is validated. H0: No abnormal returns The variable that concerns us in this paper, is the power of each one of the three methodologies, that is, the value of the conditional probability. (Abstract shortened by UMI.)