Event study methodology for the Borsa İstanbul s

Başdaş, Ülkem
The primary research question of this thesis is to try to determine the appropriate event study methodology for studies carried out on the Borsa Istanbul. In order to find the most appropriate methodology we compare the performance of different models (mean adjusted returns, market adjusted returns, market model) in the Turkish Stock Market with two parametric (portfolio time-series standard deviation test, Patell test) and two non-parametric tests (generalized sign, generalized rank tests) under different return definitions (log versus arithmetic), sample sizes, event windows, and clustering. Also, the sensitivity of results to time period, different databases (Datastream versus Matriks) as well as statistical tools (Excel macros versus Stata) are considered. This thesis basically follows the experimental design of Brown and Warner (1980; 1985) (BW, henceforth) but modifies the test statistics in line with the current developments. According to the results on Turkish stock market data of 471 securities over 1988-2012, similar to the findings of BW, the mean adjusted returns do not cause a severe specification and power problem under certain circumstances, but in case of clustering, the results suggest not to use the mean adjusted returns. In most of the cases crude adjustment test is well-specified. Besides, samples with larger number of securities and shorter event windows are preferred for the power of tests. Shortening the time period does not affect the results whereas using a different database can cause changes in specification and power.
Citation Formats
Ü. Başdaş, “Event study methodology for the Borsa İstanbul s,” Ph.D. - Doctoral Program, Middle East Technical University, 2013.