Even though many researchers have found the problem of event-induced variance in event studies, they are tended to neglect these hazards by using conventional event-study methods, such as the Patell test. This test tends to reject the null hypothesis of zero average abnormal returns too often when it is true (higher type I error). In this study, we had implemented a more advanced event-study method, Boehmer, Mucumeci, and Poulsen (BMP) test, to remedy the issue of event-induced variance. Using stock dividend, the empirical findings demonstrated that the BMP test produced six significant abnormal returns from day 10 before the event to day 30 after the event while the Patell test generated 11 significant abnormal returns. In other words, the over-rejection rate in Patell test was 83.33%. At the same time, the level of significance in test values increased from 1%-5% in the Patell test to 5%-10% in the BMP test. A possible explanation for the two main findings might be due to the presence of event-induced variance. We found that the BMP test generated equally powerful tests as the null was false as well as suitable rejection rates as it was true. In addition, there has the impact of the stock dividend event on the Malaysia stock market returns. This paper provides an empirical comparison between conventional event-study methods and the BMP test to resolve event-induced variance in event studies.