A rapidly growing literature claims to reject the efficient market hypothesis by producing large estimates of long-term abnormal returns following major corporate events. The preferred methodology in this literature is to calculate average multi-year buy-and-hold abnormal returns and conduct inferences via a bootstrapping procedure. We show that this methodology is severely flawed because it assumes independence of multi-year abnormal returns for event firms, producing test statistics that are up to four times too large. After accounting for the positive cross-correlations of event firm abnormal returns we find virtually no evidence of reliable abnormal performance for our samples.

As in previous decades, merger activity clusters by industry during the 1990s. One particular kind of industry shock, deregulation, becomes a dominant factor, accounting for nearly half of the merger activity since the late 1980s. In contrast to the 1980s, mergers in the 1990s are mostly stock swaps, and hostile takeovers virtually disappear. Over our 1973 to 1998 sample period, the announcement-period stock market response to mergers is positive for the combined merging parties, suggesting that mergers create value on behalf of shareholders. Consistent with that, we find evidence of improved operating performance following mergers, relative to industry peers.

computers to make possible the non-parametric tests described in Section 3.4. Mark Hulbert was very generous with his time, advice, and data. This project was partially supported by the Richard Hermstein Fulld at Harvard University. The Equity Performance of Investment Newsletters This paper analyzes the equity-portfolio recommendations made by investment newsletters. The dataset spans 16 years, is free of survivorship and back-fill biases, and includes the recommendations of 145 different newsletters. Overall, there is no significant evidence of superior stock-picking ability for this universe of newsletters. Some individual letters do have superior performance records, but this does not occur more often than would be expected by chance, and these records are never more extreme than would be expected for the sample size. In addition, while there is some short-term performance persistence, a strategy of buying past winners does not earn statistically significant excess returns