Nonparametric Tests for Event Studies under Cross-sectional Dependence

We propose two nonparametric tests for the null of no event-induced shifts in the location parameter of the distribution of cross-sectionally dependent stock returns and develop their distributional approximation theory. We compare size and power of our test statistics to 18 tests found in the relevant literature in a controlled simulation experiment where returns are bootstrapped from the constituents of the S&P100 and NASDAQ stock indexes. One of our tests turns out to be the most reliable in terms of size, while in terms of power, the other test we propose ranks in the top group of tests with an acceptable actual size. The test comparison presented in this paper is one of the most comprehensive and up-to-date available in the literature. Some tests in the comparison have been adapted to the cross-sectional dependence case by us. A real-world application of the tests to assess how U.S. banks' stock prices have reacted to early warnings of the 2008 financial crisis completes the article.