The specification and power of tests to detect abnormal changes in corporate investment

We evaluate methods used to measure abnormal changes in capital expenditures. We examine both statistical tests and models of expected capital expenditures. We find that commonly used research designs yield test statistics that are misspecified, even in random samples. In cases where sample firms share a common characteristic such as extremely low or high investment, size, leverage, return on assets or market-to-book ratio, it is very important to match sample firms to a control group that shares this pre-event characteristic. We also find that using control groups, rather than a single control firm, yields more powerful test statistics.