We analyze the size and power of a large number of “robust” asset pricing tests, investigating the hypothesis that the price of risk of a candidate factor is equal to zero. Different from earlier studies, our bootstrap approach puts all tests on an equal footing and focuses on sample sizes...