|
Abstract:
challenge to market efficiency. However, an equal number of
studies have challenged these anomalies, arguing that they are
merely symptoms of data-snooping biases and overfitting. In this
paper, we propose several methods for assessing the statistical and
economic significance of financial anomalies.
We begin by providing a critical review of the financial
anomalies literature, categorizing anomalies systematically and
performing citation counts to evaluate the importance of each
anomaly in the literature. We then develop methods---both classical
and Bayesian decision-theoretic---for determining the statistical
significance of an anomaly after taking into account the fact that
anomalies are often obtained by extensive data-mining algorithms.
We conclude with with an illustrative empirical example involving
multifactor linear models of US stock returns.
Work done in collaboration with Li Jin (ljin@mit.edu).
|
