Asset pricing models, specification search, and stability analysis

Testing asset pricing models is closely related to specification search analysis in quantitative economics. Most specification search processes select models based on some goodness of fit statistic (such as R2 or related F). The effects of the sequential search on the statistical tests should be taken into account when looking for the maximum goodness of fit. To avoid misspecified models it is useful to study the selected models based both on the full sample and along the sample. This paper presents a conditional sequential procedure for the specification search process with linear regression models that minimizes data snooping or data mining. It is a combined test that first considers the search for the 'best' set of regressors and, conditional on this set, studies its significance and/or stability along the sample. The characteristics of the conditional tests are presented. Its efficacy is illustrated with a model of future returns as a function of past volume and returns.