The Impact of Cross-Validation Adjustments on Estimates of Effect Size in Business Policy and Strategy Research

Abstract

The authors examined use of cross-validation in multiple regression research in organization studies by reviewing research published in three top journals in business policy and strategy (BPS) over 6 years. Application of a formula method of cross-validation to each study showed that shrinkage was often large. In some cases, shrinkage was greater than 50% of R2 when sample sizes were less than 90. The authors recommend that organization studies researchers use formula cross-validation to improve the validity of inferences while avoiding the additional data collection and split samples required by empirical methods.