Performance persistence through the lens of chance models

To what extent will performance differences persist? Prior studies have suggested that a large proportion of performance variances cannot be explained by systematic factors. The unexplained variance, i.e., chance implies that one should not expect performance differences to persist. In particular, more extreme performances are likely to regress more to the mean. We empirically examine this statistical account of performance persistence in six datasets: National Football League (2001-2016), National Basketball Association (2004-2017); Major League Baseball (1901-2016), Formula One Racing (1996-2015), return on assets of US firms (1980-2010) and Fortune500 firms (1955-2005). Strong regression effects appear in all datasets and we found a trend of decreasing performance persistence of exceptionally performing firms over the past thirty years. Moreover, the effects of regression can be so strong that they generate non-monotonic performance associations and the contexts determine whether and where these rank reversals occur. Chance models that incorporate the contextual factors are developed to account for these anomalies. We conclude by discussing the implications of chance models for theory building and testing in management and how non-monotonic performance associations imply an alternative source of strategic opportunities with illustrations from sport betting