The Risk—Return (Bowman) Paradox and Accounting Measurements

AbstractWe show that the negative relationship found by some researchers between firms' mean accounting returns and their standard deviations (the Bowman paradox) may be spurious. We show that this relationship becomes either positive or insignificantly different from zero following two adjustments. One adjustment is using the beginning-of-year rather than the end-of-year equity as the return on equity (ROE) denominator. The second is the adjustment of reported net income for accruals. We use a longer-period sample than the samples that are used in previous studies, but demonstrate that our results hold even when we use a subsample that corresponds to the period used in previous studies. We also show that our results are robust to a survivorship bias.