Scale effects in capital markets-based accounting research

Based on data simulated using a modified Ohlson (1995) valuation model, we investigate effects on inferences of five potential scale-related effects: multiplicative and additive omitted scale factors, scale-varying coefficients, survivorship, and heteroscedasticity. We find that diagnostics identified in prior research are not successful in detecting or distinguishing these scale effects. Thus, we investigate the effectiveness at mitigating scale effects of six specifications of regressions of equity market value on equity book value and earnings: undeflated, share-deflated, equity book value-deflated, lagged price-deflated, returns, and equity market value-deflated. For each specification, we compare frequency of correct rejection that the coefficients equal zero, coefficient bias and absolute error, and regression explanatory power. We find that share-deflated and undeflated specifications generally perform the best, regardless of the type of scale effect.