Our design detects differences between groups of firms in earnings smoothing, as measured by residual earnings variability, provided that the mean level of the residuals from equation (I) docs not differ significantly between the two groups of firms. For all relevant comparisons of earnings variability, untabulatcd statistics reveal no significant differences in mean residuals for each group, In addition, the frequency with which the test observation’s residual exceeds that of its matched control yields the same inference.