Although the order of magnitude of the residual variances is similar to that in Lang, Raedy, and Wilson [2006], the variances are not directly comparable between the two studies because our change in net income regression includes four additional control variables not included in the prior study-AM), NUMEX, XIJST, and CLOSE. Untabulated results indicate that each of the four variables provides incremental explanatory power to the model, thereby reducing unexplained variance