Investor Evaluation of Accounting Information: Some Empirical Evidence

Abstract

ing from Industry Effects As mentioned before, there were some industries in which several firms switched back from accelerated to straight-line depreciation in the same year. Professor Ray Ball pointed out that such coinicidenital action might have confounded the results. Since the sample size of the depreciation group is relatively small (seventy-one) and since a small but significant proportion of a security's price variation can be explained by an industry f actor,32 he argued correctly that the ability of cross-sectional averaging to eliminate factors other than the accounting change from the may have been hampered. To make a rough measure of the possible bias introduced, we recomputed the abnormal returns for the depreciation group after taking out the industry effect for the four most heavily represented industries33 in the sample. This 32. The first and definitive study of the industry factor was by Benjamin F. King, Market and Industry Factors in Stock Price Behavior, Journal of Businiess 39, suppl. (January 1966): 139-90. 33. These industries were steels, nine firms; papers, eight firms; cement, four firms; and glass and metal containers, three firms; these comprised a total of twenty-four of the seventy-one firms in the sample. No other industry was represented by more than two firms. This content downloaded from 207.46.13.128 on Tue, 06 Sep 2016 05:43:44 UTC All use subject to http://about.jstor.org/terms 245 Evaluation of Accounting Information was accomplished by adding a third explanatory variable, the return on an industry index,34 to regression model (2). Then the abnormal return was calculated net of this industry index return (and net of the total market return and the interest rate, too). The results can scarcely be distinguished from those reported in figure 1, Panel C, and in table B1, Panel C, for depreciation changes where the effect of heavily represented industries was not eliminated.35 VI. SUMMARY Earnings manipulation may be fun, but its profitability is doubtful. We have had difficulty discerning any statistically significant effect that it has had on security prices. Relying strictly on averages, however, one can conclude that security prices increase around the date when a firm announces earnings inflated by an accounting change. The effect appears to be temporary, and, certainly by the subsequent quarterly report, the price has resumed a level appropriate to the true economic status of the firm. In the present sample, firms that manipulated earnings seem to have been performing poorly. If this is generally true, one would predict that earnings manipulation, once discovered, is likely to have a depressing effect on market price because it conveys an unfavorable management view of a firm's economic condition. APPENDIX A ESTIMATORS FOR SYMMETRIC STABLE DISTRIBUTION Let x be a random variable conforming to a symmetric stable distribution function dF(x; a, 8, s). Let A A . . . A xi 1? x2 ..?XN 34. The industry index return was defined by Ri t = loge [INDj,t/INDjt_1] where 1ND, t is the Standard and Poor index for industry i at time t. It might have been better to construct an industry index that is orthogonal to the market index (see King, n. 32 above). However, the uniformly significant industry and total market coefficients in all twenty-four cases where the industry factor was included led us to believe that multicollinearity was not a serious problem. In the twenty-four regressions, the lowest t-ratio associated with the industry effect was 1.8 and only three were below 3.0. Among the t-ratios associated with the total market return, nineteen of twenty-four were above 3.0. 35. The patterns are almost identical. The only difference is a small downward shift in the abnormal returns from weeks 10 through 60. For example, the cumulative abnormal returns (Ut's) were previously -5.058 percent in week 52 and -0.31 percent in week 38 (Ut reached a relative peak in week 38). After we abstracted from the four industries, these cumulative abnormal returns were U52 = -6.00 and U38 = -1.59 percent. If anything, this strengthens our conclusion that switching back to straight-line depreciation has little permanent effect. In addition, the temporary effect is lowered. Previously, the cumulative abnormal return increased 2.24 percent from week 28 to week 38. After we netted out the industry return; this temporary increase was only 1.85 percent. This content downloaded from 207.46.13.128 on Tue, 06 Sep 2016 05:43:44 UTC All use subject to http://about.jstor.org/terms 246 The Journal of Business be the order statistics of a random sample of size N drawn from dF(x). The parameters are a, the characteristic exponent; 6, the location; and s, the scale. We examine the problem of estimating a, 6, and s from the x's. An Estimator for s Any random variable can be standardized by the linear operation