زمان بندی صنعت و اطلاعات درآمدی شرکت در قیمت های امنیتی: ارزیابی مجدد

This paper re-evaluates evidence in Ayers and Freeman [Ayers, F., Freeman, R., 1997. Market assessment of industry and firm earnings information. Journal of Accounting and Economics 24, 205–218] suggesting that investors anticipate industry-wide components of earnings earlier than firm-specific components, and that post-earnings-announcement drift following annual earnings announcements is due primarily to firm-specific components of earnings. Our tests indicate that post-announcement drift is entirely attributable to coefficient bias due to measurement errors in the use of realized earnings changes as proxies for unexpected earnings. Also, coefficient differences in the market's anticipation of subsequent-year industry and firm-specific earnings become insignificant when we introduce suitable controls for non-linearity in the return/earnings relation.

The breakdown of security returns into macro-economic, industry and firm-specific components has been a recurrent issue in studies of the timing of the dissemination of valuation-relevant information in financial markets (Roll, 1988; Morck et al., 2000; Bhojraj et al., 2003). Recently, similar disaggregations of reported earnings changes have been employed in accounting research that addresses the timing of the security price effects of earnings information (Ayers and Freeman, 1997, Ayers and Freeman, 2000 and Ayers and Freeman, 2003; Piotroski and Roulstone, 2004). Generally, these accounting inquiries evaluate the premise that investors anticipate aggregate (market-wide and/or industry) earnings components earlier than firm-specific earnings components. In addition, the more aggregate market-wide and industry components of earnings innovations are expected to be more persistent, and consequently to have larger capitalization factors than firm-specific components. Ayers and Freeman (1997), hereafter AF (1997), evaluate whether security prices anticipate industry earnings components earlier than firm-specific earnings components, based on two main hypotheses. First, AF (1997) propose that, if security prices exhibit a delayed response to earnings announcements, the resulting post-announcement drift in returns will be more significant for firm-specific earnings components than for industry earnings components. This expectation reflects the intuition that industry-level information is widely available from a variety of public sources, and its valuation implications are readily recognized and quickly impounded in security prices. In contrast, firm-specific information is assumed to be less timely, and its valuation implications less transparent, and thus is more likely to be reflected in security prices with a delay. Second, AF (1997) propose that the extent to which investors anticipate the earnings of the subsequent period will differ for the industry and firm components of earnings. Specifically, lead earnings response coefficients (the slope coefficients from regressions of current-year returns on subsequent-year earnings changes), which indicate investors’ anticipation of future earnings, will be larger (smaller) for industry (firm-specific) components of earnings. AF (1997) provide empirical results, based upon returns–earnings regressions of current year security returns on lagged, current and subsequent-year industry and firm earnings components, that strongly support both hypotheses. Current-year security returns are significantly positively associated with the prior year's firm-specific earnings changes. In contrast, current year security returns exhibit no significant relation to the prior-year's industry earnings changes. In other words, investors appear to have fully reflected the industry component of annual earnings changes in returns by the earnings announcement date, but appear to have a delayed reaction to the firm-specific component of earnings changes. In addition, the estimated slope (earnings response) coefficients relating current year security returns to the subsequent year's earnings change components show that the industry component has a significantly higher coefficient than does the firm-specific component. This relation is interpreted in AF (1997) as further corroboration that investors anticipate the industry component of the subsequent year's earnings changes earlier than the firm-specific component. This paper re-evaluates the evidence in AF (1997) of differential timing of industry and firm components of earnings changes in security prices. Two related attributes of the evidence in AF (1997) motivate this re-evaluation. First, the specification of the returns–earnings relation in AF (1997) employs realized changes as proxies for unexpected innovations in the earnings components. The extant literature, however, documents a negative serial dependency between successive earnings changes (Abarbanell and Bernard, 1992; Elgers and Lo, 1994; Fama and French, 2000). To the extent that investors’ earnings expectations impound these serial dependencies, there is measurement error in the earnings change variable as a proxy for unexpected earnings. Moreover, earnings changes in a given year are correlated with the measurement error in the subsequent year, in a manner that biases the regression estimates in a direction supporting the premise that security returns exhibit a delayed market reaction to earnings changes. Further, because these serial dependencies are more (less) pronounced for the firm-specific (industry) components of earnings changes, the coefficient bias is more severe for the firm-specific components. The greater relative bias in the measurement of the firm-specific components potentially provides spurious support for the hypothesis that post-announcement drift is attributable primarily (or entirely) to the firm-specific component of unexpected earnings. A second factor motivating our re-evaluation of the AF (1997) findings is the non-linear relation of earnings surprises to abnormal security returns (Freeman and Tse, 1992). This non-linearity implies that a linear specification of the returns–earnings relation imparts a downward bias to estimated earnings response coefficients. The extent of the downward bias varies directly with the variance of the unexpected earnings measure. We expect the industry components of earnings changes to have substantially lower cross-sectional variances than do the firm-specific changes, because industry components are measured as average earnings changes for multiple firms. Consequently, the expected downward bias in the coefficient estimate is greater for the firm-specific component of earnings changes, and for this reason we expect to find a higher coefficient for the industry component of subsequent earnings changes. The higher coefficient, however, is not interpretable as evidence that industry components of subsequent earnings changes are impounded earlier than are firm-specific components. AF (1997) recognize this potential bias and attempt to control for the non-linear response function by imposing a common truncation rule on both industry and firm components. Such an approach, however, is unlikely to equate the variances of the earnings components in this setting, and a larger variance for the firm component continues to bias the tests against the null hypothesis of coefficient equality. Note that the two factors motivating this study are not independent phenomena. The non-linear relation between earnings surprises and security returns reflects the differential persistence between “extreme” and “non-extreme” earnings changes. Extreme earnings changes have lower persistence, evidenced by their negative serial dependency. The larger the cross-sectional variance in earnings changes, the greater is the negative serial dependency in earnings changes, and the greater is the measurement error in using current earnings as proxies for investors’ expectations of subsequent earnings. The remainder of this paper is organized as follows. Section 2 describes our research design and presents the empirical results. These results indicate that the primary relations in AF (1997) replicate well in our larger and more recent sample: subsequent year abnormal security returns are significantly positively associated with firm-specific earnings changes, i.e., there is an apparent post-announcement drift in investors’ response to annual earnings changes. Also, the industry component of the subsequent year's earnings changes has a significantly higher anticipatory returns response coefficient than does the firm-specific component. Our further tests, however, show no evidence of significant bias in investors’ weightings of industry and firm-specific earnings changes. In addition, we demonstrate that the apparent post-announcement drift is entirely attributable to measurement errors in the use of realized earnings changes as proxies for unexpected earnings. Also, the coefficient differences in the market's anticipation of subsequent-year industry and firm-specific earnings become insignificant when we introduce suitable controls for non-linearity in the return–earnings relation. Moreover, we show that the primary relations reported in AF (1997) are evident in the majority of cases when “industry” designations are randomly assigned to the sample firms. Supplementary tests, however, indicate significant timing differences between investors’ valuations of non-extreme and extreme firm-specific earnings changes in security prices. Although annual earnings information appears to be fully impounded in security prices by the earnings announcement date, relatively extreme (non-extreme) changes in firm-specific earnings are reflected in security prices later (earlier) in the earnings year. The extremity of earnings changes, rather than the firm-specificity of earnings changes, underlies the observed differences in the timing of earnings information in security prices. Section 3 summarizes the implications of these findings, and concludes the paper.

This paper re-evaluates earlier evidence in AF (1997) suggesting that investors anticipate industry-wide components of earnings changes earlier than firm-specific components, and that post-earnings-announcement drifts in security returns for annual earnings changes are due primarily (or exclusively) to the firm-specific components of earnings changes. Our tests indicate that the apparent post-announcement drift is instead attributable to measurement errors in the use of realized earnings changes as proxies for unexpected earnings. Serial relations in successive years’ earnings changes, to the extent impounded in investors’ earnings expectations, bias the coefficients in multiple regressions in a manner consistent with a post-announcement earnings drift. Moreover, coefficient differences in the market's anticipation of subsequent-year industry and firm-specific earnings become insignificant when we introduce suitable controls for non-linearity in the return/earnings relation. In addition, we show that the primary relations reported in AF (1997) are evident in the majority of cases when “industry” designations are randomly assigned to the sample firms. Supplementary tests, however, indicate significant timing differences between investors’ valuations of non-extreme and extreme firm-specific earnings changes in security prices. Although annual earnings information appears to be fully impounded in security prices by the earnings announcement date, relatively extreme (non-extreme) changes in firm-specific earnings are reflected in security prices later (earlier) in the earnings year. It appears that the extremity of earnings changes, rather than the firm-specificity of earnings changes, underlies the differences observed by AF (1997) in the timing of earnings information in security prices. The tenor of our results does not reduce the plausibility of the intuitions in AF (1997) that, because industry-level information is widely available from a variety of credible public sources, its valuation implications should be readily recognized by investors and quickly impounded in security prices. Conversely, because firm-specific information may be less timely, and its valuation implications may be less transparent, such information may be reflected in security prices with a delay. The fact that we do not document these intuitions empirically, after suitable controls for measurement errors and non-linearities, may reflect difficulties in the specification and measurement of industry components in accounting earnings. Measurement of the industry component rests on judgments about how to define industry membership, the level of aggregation in classifying firms into industries, the manner of weighting individual firms in order to construct industry indexes, the classification of firms engaged in multiple lines of business, whether commonalities among firms should incorporate financing as well as operating features, the functional form and parameters of the relation between a given firm and its industry and other factors. Recent research examines a variety of alternative classification schemas to explain commonalities in security returns as well as in accounting fundamentals (Bhojraj et al., 2003; Piotroski and Roulstone, 2004). This recent line of inquiry may in time improve our ability to separate accounting earnings into firm-specific and more aggregate components that are homogeneous in terms of information timing in security prices.