[The sequential treatment adopted in many multiple signal studies is consistent with a simultaneous treatment and with other apparently diverse sequential approaches. More importantly, the conventional residual security return research design, which is also a sequential approach, can lead to downward-biased estimates of the incremental explanatory power of the accounting variables introduced at the second stage. This finding is striking in that it implies that the most pervasive use of a sequential approach is likely to lead to biased results.]
Journal of Accounting and Economics199926(1-3), 35-42
The comment reviews major features of the research design of the DMS study. Several opportunities for future research are discussed, including the exploration of the underlying information variables that influence analyst's forecasts.
The time series behavior of earnings is an important area for empirical research because of its implications for related research in several areas of and finance. Although many other examples could be provided, three accounting issues immediately come to mind: (1) income smoothing, (2) the relative forecast ability of alternative income measurements, and (3) interim reporting. The hypothesis that management uses discretionary practices to smooth income was first posited by Gordon [21] and later tested by Gordon, Horwitz, and Meyers [22] and by Copeland [14] among others. As stated by Gordon, smoothing involves minimizing the deviations of reported income from some standard, where the standard is defined in terms of normal income. Normal income has never been precisely defined at the conceptual level, but in many cases it appears to have been used in the sense of the expected value of the process at a given point in time. A variety of models could be used, and in fact have been used, to assess the normal or expected value of income for a given period. Each model makes specific assumptions about the process generating income numbers. Any inferences drawn from empirical evidence regarding the existence of income smoothing (or the lack of it) are dependent upon the validity of the assumptions made about the underlying earnings process. Moreover, as shown later in the paper, for certain processes attempts to smooth income can have exactly the opposite effect. Yet the models used in the smoothing literature represent only a narrow range of the possible alternatives, little justification (either a priori or empirical) has been offered in their behalf, nor has there been any direct, rigorous investigation of the underlying nature of the earnings process itself.
These remarks provide some perspective on my six decades of research, teaching, and participation in the AAA. A recurring theme is that my career took several unexpected turns and that my research often had unexpected outcomes. Several areas of research are discussed, including the prediction of financial distress, the information content of earnings announcements, the information content of prices, accounting and market measures of risk, discretion in financial reporting, conservatism, and value relevance of financial statements. Included is a brief summary of some of what I have learned from six decades of teaching. I review some of the major benefits of AAA participation.
The sequential treatment adopted in many multiple signal studies is consistent with a simultaneous treatment and with other apparently diverse sequential approaches. More importantly, the conventional residual security return research design, which is also a sequential approach, can lead to downward-biased estimates of the incremental explanatory power of the accounting variables introduced at the second stage. This finding is striking in that it implies that the most pervasive use of a sequential approach is likely to lead to biased results.