EMPIRICAL TESTS OF EARNINGS QUALITY
The theoretical problems have not deterred researchers from attempting to rank the quality of competing earnings measures. The goal of the empirical research is to determine which measure works best in practice. This task requires, of course, a definition of what it means to “work best.” In the literature, the following criteria have been used.
Value Relevance Method
The value relevance criterion is based on separate regressions of market price on each of the competing earnings measures.8 In these regressions, the best measure is defined as the one that yields the highest adjusted R2, the most significant slope coefficient, or the slope coefficient most consistent with a predicted amount (Barth, Beaver, and Landsman 2001). Depending on the research question and statistical issues, changes in the variables (rather than the level of the variables) may also be used in estimating the regression (Landsman and Magliolo 1988) and possibly data over long observation intervals (Dhaliwal, Subramanyam, and Trezevant 1999). Value relevance can also be defined in incremental terms. In that case, various measures (their levels or changes in their levels) are separately added to a regression of market price that already includes variables other than the competing measures. The best measure is the one that yields the greatest incremental explanatory power.
Information Content Method
The information content criterion is based on the response of stock prices, ...