Accounting information and analyst forecast errors: A study of the explanatory power of discretionary accruals and accruals quality
I investigate links between two commonly employed measures of accounting information, discretionary accruals and accruals quality, and errors in sell-side analysts’ forecasts of firms' quarterly earnings per share. I measure analyst forecast error as the difference between the mean forecast and actual earnings as reported by First Call. I improve upon prior studies' measures of discretionary accruals and accruals quality in two ways. First, the regression equations that I use in estimating both measures use quarterly, rather than annual, financial statement data. I convert the quarterly data to trailing twelve month (four quarter) data in order to run the regressions necessary to estimate discretionary accruals and accruals quality. Second, I use random coefficients regression, rather than ordinary least squares regression, to model accruals in estimating discretionary accruals and to model certain changes in working capital in estimating accruals quality. My use of random coefficients regression allows me to leverage firm-specific and industry-specific information, as well as information about my entire sample of firm quarters. I find that discretionary accruals tend to decrease as the absolute size of analyst forecast error increases. One possible reason for this finding is that firms record income-decreasing accruals as their earnings deviate from analysts' forecasts by large amounts, either positive or negative. I find that analyst forecast errors tend to grow larger, in absolute terms, as accruals quality decreases, suggesting that analysts have a relatively difficult time forecasting firms’ earnings when those firms’ accruals are of relatively low quality. I find that discretionary accruals and accruals quality are both useful in explaining variation in analyst forecast error levels.