To provide evidence on whether firms who engage in REM affect the information
content of their earnings, this study uses a large sample of 19,356 firm year observations
from 1989-2009. These observations are ranked into deciles based on two types of
common REM – abnormal production costs and abnormal discretionary expenses. Firms
which display abnormally high levels of production costs or abnormally low
discretionary expenses relative to other firms in their industries are considered to have
high levels of REM. OLS regression of current period returns on multiple measures of
earnings, an indicator variable based on high levels of REM, and interactions of the
indicator variable and the various earnings variables, as well as control variables is then
used to compute ERCs. This is specifically designed to examine the marginal effect of
REM on the ERCs of high REM firms. The model is run once with abnormal production
costs as the basis for the high REM indicator variable and once with abnormal
discretionary expenses as the basis for the indicator variable. Results for both iterations
of the model suggest that firms which engage in REM do impact the information content
of their reported earnings. Furthermore, results suggest that these high REM firms are
adding noise to their reported earnings and, therefore, degrading the information content
of those earnings.