Results (
Arabic) 1:
[Copy]Copied!
We then incorporate the same sample reductions as DGM 2004, first deleting4,166 firm-year observations whose differences between their I/B/E/S consensusforecasts and actual earnings per share (EPS) are not within five cents per shareand then eliminating 6 observations with total assets of less than $10 million. Next,we delete 310 observations that are within the top and bottom 1 percent of the distributionsofETRQ3,ETR4_ETR3, andInduced_Chg_ETR. Then, we remove4,125 observations whose earnings absent tax expense management are not withinfive cents per share of their consensus forecasts, leaving 3,035 observations.Although our sample is smaller than the 4,656 firm-year observations of DGM2004, who use observations from 1986 to 1999, our sample is proportionatelylarger because of the number of sample years in our study. We add auditor-providedtax service fee data from Audit Analytics, reducing the sample by an additional1,056 firms, thereby leaving 1,979 firm-year observations. Finally, we follow Millset al. 1998 and scale firms’ tax fees paid to their auditors by selling, general, andadministrative expense from COMPUSTAT, eliminating an additional 177 observationsand yielding a final sample size of 1,802. Of these 1,802 firm-year observations,1,473 reported positive amounts of tax fees paid to auditors, and the remaining 329reported values of $0 for this field. As we discuss in the next section, values of $0indicate that firms did not pay tax fees to their auditors and purchased tax servicesfrom nonauditors or conducted in-house tax planning instead
Being translated, please wait..