We obtain our sample of ERP system adopters from a proprietary database supplied
by a leading international provider of ERP systems. The unique dataset is a record of all
license agreements sold by the company to U.S. firms from 1993 to 1999.6 The dataset
includes the name of the firm that purchased the license, the start date of implementation,
the date when the installation was complete and the system went live, and the modules
(e.g., financial accounting, human resources) that the firm implemented. This dataset is used
for real operational decisions at the ERP provider; thus we believe that the dataset is
extremely accurate.
We begin with 625 unique firms that purchased the ERP system and started the implementation
process between 1993 and 1999. The years when the ERP systems went live
range from 1994 to 1999. The average implementation period was 1.73 years. We searched
the Ticker IDs for these firms on COMPUSTAT by firm names, and 315 firms that did not
have Ticker IDs were removed from our sample.7 As described below, we use the absolute
value of discretionary accruals as our measure of earnings management. Therefore, data
from COMPUSTAT needed to estimate discretionary accruals using the cross-sectional
modified Jones model (Dechow et al. 1995) are required. Earnings announcement dates,
available on COMPUSTAT, must also be available as we use reporting lag (i.e., days between
the fiscal year-end and earnings announcement date) in our analysis of earnings
release date management.