Survivor Bias In Firm Specific Longitudinal Studies: The Case Of ERP Systems

Abstract

Researchers in the field of accounting and finance often use event studies to measure the impact of various business decisions on the market value of stock.  They argue that market reaction to the event is predictive of the future effect on the company.  Sometimes researchers would like to follow up on these initial event studies to see if the market was correct in its predictions.  However, many times the original sample used for the event study does not survive intact into future time periods, which then raises the issue of survivor bias.  This paper examines one such event study in which the markets react favorably to announcements that ERP systems are being implemented.  However, only 55% of the firms in the original sample survive through 2009, making any follow-up study subject to the survivor bias argument.  The study examines the circumstances related to the non-survivors and, using proxies for the missing data, concludes that the results of a follow-up study would not have been impacted by survivor bias.