Industry 'Closeness' and the Contagion Effects of Accounting Information Announcements

Abstract

Firm closeness or comparability is an important concept to investors. Knowing that two firms have been historically close and observing an information announcement by one of the firms gives the investor cues as to the future performance of the other firm. Furthermore, from a methodological point of view, researchers commonly control for firm-group closeness by using industrial classification schemes such as the SIC. To the extent that these schemes group dissimilar firms, the advantage of using these schemes is undermined.This paper more formally examines the comparability concept and develops a new measure of firm -- group accounting closeness. The measure is based on the co-movement of accounting fundamentals. This measure adds to the De Franco et al. measure in the sense that it is designed to judge the closeness of a group of firms, rather than just firm-pairs. I provide insight regarding the extent to which the SIC scheme groups similar firms. The mean SIC two-digit industry closeness over the time period 1999-2010 is 0.15 (on a scale of 0 to 1). I demonstrate that this is only slightly better than if the groups were randomly chosen. Furthermore, a trading strategy utilizing information in the measure is developed and tested. Results indicate that, when industry closeness is high enough, abnormal returns in the range of 1-3% over a 1-3 day window can be earned around leader information announcement dates. Finally, I replicate Gleason et al. (2008) and document that the industry contagion effect of an accounting restatement is increasing with the closeness of that industry.Together, these two results contribute to the contagion/information transfer literature by empirically establishing a positive relationship between the industry contagion effects of an accounting information announcement and the closeness of that particular industry. These two results also help to bolster the construct validity of the closeness measure.