A B S T R A C TWe employ Vector Autoregressions and quarterly financial statement data to explicitly test for both own-and cross-industry earnings management "spillovers". We find that earnings management behavior can transmit from one industry to another, similar industry. Further, we find an indirect, information-based channel whereby earnings management information may reach less-related industries with a lag of about six months. These results provide initial evidence that managers within similar industries "herd" in their discretionary accrual strategies. That is, earnings management in one sub-industry classification provides an informative signal to other industries. Others, then, follow suit by adopting like strategies either for competitive parity or purely wealth-maximizing purposes. In addition, we find that earnings management is persistent and does not mean-revert. Thus, earnings management can have long-term permanent impacts on a firm and its financial statements. This paper presents both methodological and policy implications for the academic accounting literature.
Keywords: Earnings Management, Accrual Strategy
Ⅰ. IntroductionFormer SEC Chairman Levitt commented that earnings management was a "widespread" numbers game and that, if not addressed soon, would have adverse consequences for America's financial reporting system (Levitt 1998). Chairman Levitt's warnings are particularly relevant given that some investors completely rely on firms' reported financial statements, at their face value, due to their inability to process more sophisticated (i.e., footnote) information. In this paper (708) 534-4958 mwilliams15@govst.edu we examine two aspects of earnings management behavior. The first aspect is whether earnings management is (own) persistent or mean-reverting across time. The second aspect is whether, after controlling for own-persistency in earnings management, can earnings management in one industry "spillover" to another, related industry in a lead-lag fashion. Our results can help investors partially "undo" earnings management and hence have an unobstructed view of a firm's financial health.Unlike the prior, disjoint literature on these two issues, this study uses a modeling framework which simultaneously allows for both own-to-own and cross-to-own earnings management impacts across time. Additionally, the modeling framework explicitly incorporates time-lags clarifying causality unlike prior literature which uses effectively-contemporaneous, instrumental variable, two-stage least squares approaches. Using our methodologically-correct approach, we find evidence of both own-and cross-earnings management spillovers across time.