Tests of the Hypothesis

2403 Where: Earn Decrt+1 : a dummy variable equal 1 if a firm experienced a reduction in earnings before extraordinary items in year t+1 as compared to year t. Div Decrt+1 : a dummy variable equal 1 if a firm experienced a reduction in dividends per share in year t+1 as compared to year t. CONSt: conservatism that measure according to Ball and Sivakumar 2005 Kim and Pevzner 2008: coefficient βγ from the following regression, multiplied by -1. OIt = αt + β1 D+ ββ OIt-1 + βγ DOIt-1 +et Where: OIt is change in operating income in year t deflated by beginning market value of equity. OIt-1 is change in operating income in year t-1 deflated by 2404 beginning market value of equity D is 1 if _OIt-1 is negative and 0 otherwise. Control variables: ROAt : year t return on assets. EARNt : year t change in earnings before extraordinary items, deflated by prior assets. MVEt: year t size The hypothesis in this study is test using logistic regression.

4. Results

In this study, we use two proxies for future bad news that is the likelihood of future earnings decrease and the likelihood of future dividend decreases. We examine the effects of conservatism on avoiding future earnings decrea se, because the stock market looks upon earnings changes as an additional benchmark of credibility of earnings news Dopuch et al. 2003. Thus, managers have incentives to avoid earnings decreases as well, and 2405 conservatism could serve as an additional device constraining managerial efforts in upward manipulation of earnings changes. We also examine the effects of conservatism on the likelihood of future dividend decreases because literature shows that market generally reacts very negatively to such events. Moreover, due to negative effects of dividend decreases on firms‘ stock prices, Daniel et al. β007 show that firms manage earnings upward to avoid showing dividend decreases. Kothari, Wysocki and Shu 2006 show that market reacts particularly negatively to the news of dividend decreases. They interpret this result as indicating that managers are more likely to withhold bad news releases relating to dividend cuts. Thus, it is interesting to see whether greater conservatism constrains manager‘s earnings management behavior and forces bad dividend news ―out‖. To further investigate the relationship between accounting conservatism and future earnings bad news, the formal test of regression model 1 and 2 are employed. The regression of model 12 are estimated using logistic regression.