In this section I discuss the empirical proxies of the factors that are predicted to influence firm’s disclosure choices for bad news.
Effective on October 23, 2000, the Securities and Exchange Commission in the US passed Regulation Fair Disclosure that prohibits selective disclosure of mate- rial information to experts and other investment professionals. Under the regulation, any intentional disclosure of material non-public information by firms to experts or other parties must be simultaneously released to the general public. RegFD reduces the amount of asymmetric information in the securities markets by forcing firms to either disclose information to everyone or disclose no information (Eleswarapu, Thompson, and Venkataraman, 2004). RegFD delays price discovery to quarterly earnings release by stifling information leakage (Dong, Li, Ramesh, and Shen, 2011). I argue that passage of RegFD reduces managers’ incentives to leak bad news well before the mandatory quarterly earnings announcements.
The litigation reduction hypothesis proposed by Skinner (1994) states that firms reveal bad news to lower the likelihood of litigation. Donlson, McInnis, Mer- genthaler, and Yu (2012) study bad earnings news and conclude that litigation risk measured by predicted litigation probability is negatively associated with the amount of bad news leaked. I argue that firms with low litigation risk benefit from revealing more bad mandatory earnings news and are thereby more likely to continue the practice. Despite firms with high litigation risk might have incentive to withhold
bad news in relation to discretionary announcements, they may release more bad earnings news in the run-up to disappointing mandatory earnings announcements. The measure of litigation risk is calculated using the coefficient estimates obtained from Rogers and Stocken (2005). The explanatory variables used in their model are primarily stock return-based variables such as market value, stock turnover, beta, and volatility (see Rogers and Stocken, 2005 appendix B for more details). Same as Kothari, Shu, and Wysocki (2009), I only include the independent variables that are significant into the prediction. All the variables are computed at the end of the last fiscal quarter.
Information asymmetry between management and investors affects news dis- closure. In the traditional setting, high information asymmetry tends to provide incentives for managers to disclose all types of news to avoid market penalties from investors. Thus, higher information asymmetry corresponds to larger amount of bad news released. Following Kothari, Shu, and Wysocki (2009), I construct a factor that potentially measure information asymmetry based on five variables, namely market-to-book ratio, stock volatility, leverage ratio, membership in high-tech in- dustries, and regulatory status. The market-to-book ratio is computed by dividing the market value of equity by the book value of equity. Stock volatility is calcu- lated as the standard deviation of daily stock returns within a quarter. Leverage is measured as long-term debt scaled by total assets. High-tech firms are firms with the Standard Industrial Classification (SIC) codes 2833-2836, 3570-3577, 3600-3674, 7371-7379, and 8731-8734. Regulated industries excluding the financial institutions are considered with SIC codes 4812-4813, 4833, 4841, 4891-4899, 4922-4924, 4931, and 4941. The factor analysis is conducted to extract an underlying information asymmetry factor. I define firms with above-median asymmetry factor score as high information asymmetry firms and vice versa for low information asymmetry firms. All the ratios are measured at the end of the previous fiscal quarter.
asymmetric loss function in choosing their voluntary disclosure policies. That is, managers behave as if they bear large costs when investors are surprised by large negative earnings news, but not when other types of news are announced (Skinner, 1994). Both litigation risk and reputation costs possibly create this asymmetric loss function. The costs borne by managers as a result of large negative earnings surprises further increase when firms are in distress (Gilson, 1989). I argue that the career concern incentivizes managers to release more bad news. To capture these incentives, I classify a firm quarter as being financially distressed if the Zmijewski (1984) Z-score financial distress rank is in the top decile of all firms in that quarter. Our predications are the opposite of those of Kothari, Shu, and Wysocki (2009). The reasons are mainly twofold. Firstly, this chapter investigates the stock market reaction to mandatory quarterly earnings announcement while Kothari, Shu, and Wysocki (2009) look at stock price behaviour surrounding discretionary corpo- ration information including dividend change and voluntary management earnings forecasts. Secondly, the association study framework is adopted to examine the stock return-news relation. With this framework, the return calculation windows are the fiscal quarters for which the mandatory earnings announcements cover, and the public announcements are typically two to three months after the fiscal quar- ter. This framework works based on the assumption that earnings news is leaked well before the actual announcements. Due to these two characteristics, the overall prediction is that firms may release more bad news in the run-up to disappointment mandatory earnings announcements even if they have a general incentive to with- hold bad news in relation to discretionary announcements. The release increases with lower litigation risk and high information asymmetry, and decreases after the passage of RegFD.
4.2.3 Summary statistics
Table 4.1 Panel A reports summary statistics for some of the important variables. There are 9345 firms in the final sample. The average log return across firms is around 0.3% per quarter, with a standard deviation 25.6%. The earnings surprise measure, dEP, has a mean 0.2% and a standard deviation 0.045. the measure for difference of opinion,DoO, has a average of 0.450 and a standard deviation of 0.467. Ambiguity which is equal to 1−DoOhas a average of 0.550 and the same precision. From IBES, the number of experts (N U M EST) following each firm ranges from 2 to 50 with an average around 7. This range is used as a measure of difference of opinion in the model. The rational is that more experts coverage leads to lower standard deviation of forecasts. The correlations between those variables are presented in the Panel B of Table 4.1.