Factors associated with the level of superfund liability disclosure in 10K reports: 1991-1997

Academy of Accounting and Financial Studies Journal, Sept, 2008 by Carol A. Cox

RESULTS

Table 4 provides the estimation results for the current study. The p-values for all model coefficients are based on White's (1980) heteroskedasticity-corrected standard errors. The adjusted R squared for the base model (Model 1) is .181 (p-value .000). The results for Model 2, which introduces industry indicators, show a larger adjusted R squared (.360, p-value .000) than the base model. The comparison of the two models illustrates the incremental value of the industry indicators. The adjusted R squared is improved when industry dummy variables are included in model (Model 2).

The reported findings are insensitive to outliers as identified by the Belsley, Kuh, and Welsch (1980) DFFITS statistics. No DFFITS statistic exceeds the size-adjusted cutoff; therefore, outlier analysis reveals no influential data affecting model coefficients. The Durbin Watson statistics for Model 1 (1.95) and Model 2 (2.01) do not indicate autocorrelation (Studenmund, 1997). Condition Numbers below 30 for Model 1 (22.34) and Model 2 (24.31) do not indicate severe multicollinearity (Studenmund, 1997). As all test statistics are reported using White (1980) standard error estimates, standard diagnostics do not reveal significant problems with the model specifications.

The primary results (Model 2) indicate a positive association between company size and the amount of ED (p-value =.000). The findings support the current hypothesis (H1), and are consistent with prior research (Barth et al., 1997; Patten, 1992, 1991; Cowen et al., 1987). The findings indicate that firms attempt to reduce political costs by increasing social disclosure. Since the magnitude of political costs is highly dependent on size, Watts and Zimmerman (1978) hypothesized that there is positive relationship between size and social disclosure.

The results indicate a negative association between profitability and the amount of ED (pvalue = .000). Prior studies that examine the impact of profitability on social disclosure find no association; however, these studies do not specifically include ED (Cowen et al., 1987; Patten, 1990). Since, the "deep pocketed" party is often legally liable for all remediation costs under the Superfund Act, the current study posits that profitability would impel management to disclose more about the environmental activities of the firm (H3). However, the negative association found between profitability and the amount of ED suggests that less profitable companies disclose more. This finding may be explained within the context of voluntary disclosure literature, which focuses on management's earnings forecasts (Pownall et al., 1993; Skinner, 1994, 1995).

Skinner (1994, 1995) provides evidence consistent with the idea that firms with negative earnings news tend to voluntarily disclose bad news. Pownall, Wasley, and Waymire (1993) also provide evidence that voluntary earnings disclosures tend to convey bad news. Since much variation exists in the ED practices of firms, ED may be explained in the context of voluntary disclosure. Environmental liability exposure may be viewed as bad news, and less profitable firms may be more likely to disclose ED. This is also consistent with more recent studies that suggest there are legal incentives for firms to disclose bad news (Skinner, 1997; Evans et al., 2002).