Disclosure and the outcome of securities litigation

Any litigation, regardless of the outcome, imposes incremental costs on sued firms including legal fees, short-term stock price effects, and reputation costs, which are difficult to measure. Although a component of these costs is probably fixed, they also likely include a large variable component that depends on the outcome of litigation. Most firms purchase directors and officers (D&O) liability insurance, which can shield them from settlement costs up to their coverage limits. However, Baker and Griffith (2007) point out that the pricing of D&O insurance incorporates the claims history of the corporation, suggesting that settlements are priced into future D&O premiums. Managers must also consider that a negative litigation outcome can impose nonmonetary costs such as damage to reputation or management turnover.

Although Rogers and Van Buskirk’s results are consistent with managers behaving as if they believe more disclosure increases litigation costs, they do not provide direct evidence of this association. Managers’ perceptions may be influenced in a way that leads them to overreact to or misinterpret the effect of disclosure in litigation. More disclosure may also increase the cost to defend a lawsuit but not necessarily the likelihood of a negative outcome.

The purpose of our study is not to evaluate the net impact of the PSLRA on social welfare. It is possible that the PSLRA has had an overall positive effect on social welfare by decreasing frivolous lawsuits (Grundfest and Perino 1997) and/or potentially encouraging forward-looking disclosures through its safe harbor provisions (Johnson et al. 2001).

For example, in In re Bridgestone Sec. Litig., (2006), where the court was influenced by multiple disclosures referencing a faulty product throughout the class period, and City of Sterling Heights Pol. And Fire Ret. Sys. v. Abbey Nat’l, PLC, (2006), where the decision turned on repeated disclosures about the defendant’s investments.

Companies may have duties to disclose when there is a recognized relationship of trust and confidence, when the company itself is trading, or when the company knows insiders are trading on nonpublic information. See Palmiter (2011, p. 386).

The complaint and decision retrieved from http://​securities.​stanford.​edu/​filings-case.​html?​id=​104341.

However, press releases frequently announce when a firm will present at investor conferences, and these are counted in our measures.

There are more securities than lawsuits in the data because of lawsuits filed against mutual fund families and other securitized assets as well as firms with multiple classes of stockholders. McShane et al. (2012) discuss this and other issues encountered when using the SCAS data (McShane et al. 2012) and reference RiskMetrics as the source of their data. MSCI Inc., the parent of ISS, purchased RiskMetrics in 2010).

We also construct measures of tone based on word lists derived from Diction and LIWC, which are general word lists that have been used previously. We find similar results in all specifications using these alternative tone measures. We choose to report results based on the word lists from Henry (2008), because they are domain specific, and word lists derived from earnings press releases seem most appropriate in our setting.

In untabulated analyses, we confirm that are results are not sensitive to the inclusion of any particular fixed effect.

We cluster by circuit because each of the 12 circuits has its own distinctive case law precedent, and therefore the error terms might be correlated given the general model across circuits. Our inferences do not change if we do not cluster by circuit.

In untabulated analyses, we confirm that, as the market value of defendant firms increases, the value of settlements paid by those firms increases substantially, even as the likelihood of paying a settlement decreases.

We use LexisNexis to obtain press releases issued via PR Newswire and Business Wire for the pre-PSLRA period. We use LexisNexis because press releases are more readily available during the pre-PSLRA period via LexisNexis, relative to Factiva. We also collected press releases for the pre-PSLRA period using Factiva and find similar results.

The results are similar when using an alternate nuisance classification in which all settlements less than $3 million dollars are considered nuisance settlements.

The sample size is much smaller in the pre-period (128 observations) versus the post-period (909 observations), which may indicate a lack of power. To examine whether this is the case, we perform a bootstrap procedure in the post-period in which we restrict the sample in the post-period to be the same number of observations as the pre-period and re-estimate the regression in Table 4. We repeat this 1000 times and get an average coefficient estimate for the post-period using a similar sample size. The bootstrapped results provide similar coefficient point estimates in the post-period (0.00467 for Disc_words and 0.00503 for Disc_PR in each specification of Table 4). However, the standard errors are noticeably higher in these bootstrap tests, resulting in coefficients that are statistically insignificant from zero. This suggests that the small sample size in the pre = period may be decreasing the lack of power in of our tests. However, we note that, if the point estimates are accurate in the pre-period (albeit imprecise) in Table 4, then the fact that the coefficients on our disclosure measures substantially increase or switch signs suggests the relation is more tenuous in the pre-PSLRA period.

We use both periods because the class period is arguably more relevant and closer in time to the post-period, reducing the chance of some confounding factor causing a change in disclosure. However, to be consistent with Rogers and Van Buskirk (2009) and in case of the possibility that firms’ disclosures were somehow abnormal during the period of alleged misbehavior, we also use the pre-period measured before the start of the class period.