An Empirical Assessment of Voluntary Disclosure Theory

By Mark LangRussell Lundholm
1992| Working Paper No. 1188

In this paper we examine the determinants of a firm’s level of voluntary disclosure. As a comprehensive measure of a firm’s disclosure level, we use analysts’ evaluations of firms’ disclosures as reported in the Financial Analysts Federation Corporate Information Committee Report. This report is prepared by industry subcommittees which evaluate the informativeness of firms’ disclosure on three dimensions: annual published information, non-required published information and investor relations. This measure of disclosure has many advantages: it is based on the perceptions of actual users of disclosed informtion, it measures the effort firms put into communicating information, it is a comprehensive analysis of a firm’s chosen level of disclosures and it is naturally controlled by industry.To explain a firm’s chosen level of disclosure we use variables suggested by the theory of voluntary disclosure as developed in Verrecchia (1983, 1990) and Dye (1985, 1986). In particular, the firm’s performance must exceed some threhold value before it warrants incurring the cost to disclose, and the more sensitive a firm is to the perceptions of outsiders the more it will disclose. Thus, a firm’s level of disclosure is predicted to be increasing in its own performance, decreasing in the threhold level and increasing in the sensitivity to outside perceptions.We find evidence largely consistent with the theory. A firms’ level of disclosure is increasing in its past, present and future return on equity and increasing in its past and present stock returns (proxies for performance); a firm’s level of disclosure is increasing in the firm’s size and decreasing in the standard deviation of past return on equity (proxies for the threshold); and a firm’s level of disclosure is higher if the firms issues securities in a current or future period (a proxy for the sensitivity t outsiders’ perceptions). These results hold both unconditionally and on the margin (i.e after conditioning on the other variables in the model). They are also robust to changes in the specification of performance and threshold levels an after controlling for possible time-series dependence.