A New Measure of Disclosure Quality

In our paper, A New Measure of Disclosure Quality: The Level of Disaggregation of Accounting Data in Annual Reports, recently featured in the Journal of Accounting Research, we develop a new measure of disclosure quality (DQ), which captures the level of disaggregation of accounting line items in firms’ annual reports, with greater disaggregation indicating higher disclosure quality. This measure is based on the premise that more detailed disclosure gives investors and lenders more information for valuation (Fairfield et al., 1996; Jegadeesh and Livnat 2006) and a higher level of disaggregation enhances the credibility of firms’ financial reports (Hirst et al. 2007; D’Souza et al. 2010).

We use the number of non-missing Balance Sheet and Income Statement items reported in Compustat to proxy for disclosure quality. A higher count indicates higher disclosure quality. In developing DQ we employ the natural nesting feature of the Balance Sheet (and to a lesser extent the Income Statement) to impose multiple screens to filter out the impact of Compustat systematic coding scheme in the count of missing items. In particular, our screening steps mitigate Type I error—counting an item as missing when in fact it is not missing.

We validate DQ through three sets of tests: if DQ captures disclosure quality, then it should 1) be related to lower analyst forecast dispersion and higher analyst forecast accuracy, 2) be negatively associated with information asymmetry as proxied by bid-ask spread, and 3) be negatively associated with cost of equity. All three sets of tests yield evidence consistent with the predictions above and with DQ capturing disclosure quality. The tests on analyst forecasting properties further reinforces that DQ captures disclosure quality not complexity, as complexity should be associated with higher analyst forecast dispersion and lower forecast accuracy, exactly opposite to the relationship documented. These results continue to hold after we control for firm fundamentals, such as operating complexity, which can drive the cross-sectional variation in DQ.

The consistent results across all three sets of validation tests also provide us further confidence that DQ is not simply reflecting Compustat’s coding of missing items, as it is extremely unlikely that the way Compustat collects and codes data would be systematically associated with established information asymmetry metrics, or the cost of equity.

We contribute to the existing literature by developing a unique disclosure measure that captures an important aspect of firms’ disclosure behaviour that has not received much research attention: the level of disaggregation of accounting data items in firms’ annual reports. DQ differs from existing measures that either capture managers’ voluntary disclosure behaviour (e.g., management earnings forecasts, conference calls) or self-constructed measures based on researchers’ or analysts’ evaluation of selected items in the financial statement (e.g., AIMR scores). Furthermore, DQ is a parsimonious measure that can be constructed for the universe of Compustat industrial firms for all years. This contrasts with existing measures, which are usually only applicable to a subset of firms (e.g., management forecasts, conference calls), or to a subset of financial statement items (e.g., AIMR), or capture the narrative aspect of MD&A (e.g., Fog index). DQ can be used by researchers for replication or to study new questions on firms’ disclosure behavior on a much wider set of firms in the economy.

We caution that the applicability of DQ is limited by the following factors. First, future research intending on establishing causality will need to include controls variables that will likely result in considerable reduction of sample sizes. Second, DQ, as a measure of annual report disaggregation level, does not capture the timeliness of new information, because annual reports provide perhaps predominantly a confirmation role to earlier or more timely voluntary disclosures. Third, it is possible that the complementarity between mandatory and voluntary disclosure can induce an upward bias in the estimation of the impacts of DQ. Future researchers interested in using DQ should take these limitations into consideration.

The full paper is available for download here.