A review of Francis, J., R. LaFond, P. M. Olsson, and K. Schipper, 2004, “Costs of Equity and Earnings Attributes,” The Accounting Review 79 (4), 967-1010.
I. Research question and its importance
Drawing the relation between investor’s resource allocation decision and firm’s accounting earning, the authors examine the relation between earning attributes and cost of equity. Although earning has been studied in many empirical and survey-based researches, the authors identify a gap where earning attributes were not thoroughly covered. Therefore, the authors investigate two groups of attributes, accounting- vs. market-based attributes, and their differential associations with cost of equity. The authors (p. 969) “characterize the seven earnings attributes […] to capture differences in underlying assumptions about the function of earnings, which are, in turn, reflected in the way the attributes are measured” and hence facilitate their rationales to connect accounting earning and investor resource allocation via cost of equity. Moreover, they delve into three particular questions to augment this research, encompassing (1) the directions, (2) magnitudes, and (3) the conditional effects of this relation, whereas the last two were missing in previous research.
II. Method and findings
The authors empirically test their hypotheses from annual cross-sectional data from Value Line reports between 1975 to 2001. To mitigate the concern that earning attribute is itself correlated with firm innate features, they include a series of control variables which are known determinants of earning attributes. In addition, they also provide many robustness checks, such as alternating measurement to dependent variable, and analysing on yearly basis. Their results hold. In sum, they find support that firms with the least favourable earning attributes are, on average, bearing larger cost of equity. Amongst these effects, accounting-based attributes have larger magnitudes compared to market-based ones. Particularly, accrual quality has the highest magnitude, while predictability and conservatism do not have constant support in this relation. In brief, the finding that accrual quality with other accounting-based attributes are associated cost of equity is consistent with other preliminary researches.
III. Limitation and future research
The major limitation of this paper is causality. Although not mentioned explicitly by the authors, I can see their intention to make an argument that having more favorable earning attributes can reduce cost of equity. Yet, their research design fails to support this claim, because they do not have any time series specification although they have time series data. For a quick revision, they might adopt a lagged DV design (i.e., cost of equity at t+1) and rerun the regression of Eq. (5). Also, they better provide reverse causality test for robustness. However, a more compelling design might require an exogenous shock that changes earning attributes, such as a natural experiment design exploiting policy change, and compare the potential outcomes of treatment and control groups (e.g., Canadian firms with comparable characteristics). Finally, I would recommend difference-in-differences (DiD) as their specification since they have 27 years of longitudinal data which can illustrate treatment effect change over time, if assumptions hold.