Rajiv D. Banker, Sudipta Basu, Dmitri Byzalov, Janice Y.S. Chen
8 min
Abstract
Sales decreases affect earnings more than sales increases because of cost stickiness. We hypothesize that this correlated omitted variable constitutes a confounding effect in standard asymmetric timeliness models. Controlling for a piecewise linear effect of sales changes in these models decreases the measured asymmetric timeliness significantly and changes inferences about the average level of conservatism and the extent of crosssectional variation in conservatism. Validation tests confirm that the asymmetry for sales changes is consistent with sticky costs and is distinct from conditional conservatism. Future empirical research on conditional conservatism should recognize the potential confounding effect of sticky costs.
Sam: Then the distinction is between an asymmetry in what happens economically and an asymmetry in how accountants record it.
Alex: That’s the paper’s central distinction. Conservatism concerns the information system translating economic activity into accounting numbers. Sticky costs concern the economic activity itself. Standard models risk attributing both to reporting.
Sam: What do the authors add that lets them distinguish those explanations?
Alex: They keep the usual separate earnings responses to positive and negative stock returns. Then they also allow separate responses to rising and falling sales. The model can assign some earnings asymmetry to sales behavior rather than automatically assigning it to conservatism.
Sam: But falling sales could also trigger genuine conservative reporting, such as an asset write-down. How do they know what their added terms capture?
Alex: They explicitly test that objection. Consider depreciation, the expense recording the use of fixed assets over time. Retaining unused assets after sales fall can keep depreciation from falling, while expansion requires assets that add depreciation.
Sam: Whereas a write-down triggered by bad news would show up somewhere else?
Alex: The authors say those write-downs flow through special items, not depreciation. They find the predicted sales asymmetry in depreciation and interest expense. Those results fit sticky costs rather than their competing conservatism explanation.
Sam: Do they also test a component where conservative reporting should matter, but sticky costs shouldn’t?
Alex: They examine changes in accounts receivable, money customers owe the firm. Falling sales could signal write-downs there, but sticky costs concern costs rather than revenue. The sales asymmetry is not statistically significant for that component.
Sam: That’s more informative than simply showing that the expanded model fits better. The explanations predict different patterns across accounting components.
Alex: They also examine prior sales declines and firms’ resource intensity. The patterns again support sticky costs. But the authors stop short of calling their sales measure pure: conservatism may contribute too.
Sam: With that qualification, how large is the change in the main estimate?
Alex: They analyze about fifty-five thousand firm-year observations from publicly traded US companies. The records span nineteen eighty-seven through two thousand seven. Financial firms and observations with very large sales changes are excluded from the main sample.
Sam: So these are repeated company observations, not fifty-five thousand distinct companies. What happens when they add the sales controls?
Alex: The standard Basu estimate is about twenty-seven percent higher than the estimate with sticky-cost controls. That difference is statistically significant. Both conservatism and sticky costs remain significant in the expanded model.
Sam: That percentage is relative to the corrected estimate. It doesn’t mean conservatism disappears, or that the coefficient falls by that same percentage.
Alex: Correct. The authors report an upward bias relative to their full model. Their conclusion is not that asymmetric timeliness is useless. It’s that the uncorrected measure mixes mechanisms.
Sam: Does this survive other researchers’ proposed improvements, or only the original model?
Alex: It survives adding firm fixed effects, which address differences in expected earnings and returns across companies. The remaining upward bias is about twenty-five percent. So that earlier correction and this one address different problems.
Sam: Another approach is to remove operating cash flow and focus on accounting adjustments. Would that solve the problem?
Alex: Not here. Those adjustments are operating accruals: earnings components not represented by current operating cash flow. Sticky costs can affect them too, including through depreciation. The accruals-based estimate shows an upward bias of sixty-two percent relative to the expanded model.
Sam: Then changing the earnings component doesn’t automatically isolate reporting choices. What changes when researchers compare firms rather than estimate an average?
Alex: Industry comparisons change substantially. In the more detailed industry groupings, adding the controls significantly changes the conservatism estimate in about half the industries. The relative ordering also changes, as do firm-level scores built from company characteristics.
Sam: Does that alter a substantive claim about what explains conservative reporting?
Alex: The authors examine managerial stock ownership. Without sticky-cost controls, greater ownership is significantly associated with less conservatism. After the correction, that association weakens and is no longer statistically significant.
Sam: That doesn’t establish that ownership has no effect. It shows that the original association isn’t robust to separating these mechanisms.
Alex: And the authors explicitly warn against causal interpretation. Their models do not resolve the endogeneity of managerial ownership. More broadly, the sales proxy mainly captures sticky costs, not exclusively, and this historical US sample doesn’t establish universality.
Sam: So who should read the full document, and where should they begin?
Alex: Researchers using asymmetric timeliness to measure reporting choices should read it closely. Start with the empirical models, then the validation tests asking whether sales changes capture sticky costs or conservatism. If you compare firms or industries, read the variation analysis next.
Sam: And for someone interested in debt contracts rather than reporting choices?
Alex: Read the conclusion’s discussion of contracting. Total earnings asymmetry may matter there, but the authors argue its components affect creditors differently. For everyone else, remember: earnings can react unevenly because businesses adjust unevenly, not just because accountants report unevenly.
Sam: Keep the operating mechanism separate from the reporting story.