GB791 Unit 7 Empirical Study Appraisal Example

Forty Years of R-Squared: A Unit 7 Appraisal of a Landmark Value Relevance Study

[Student Name]

Purdue University Global

GB791: Advanced Topics in Accounting

Unit 7 Empirical Study Appraisal

[Instructor Name]

[Date]

This appraisal concerns a real published study; its reading of the design and its limits is the candidate's.

The Study

Collins, D. W., Maydew, E. L., & Weiss, I. S. (1997). Changes in the value-relevance of earnings and book values over the past forty years. Journal of Accounting and Economics, 24(1), 39-67.

The Question and Why It Mattered

By the mid-1990s, a common complaint among practitioners and some academics was that traditional financial statements were losing their relevance in an economy increasingly driven by intangible assets, technology and services. Collins et al. (1997) set out to test that complaint. If accounting numbers had become less useful to investors, the relation between those numbers and stock prices should have weakened over time.

The Design

The study regressed each firm's share price on its earnings per share and its book value per share, year by year, using Compustat data for firms over the forty years from 1953 to 1993. The model follows from valuation theory, in which a firm's value can be expressed in terms of its book value and its expected future earnings, and current earnings serve as a proxy for those expectations. The measure of value relevance was the adjusted R-squared of each year's regression: the share of variation in prices across firms that earnings and book value together explained. The study also decomposed that explanatory power into the part attributable to earnings alone, the part attributable to book value alone and the part they shared.

The Findings

Contrary to the complaint, the combined value relevance of earnings and book value did not decline over the forty years and, if anything, rose slightly. What changed was the mix. The incremental explanatory power of earnings fell, while that of book value rose. The authors linked the shift to several changes in the population of firms: a growing frequency of one-time items in earnings, a growing share of firms reporting losses, a shift toward smaller firms in the sample and an increase in firms with large intangible assets. In each of these cases, current earnings are a weaker guide to future earnings, and book value takes on more of the explanatory work.

Appraising the Measure: What R-Squared Captures

The study's conclusions depend on a single measure, and the first question an appraisal must ask is what that measure captures.

R-squared from a cross-sectional price regression measures how well accounting numbers explain differences in prices across firms at a point in time. It does not measure whether accounting numbers provide new information to investors, since prices could be explained by accounting numbers that investors learned about from other sources earlier. It does not measure whether accounting numbers are useful for any particular decision, such as lending or contracting. And it cannot say whether the numbers are relevant in the sense standard setters use, which includes predictive value for future cash flows. Holthausen and Watts (2001) made a broader version of this argument, contending that value relevance studies rest on an implicit theory of what accounting is for that excludes most of its uses, and that their findings therefore offer limited guidance to standard setters.

Appraising the Measure: The Problem of Scale

A second problem is statistical. A regression of price on per-share earnings and book value combines firms of very different sizes, and the variation in scale across firms can drive R-squared independently of how well accounting explains value. Brown et al. (1999) showed that R-squared in such regressions rises with the cross-sectional variation in scale, and that this variation changed over the forty years. When they controlled for scale, by deflating variables or including a scale proxy, the value relevance of earnings and book value declined over the period rather than holding steady. That finding does not show that the 1997 results were wrong, but it shows that their main conclusion, no decline in combined value relevance, depends on a modeling choice that the original study did not fully address.

A rising R-squared can mean accounting explains more of value, or only that the firms in the sample differ more in size, and the 1997 design could not tell which.

Appraising the Sample and Period

The sample consists of firms with data on Compustat, which over the forty years expanded from mainly large industrial firms to a much broader population including small firms, technology firms and firms with losses. Changes in value relevance over time are therefore changes in a population that was itself changing. The study treated this partly as a finding, since the shift toward small and loss-making firms helps explain the declining role of earnings, but it also means the forty-year comparison is not of the same kinds of firms at different times. Francis and Schipper (1999) examined the same question with several alternative measures and found evidence of decline in some, which suggests the conclusion is sensitive to how value relevance is defined.

Appraising the Model Itself

A further question concerns the price regression itself. Regressing price on earnings and book value assumes that the two numbers capture what accounting contributes to value, and that their coefficients are stable across firms within a year. Neither assumption is innocent. Loss-making firms, whose numbers in the study's own account grew more common, have a different relation between earnings and price than profitable firms, since investors value a loss firm partly on its option to abandon the business rather than on its current earnings. Pooling profitable and loss firms in one regression therefore mixes two relations, and some of the decline in the role of earnings may reflect that mixing rather than a change in how investors use earnings for any given firm.

How Far the Conclusion Travels

What the study supports, appraised in this way, is a narrower conclusion than its headline. Within the price-level framework and its scale assumptions, the combined explanatory power of earnings and book value for cross-sectional differences in price did not decline between 1953 and 1993, while the relative roles of earnings and book value shifted for identifiable reasons. That shift, particularly the growing importance of book value when earnings contain transitory items or losses, has proved durable in later research and is the study's most useful contribution.

The conclusion does not establish that financial statements remained equally useful to investors over the period, because the measure does not capture usefulness in that sense. Nor does it extend beyond 1993, a period that ended before the growth of intangible-intensive firms accelerated. A researcher citing the study today should cite it for the decomposition and its explanations, not as evidence that accounting kept its relevance.

References

Brown, S., Lo, K., & Lys, T. (1999). Use of R2 in accounting research: Measuring changes in value relevance over the last four decades. Journal of Accounting and Economics, 28(2), 83-115. https://doi.org/10.1016/S0165-4101(99)00023-3

Collins, D. W., Maydew, E. L., & Weiss, I. S. (1997). Changes in the value-relevance of earnings and book values over the past forty years. Journal of Accounting and Economics, 24(1), 39-67. https://doi.org/10.1016/S0165-4101(97)00015-3

Francis, J., & Schipper, K. (1999). Have financial statements lost their relevance? Journal of Accounting Research, 37(2), 319-352. https://doi.org/10.2307/2491412

Holthausen, R. W., & Watts, R. L. (2001). The relevance of the value-relevance literature for financial accounting standard setting. Journal of Accounting and Economics, 31(1-3), 3-75. https://doi.org/10.1016/S0165-4101(01)00029-5