Value Relevance of Accounting Information for Intangible-Intensive Industries and the Impact of Scale: The US Evidence

Abstract

The structural shift in the USA from a tangible- to an intangible-intensive economy raises a concern that reporting based on generally accepted accounting principles (GAAP) might have lost its usefulness to investors. Amir and Lev [(1996) Value relevance of nonfinancial information: the wireless communications industry, Journal of Accounting and Economics, 22(1–3), pp. 3–30] argue that accounting information is not useful for intangible-intensive firms. In contrast, Collins et al. [(1997) Changes in the value relevance of earnings and book values over the past forty years, Journal of Accounting and Economics, 24(1), pp. 39–67] find that the value relevance (measured by R-squared) of accounting information has increased over time and that value relevance for intangible-intensive industries is as high as that for tangible-intensive industries. In this article, we attempt to resolve the above discrepancy by examining the impact of scale on R-squared (Brown, S., Lo, K. and 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), pp. 83–115). We find that, after controlling for scale, R-squared is lower for intangible-intensive industries than for non-intangible-intensive industries and has declined over time for intangible-intensive industries but remained stable for non-intangible-intensive industries. Interestingly, the declining trend ended with the demise of the ‘New Economy’ period (NEP) (Core, J. E., Guay, W. R. and Van Buskirk, A. (2003) Market valuations in the New Economy: an investigation of what has changed, Journal of Accounting and Economics, 34(2–3), pp. 43–67), and value relevance for both industry groups appears to be restored in the post-NEP to the pre-NEP level. We also find that R&D capitalisation increases value relevance for intangible-intensive industries, but does not completely eliminate the gap between the two groups.

Notes

¹BLL show that after controlling for variability of scale, VR of accounting information has in general declined over the period from 1958 to 1996. They further argue that the temporal increase in VR documented by CMW and FS is due to the increase in variability of scale.

²Specifically, Darrough and Ye (2007) underscore the importance of unrecorded intangibles in resolving the puzzling negative relationship between market values and earnings for loss firms. Similarly, Franzen et al. (2007) document the impact of unrecorded R&D on measures of distress risk, such as the Altman z-score. These studies suggest that INT firms increasingly shape the characteristics of the financial reporting system as a whole. Ciftci and Cready (2011) show that the size effect in earnings and returns is heavily influenced by R&D-intensive firms.

³http://www.ifrs.com/ifrs_faqs.html#q3

⁴Satisfying informational needs of investors and creditors is the primary concern for standard setters. Statement of Financial Accounting Concept (SFAC) No. 1 states that the primary purpose of accounting is to provide useful information to investors and creditors in making rational investment decisions. Consistent with its importance, there is a voluminous literature about the VR of various accounting numbers.

⁵Barth et al. (2001) further state that ‘ … in the extant literature, an accounting amount is defined as value relevant if it has a predicted association with equity market values. The primary purpose for conducting tests of VR is to extend our knowledge regarding the relevance and reliability of accounting amounts as reflected in equity values. Equity values reflect an accounting amount if the two are correlated. Relevance and reliability are the two primary criteria the FASB uses for choosing among accounting alternatives, as specified in its Conceptual Framework. Under SFAC No. 5 (FASB, 1984), an accounting amount is reliable if it represents what it purports to represent. Because the Conceptual Framework sets forth the FASB's objective criteria for evaluating accounting amounts, research needs only to operationalize the criteria, and not determine them’.

⁶Holthausen and Watts (2001) suggest that there are three types of VR studies: (1) R² as a measure of VR; (2) incremental-association; and (3) marginal-information-content studies (event studies). We follow the study design in CMW to focus on R² to reconcile the results in Amir and Lev and CMW.

⁷See Kothari and Shanken (2003) and Ye (2007) for a discussion of the problems with using R² to compare the explanatory power of a model across samples.

⁸Berkshire Hathaway reached its peak on 12 October 2007 with a stock price of $149,200 when IBM was at $109.39.

⁹To illustrate the impact of scale on R², we highlight the argument in BLL. Consider the following bivariate linear relation between z = (z₁, … ,z_n) and w = (w₁, … ,w_n):

z_i = a + bw_i + e_i, (1)

where z_i is the value of a stock at the end of a period and w_i is the variable of interest to explain the value. Equation (1) is assumed to be free of scale. On the other hand, a researcher can usually observe only possibly scale-affected data. Now assume that the data are affected by a scale factor, s = (s₁, … ,s_n). Equation (1) then becomes:s_i z_i = as_i + bs_i w_i + s_i e_i. (2) BLL show that the R² of the estimated relation between z and w is affected by both the variation in s, the scale factor, and the variation in w. Researchers who regress prices on accounting data (e.g. EPS) usually do not control for the impact of the scale effect. Thus, they estimate the following model.y_i = α + β x_i + g_i, (3)where y_i=s_i z_i, x_i=s_i w_i, and g_i=s_i e_i.

Compared with Equation (2), Equation (3) is missing as_i, leading to a correlated omitted variable problem.

¹⁰Similar to R&D, other expenditures such as advertising and human capital training, which are also investments for the future, are immediately expensed as well.

¹¹Pastor and Veronesi (2006) characterise a stock run-up and a crash as a bubble. They state that ‘on 10 March 2000, the Nasdaq Composite Index closed at its all-time high of 5048.62. In comparison, the same index stood at 1114 in August 1996 as well as in October 2002. The unusual rise and fall in the prices of technology stocks have led many academics and practitioners to describe the event as a stock price “bubble” … a more common interpretation is that the prices of technology stocks exceeded their fundamental values in the late 1990s’.

¹²Aboody and Lev (1998) suggest that capitalised SDCs are associated with contemporaneous returns, indicating that these costs are value relevant. However, Eccher (1995) finds that capitalised SDCs are not value relevant. Ciftci (2010) finds that firms that capitalise SDCs have lower earnings response coefficient than those that expense SDCs. In sum, there is no consensus about the value relevance of SDCs in prior literature.

¹³Price regressions are more appropriate for investigating our research questions than return regressions. Barth et al. (2001) argue that price regressions capture VR, while return regressions capture timeliness.

¹⁴Since the scale factor is inherently unobservable, there may be other proxies for scale. For instance, Barth and Clinch (2009) use market value of equity, book value of equity, and number of shares as proxies for scale. Since our study relies on the methodology developed by BLL in dealing with scale, we restrict our analysis to scale proxies used by BLL.

¹⁵The industries in the INT group are defined by CMW as those with the following SIC codes: 282 plastic and synthetic materials; 283 drugs; 357 computer and office equipment; 367 electronic components and accessories; 48 communications; 73 business services; and 87 engineering, accounting, R&D, and management related services. Non-INT industries are all industries except INT industries.

¹⁶BLL suggest that dividend payouts might affect scale. A profitable firm that does not pay dividends will increase its share price faster and increase its scale. In Table 1, we do not condition dividend payout on profitability. However, when we condition the dividend payout ratio on ROE, we find that dividend payout is still larger for INT industries.

¹⁷In discussing the factors that might affect variability of scale, BLL mention firm-specific factors, while CV_EPS is cross-sectional variation in EPS at the portfolio level. An alternative way to evaluate the variability of earnings is to look at the firm-specific variability of earnings. We calculate the standard deviation of EPS as per Kothari et al. (2002) over years t + 1 to t + 5 and find that the mean standard deviation of earnings for INT industries is much larger that for non-INT industries. Hence, the evidence with firm-specific earnings variability is consistent with cross-sectional variability.

¹⁸We do not investigate to what extent the variables suggested by BLL and other potential variables affect the scale differences between INT and non-INT industries. Consequently, we are unable to explain how corporate policies such as stock splits and dividend payouts impact on the coefficient of variations in our scale proxies.

¹⁹If R&D is growing over time, R&D expenditures in the current year will be on average larger than amortisation expenses if R&D were capitalised. Therefore, accounting numbers under capitalisation will not be equal to earnings under expensing.

²⁰We assume that the CV of P or BVPS captures variation in the scale factor. However, as BLL suggest, these CVs might capture variation in ‘real effects’ as well. To understand the impact of these real effects, let us assume the extreme case when the CVs capture only variation in real effects. We are not aware of any economic reason why CV of real effects would be correlated with INT_D*TIME. Hence, the addition of controls for CV of P and BVPS in Equations (8) and (9) should not result in the coefficient for INT_D*TIME becoming significant. In general, the presence of real effects in the CV of our scale proxies is likely to add noise to our scale proxy, resulting in a reduction in the power of our tests. However, we do not expect this to bias our results, unless one expects the real effects to be correlated with IND_D*TIME. Nevertheless, we acknowledge that we cannot completely rule out this possibility. We would like to thank an anonymous referee for mentioning this possibility.

²¹The conditions are: (a) the technical feasibility of completing the intangible asset; (b) its intention to complete the intangible asset and use or sell it; (c) its ability to use or sell the intangible asset t; (d) how the intangible asset will generate probable future economic benefits; (e) the availability of adequate technical, financial, and other resources to complete the development; and (f) its ability to measure reliably the expenditures attributable to the intangible asset (IAS 38 paragraph 57).