You are on page 1of 34

Accounting Research Center, Booth School of Business, University of Chicago

The Boundaries of Financial Reporting and How to Extend Them Author(s): Baruch Lev and Paul Zarowin Source: Journal of Accounting Research, Vol. 37, No. 2 (Autumn, 1999), pp. 353-385 Published by: Blackwell Publishing on behalf of Accounting Research Center, Booth School of Business, University of Chicago Stable URL: http://www.jstor.org/stable/2491413 Accessed: 27/11/2009 02:23
Your use of the JSTOR archive indicates your acceptance of JSTOR's Terms and Conditions of Use, available at http://www.jstor.org/page/info/about/policies/terms.jsp. JSTOR's Terms and Conditions of Use provides, in part, that unless you have obtained prior permission, you may not download an entire issue of a journal or multiple copies of articles, and you may use content in the JSTOR archive only for your personal, non-commercial use. Please contact the publisher regarding any further use of this work. Publisher contact information may be obtained at http://www.jstor.org/action/showPublisher?publisherCode=black. Each copy of any part of a JSTOR transmission must contain the same copyright notice that appears on the screen or printed page of such transmission. JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of content in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new forms of scholarship. For more information about JSTOR, please contact support@jstor.org.

Accounting Research Center, Booth School of Business, University of Chicago and Blackwell Publishing are collaborating with JSTOR to digitize, preserve and extend access to Journal of Accounting Research.

http://www.jstor.org

Journal of AccountingResearch Vol. 37 No. 2 Autumn 1999


Printed in US.A.

The Boundaries of Financial Reporting and How to Extend Them


BARUCH LEV AND PAUL ZAROWIN*

In this study we investigate the usefulness of financial information to investors in comparison to the total information in the marketplace.1 Our evidence indicates that the usefulness of reported earnings, cash flows, and book (equity) values has been deteriorating over the past 20 years. We document that this deterioration in usefulness, in the face of both increasing investor demand for relevant information and persistent regulator efforts to improve the quality and timeliness of financial information, is due to change. Whether driven by innovation, competition, or deregulation, the impact of change on firms' operations and economic conditions is not adequately reflected by the current reporting system. The large investments that generally drive change, such as restructuring costs and R&D expenditures, are immediately expensed, while the benefits of change are recorded later and are not matched with
* New York University. Helpful comments and suggestions were obtained from David Aboody, Mary Barth, William Beaver, Christine Botosan, Amihud Dotan, Ron Kasznik, Nahum Melumad, Jim Ohlson, Fernando Penalva, Richard Sansing, Brett Trueman, Paul Wachtel and Gregory Waymire. 1 We assume that the major objective of financial reporting is the provision of decisionrelevant information to investors, as stated in the FASB's Statementof Financial Accounting ConceptsNo. 1: "Financial reporting should provide information that is useful to present and potential investors and creditors and other users in making rational investment, credit and similar decisions.. . . Financial reporting should provide information to help present and potential investors and creditors and other users in assessing the amounts, timing, and uncertainty of prospective cash receipts from dividends or interest and the proceeds from the sale, redemption, or maturity of securities or loans" (FASB[1978, paras. 34, 37]).

353 Copyright?, Instituteof ProfessionalAccounting, 1999

354

JOURNAL

OF ACCOUNTING

RESEARCH,

AUTUMN

1999

the previously expensed investments. Consequently, the fundamental accounting measurement process of periodically matching costs with revenues is seriously distorted, adversely affecting the informativeness of financial information.2 We validate our conjecture, that business change is an important factor responsible for the deterioration in the informativeness of financial information, first by providing evidence that the rate of change experienced by U.S. business enterprises has increased over the past 20 years, and then by linking the increased rate of change with the decline in the usefulness of financial information. We extend our inquiry by considering the accounting for innovative activitiesof business enterprises-the major initiator of change in developed economies. These activities, mostly in the form of investment in intangibleassetssuch as R&D, information technology, brands, and human resources, constantly alter firms' products, operations, economic conditions, and market values. We argue that it is in the accounting for intangibles that the present system fails most seriously to reflect enterprise value and performance, mainly due to the mismatching of costs with revenues. We demonstrate the adverse informational consequences of the accounting treatment of intangibles by documenting a positive association between the rate of business change and shifts in R&D spending, and an association between the decrease in the informativeness of earnings and changes in R&D spending. Having linked the increasing importance of intangible investments, through their effect on the rate of business change, to the documented decline in the usefulness of financial information, we address the normative question of what can be done to arrest this decline. We advance two proposals: a comprehensive capitalization of intangible investments and a systematic restatement of financial reports. The first proposal expands on a practice currently used in special circumstances (e.g., software development costs), while the second proposal implies a radical change in current accounting practices.

1. The Decreasing Usefulness of Financial Information


We use statistical associations between accounting data and capital market values (stock prices and returns) to assess the usefulness of financial information to investors.3 Such associations reflect the consequences of investors' actions,whereas alternative usefulness measures, such as those
2 It is not change per se that distorts financial reporting, rather it is the increased uncertainty generally associated with change (e.g., uncertainty about the consequences of a substantial restructuring, product development, or deregulation). If the consequences of change were perfectly predictable, the accounting system would have no problem matching costs with revenues. The uncertainty associated with change provides the justification or excuse for the immediate expensing of practically all change-related outlays. 3 Since we are concerned with the usefulness of financial information to investors, the contractual and stewardship (compensation) functions of such information are not examined here.

BOUNDARIES

OF FINANCIAL

REPORTING

355

based on questionnaire or interview studies, reflect investors' opinions and beliefs. Furthermore, empirical associations between market values and financial data allow for an assessment of the incrementalusefulnessof accounting data relative to other information sources (e.g., managers' voluntary disclosures or analysts' recommendations). Interviews or prediction studies, where usefulness is assessed in terms of predictive power (e.g., Ou and Penman [1989]), generally do not compare the usefulness of accounting data with that of other information sources.4
1.1
THE WEAKENING RETURNS-EARNINGS ASSOCIATION

It has been previously documented (e.g., Lev [1989]) that the association between reported earnings and stock returns is weak. Over return intervals of up to a year, earnings account for only 5% to 10% of the variation in stock returns.5 This result holds in cross-section and timeseries studies and applies to reported earnings as well as to earnings surprises. In the current study we examine changesover time in the informativeness of earnings, as well as cash flows and book values. Because our interest is in linking the rate of business change with shifts over time in informativeness, we focus on the past 20 years, since that is the period of greatest change affecting business enterprises (e.g., globalization of business operations, the advent of many high-tech industries, and extensive worldwide deregulations). Our first analysis examines the usefulness of reported earnings, using the following cross-sectional regression to estimate the association between annual stock returns and the level and change of earnings:6
Rit = 0
+ alEit + a2AEit + cit,

1977-96

(1)

where: Rit = firm i's stock return for fiscal year t. Eit = reported earnings before extraordinary items (Compustatitem #58) of firm i in fiscal year t.
AEjt = annual change in earnings: AEjt = Eit- Eit-1, proxying for the

surprise element in reported earnings.


4Association studies, such as those presented here, indicate an upperboundof usefulness of the financial data examined. Unless the stock return interval around the announcement is very narrow (e.g., a day), an association between an information item and stock return does not necessarily imply that the information item indeed triggered investors' reaction. It may be that other, more timely information caused the stock price change. 5 Nonearnings accounting data (e.g., inventories, R&D, capital expenditures) increase the explanatory power of financial information with respect to stock returns to 15-25% (Lev and Thiagarajan [1993] and Livnat and Zarowin [1990]). 6 A complete characterization of the returns-earnings relationship includes, in addition to current and lagged earnings, the impact of earnings on forecasts of future earnings (Lev [1989, sec. 5]). Several recent studies (e.g., Liu and Thomas [1998]) include analysts' forecasts in the returns-earnings association. But analysts' forecasts are affected by multiple information sources in addition to reported earnings. In fact, such forecasts reflect the entire information set available to analysts (e.g., managerial voluntary disclosures), thereby overstating the informativeness of reported earnings.

356

BARUCH

LEV AND

PAUL ZAROWIN

Both Eit and AEit are scaled by firm i's total market value of equity at the beginning of year t. Our sources of data are the 1996 versions of the Compustat(both Currentand ResearchFiles) and CRSPdatabases. Table 1 presents estimates of regression (1) for each of the years, 1978-96 (1977 is "lost" due to the first differencing of earnings). The "total sample," containing 3,700 to 6,800 firms per year, includes all Compustat firms with available data. The "constant sample" is 1,300 firms with data in each of the 20 years examined. Panel A of table 1 shows that the association between stock returns and earnings, as measured by R2, has been declining throughout the 1977-96 period: from R2s of 6-12% in the first ten years of the sample to R2s of 4-8% in the last ten years.7 A regression of the annual R2s in panel A on a Time variable indicates (panel B) that the decrease is statistically significant (the estimated Time
coefficient is -0.002, t = -2.97).8

A different perspective on the informativeness of earnings is provided by the combined ERC (earnings response coefficient), defined as the sum of the slope coefficients of the level and change of earnings (a1 + a2 in regression (1)). This measure reflects the average change in the stock price associated with a dollar change in earnings. A low slope coefficient suggests that reported earnings are not particularly informative to investors, perhaps because they are perceived as transitory or subject to managerial manipulation. In contrast, a high slope coefficient indicates that a large stock price change is associated with reported earnings, reflecting investors' belief that earnings are largely permanent. It has been shown (e.g., Lev [1989]) that the estimated slope coefficient is a function of the
precision of earnings.

The estimated slope coefficients (ERCs) in table 1 (fourth column from left) have been decreasing over 1977-96, from a range of 0.750.90 in the first five years of the sample, to 0.60-0.80 in the last five years. A regression of the yearly ERCs on Time (panel B) confirms that the ERC's decline is statistically significant (the estimated coefficient of Timefor the total sample is -0.011, t = -3.04).9 The evidence on the declining slope coefficients of earnings complements the inferences based on declining R2s. While the declining R2s in table 1 might be driven by an increase in the relative importance of nonaccounting information, with no change in the informativeness of earnings on a stand-alone basis,
7 All R2s reported in this study are adjusted R2s. 8All regressions on Time were also run with one- and two-lag autocorrelation correc-

tions, with virtually identical results. 9 Ramesh and Thiagarajan [1995] provide similar evidence of a temporal decrease of the returns-earnings slope coefficient (ERC). They subject the data to various statistical and specification tests and conclude that the intertemporal decrease in ERCs is both statistically significant and robust to different model specifications (e.g., accounting for firm-size effect). Ramesh and Thiagarajan also examine the pattern of firm-specific (time-series) ERCs and find a similar phenomenon of temporally declining response coefficients. The temporal decline in ERCis documented also when unexpected earnings relative to analysts' forecasts are considered (Cheng- Honwood. and McKeown [19921).

BOUNDARIES

OF FINANCIAL

REPORTING

357

TABLE 1 The Association between Earnings and StockReturns Estimatesfrom YearlyCross-Sectional Regressionsof Annual StockReturns on the Level and Change of Reported Earnings Panel A: Equation (1): Rit = ao + a, Eit + a2AEit + Fit
Total Samplel Year 1978 1979 1980 1981 1982 1983 1984 1985 1986 1987 1988 1989 1990 1991 1992 1993 1994 1995 1996 Number of Observations 3,689 3,851 4,141 4,347 4,822 4,751 5,074 5,057 5,048 5,318 5,350 5,206 5,162 5,007 5,245 5,501 6,532 6,791 6,593 Constant Sample'

R2
0.115 0.072 0.059 0.119 0.066 0.053 0.111 0.109 0.076 0.069 0.074 0.082 0.070 0.061 0.061 0.050 0.064 0.056 0.037

ERC
0.907 0.865 0.768 0.909 0.755 0.711 0.753 0.701 0.633 0.646 0.575 0.657 0.537 0.663 0.635 0.719 0.671 0.826 0.610

R2
0.167 0.114 0.092 0.173 0.099 0.070 0.245 0.159 0.169 0.107 0.079 0.117 0.135 0.104 0.062 0.064 0.098 0.124 0.031

ERC
1.689 1.368 1.367 1.648 1.190 0.939 1.177 0.936 1.067 0.988 0.609 0.872 0.788 0.851 0.534 0.717 0.826 1.081 0.418

Panel B: Time Regressions


a + b (Timet) + ct; t = 1978-96 R2= ERCt = a + b (Timet) + ct; t = 1978-96 (t-values in parentheses)

a
Total Sample R2 0.285 (4.00) 1.688 (5.25) Constant R2 Sample 0.470 (2.80)

b
-0.002 (-2.97) -0.011 (-3.04)

R2
0.30

ERC

0.31

-0.004 (-2.11) -0.050 (-5.76)

0.16

ERC

5.353 (7.08)

0.64

Variable definitions for panel A: Rit = annual stock return of firm i in fiscal t, Eit and AEit level and change of annual earnings of firm i in fiscal t, and ERC= combined slope coefficients or "earnings response coefficient," the sum, of the estimated regression coefficients of Eit and AEit. Both Eit and AEit are scaled by market value of equity at the beginning of t. = Variable definitions for panel B: R2 and ERCt estimated coefficients of determination (adjusted R2) and earnings response coefficients (ERC), presented in panel A, and Timet= a time variable, 1978-96. Files. 'The total sample includes all firms with the required data on Compustat'sCurrentand Research The constant sample includes about 1,300 companies with the required data for the 20-year sample period, 1977-96.

358

BARUCH

LEV AND

PAUL ZAROWIN

the declining slope coefficients indicate a deterioration in the value relevance of earnings to investors, irrespective of the effects of other information sources. To assess whether the documented weakening of the returns-earnings association is due to the addition of new firms to the Compustat database (and hence to our sample), we replicated the analysis with a "constant sample" of 1,300 firms which operated throughout the sample period. This sample is clearly subject to a survivorship bias, while the total samFile (that is deleted, ple which includes firms from the Compustat Research bankrupt, or merged companies) is not subject to such a bias. The estimates reported in the right two columns of table 1 indicate that the declining returns-earnings association is not the result of new firms joining the sample; both the R2s and slope coefficients of the constant sample have been decreasing over time. The regressions on Time, reported in panel B, indicate that the decreases in R2 and ERCs of the constant sample are even more pronounced than those of the total sample. We also note that the R2s of the constant sample in table 1 are in every year substantially larger than those of the total sample, indicating that earnings are more informative for firms with extended operating histories (for a similar result, see Lang [1991]). We return to this point in section 4. Also, both the R2s and ERCs in table 1 exhibit substantial volatility over time, a phenomenon noted in earlier research (e.g., Lev [1989]), which implies the limited predictive usefulness of earnings. To summarize, our findings indicate that the cross-sectional association between stock returns and reported earnings, our measure of the usefulness of earnings to investors, has declined over the past 20 years. Our measure is not sensitive to changes over time in the quality of analysts' earnings forecasts because we do not measure the reactionto an earnings announcement, which is determined in part by the extent of earnings surprise. Rather, our analysis reflects the consistencybetween the information conveyed by earnings and that which affected investors' decisions during the entire year. Accordingly, our findings indicate that the consistency between the information conveyed by reported earnings and the information relevant to investors has decreased over time, irrespective of the quality of analysts' forecasts. Nor is the increase in the availability of nonaccounting information to investors solely responsible for the decrease in earnings usefulness, as indicated by the declining earnings response coefficient.10 1.2
THE CASH FLOWS-RETURNS RELATION

Cash flows are often claimed to be more informative than earnings because they are less subject to managerial manipulation than accrual earnings, and because they are less affected by questionable accounting
10An increase in the variability of stock returns may also have contributed to the weakening of the returns-earnings association. See Francis and Schipper [1999] on this issue.

BOUNDARIES

OF FINANCIAL

REPORTING

359

rules, such as those which require the expensing of investments in intangibles. To probe this claim we estimate the following cross-sectional regression for each sample year (1977-96):
Rit = + + PO PICFit + f2ACPFit+ P33ACCit 34AACCit+ ,it, = firm i's stock return for fiscal year t. (2)

where:
Rit

= cash flow from operations and the yearly change in cash flow from operations, respectively.11 = ACCitand AACCit annual reported accruals and the change in annual accruals, where accruals equal the difference between reported earnings and cash flow from operations. The four independent variables in (2) are scaled by the beginning-ofyear market value of equity. Regression (2) thus estimates the association between annual stock returns, on the one hand, and operating cash flows plus accounting accruals (the difference between earnings and cash flows), on the other hand. Table 2 reports yearly coefficient estimates of this regression. Our results indicate that the association between operating cash flows (plus accruals) and stock returns, as measured by R2, is not appreciably stronger than the association between earnings and returns (R2s in table 1).12 As to the pattern of temporal association, the R2s of both the total and constant samples in table 2 decrease over the period examined, although only the former is statistically significant at the .05 level (see the Timecoefficients in panel B of table 2). Similarly, the combined slope coefficients of the level and change of cash flows (1I + 12 in expression (2)), denoted as CFRC,tend to decrease over time, although only the decrease of the constant sample is statistically significant at the .05 level, as evidenced by the Time coefficients in panel B. As was the case with earnings, the R2s of the constant sample are substantially larger than those of the total sample, indicating that cash flows are more informative for firms with long operating histories. To summarize, for a broad cross-section of firms operating cash flows do not augment appreciably the informativeness (usefulness) of accrual earnings to investors.13 The declining association with stock returns documented in the preceding section for earnings is also evident for cash flows, though it is less pronounced. We believe that the milder decline in CFjtand A CFjt
11 Since in the early sample period firms did not report cash flow from operations, we computed this item as follows: Cash Flow from Operations = Net Income before Extraordinary Items + Depreciation + Annual Deferred Taxes - Annual Change in Current Assets - Cash + Annual Change in Current Liabilities - Current Maturities of Long-Term Debt. 12 A similar result was noted by Livnat and Zarowin [1990] and Bowen, Burgstahler, and Daley [1987]. 13 It may still be the case that in special circumstances (e.g., financially distressed companies) cash flows provide incremental information over earnings.

360

BARUCH

LEV AND

PAUL ZAROWIN

TABLE

The Association between Cash Flows and Stock Returns Estimates from Yearly Cross-Sectional Regressions of Annual Stock Returns on Operating Cash Flows + Accruals

+ + Panel A: Equation (2): Rit = Po + PICFit+ P2ACFit P3ACCit 4AACCit+ t Total Sample Constant Sample Yearl 1979 1980 1981 1982 1983 1984 1985 1986 1987 1988 1989 1990 1991 1992 1993 1994 1995 1996 Number of Observations 3,276 3,432 3,571 3,945 3,948 4,169 4,163 4,098 4,361 4,361 4,232 4,179 4,097 4,321 4,543 4,953 5,142 4,953
R2 CFRC R2 CFRC

0.074 0.052 0.124 0.059 0.041 0.111 0.092 0.063 0.052 0.064 0.078 0.058 0.042 0.052 0.048 0.071 0.051 0.036

0.750 0.574 0.853 0.560 0.536 0.679 0.573 0.515 0.518 0.496 0.642 0.472 0.467 0.548 0.666 0.685 0.704 0.416

0.074 0.065 0.187 0.091 0.068 0.240 0.146 0.122 0.114 0.110 0.134 0.124 0.054 0.057 0.090 0.136 0.163 0.029

0.772 0.797 1.857 1.112 0.869 1.169 0.918 0.939 0.916 0.447 1.014 0.823 0.434 0.535 0.991 0.928 0.949 0.288

Panel B: Time Regressions


a + b (Timet) + ct; t = 1979-96 R= CFRCt = a + b (Timet) + ct; t = 1979-96

(t-values in parentheses)
a b R2

Total Sample R2

0.242 (2.77) 1.159 (2.62)

-0.002 (-2.04) -0.006 (-1.28)

0.16

CFRC

0.04

Constant Sample R2

0.241 (1.13) 3.424 (2.72)

-0.001 (-0.61) -0.029 (-2.03)

0.00

CFRC

0.15

'The time series for cash flows starts with 1979 since the number of observations for 1977 (required for the cash flow change of 1978) was unusually low. The total sample includes all firms with the required data on Gompustat'sCurrentand Research Files. The constant sample includes about 1,000 companies with the required data on Compustatfor the 19-year sample period (1978-96). Variable definitions for panel A: Rt = annual stock return of firm i in fiscal t, CFit and ACFit= level = and change of cash flows from operations for year t, and ACCit and AACCit level and change of accruals (earnings minus cash flows from operations) for year t. CFit,A CFit,ACCitand AACCit are scaled by market value of equity at the beginning of t. CFRC= combined slope coefficients of the cash flow variables;

P, +

f2 in (2).

Variable definitions for panel B: Rt2 and CFRCt are derived from panel A. Timetis a year variable, 1978-96.

BOUNDARIES

OF FINANCIAL

REPORTING

361

TABLE 3 The Association betweenStockPrices and Book Values+ Earnings Estimatesof Yearly Cross-Sectional Regressionsof StockPrices on Earnings + Book Values
Panel A: Equation (3): Pit = a0 + a, Eit + a2 BVit + Fit

Year 1977 1978 1979 1980 1981 1982 1983 1984 1985 1986

R2 0.923 0.932 0.796 0.866 0.867 0.867 0.899 0.832 0.887 0.874

Year 1987 1988 1989 1990 1991 1992 1993 1994 1995 1996

R2 0.993 0.837 0.525 0.538 0.780 0.469 0.546 0.558 0.560 0.618

Panel B: Time Regressions


R2 = a + b (Timet) + ct; t = 1977-96

Total Sample R2

a 2.649

b -0.022

R2 0.57

(7.09) (-5.07) Pit,Eit, and BVit= share price at end of the fiscal t, earnings per share, and book value per share, respectively, of firm i in fiscal t. The sample includes all firms with the required data on Compustat's Current and Research Files, an average of 5,500 firms per year. Rt2is the adjusted R2 from panel A. Timetis a year variable, 1977-96.

the informativeness of cash flows relative to earnings is due to cash flows' relative immunity to the effects of some change-related items, such as accrued restructuring charges. 1.3
FROM STOCK RETURNS TO PRICES

Following Ohlson [1995], it has become popular in accounting research to examine the relevance of financial data by regressing stock prices on earnings plus book value:14
Pit =

ao + alxEit + a2BVit

+ cit,

t =

1977-96

(3)

where: Pit = share price of firm i at end of fiscal year t, Eit = earnings per share of firm i during year t, BVit = book value (equity) per share of firm i at end of t, Fit = other value-relevant information of firm i for year t, independent of earnings and book value. As indicated in table 3, the association between stock prices and earnings + book value, as measured by R2, decreased during 1977-96, from
14Strictly speaking, OhIson's model relates prices to book value plus the present value
of excess earnings.

362

BARUCH

LEV AND

PAUL ZAROWIN

R2 levels of 0.90 in the late 1970s, to 0.80 in the 1980s, and to 0.50-0.60 in the 1990s. A regression of the yearly R2s on a Timevariable (panel B) yields a negative and statistically significant Time coefficient (-0.022, t = -5.07). The estimates reported in table 3 pertain to the total sample. We obtained similar results for the constant sample (1,130 firms per year): the estimated coefficient of Timefrom a regression of annual R2 on Time is: -0.016 (t = -4.04). This finding of decreasing value relevance of earnings + book value is consistent with our previous results derived from the returns-earnings and returns-cash flow relationships. Collins, Maydew, and Weiss [1997], who estimated regression (3) over the 1953-93 period, reached the conclusion that the combined value relevance of earnings and book values has not decreased. The source of the inconsistency appears to lie in the periods examined. Our focus is on the last portion of Collins, Maydew, and Weiss's 40-year sample period. In the March 1996 version of their paper they report yearly coefficient estimates and R2s of regression (3). We regressed these R2s for the period 1977-93 on Time and obtained a negative coefficient (-0.003, t = -0.53). Our sample period extends Collins, Maydew, and Weiss's by three years (1994-96), each having a low R2 (see table 3). This extension may have increased the statistical significance of our negative Timecoefficient (panel B of table 3).15 Thus, while the association between stock prices and earnings + book value may have been stable over the past 40 years, our evidence indicates that it has decreased over the latter half of that period. The temporal association between capital market variables and financial data has also been studied by Francis and Schipper [1999], Ely and Waymire [1996], Ramesh and Thiagarajan [1995], Chang [1998], and Brown, Lo, and Lys [1998]. While all these studies report a weakening returns-earnings association when the association is measured by R2, results from levels regressions (price on earnings + book value) are mixed. Collins, Maydew, and Weiss [1997] and Francis and Schipper [1999] report a stable association over the 40+ years 1951-93. In contrast, Chang [1998], using various alternative methodologies, concludes that the value relevance of earnings and book value has decreased over the past 40 years. A decreasing association between price and earnings + book value is also reported by Brown, Lo, and Lys [1998], accounting for scale differences. Our overall results indicate a weakening of the association between market values and accounting information (earnings, cash flows, and book values) over the past 20 years.

15In their published version (table 3), Collins, Maydew, and Weiss report an average R2 of expression (3) of 0.754 for the period 1983-93. The corresponding average R2 of our estimates (our table 3) is 0.744. It appears, therefore, that our estimates of regression (3) conform closely to those of Collins, Maydew, and Weiss (which is not surprising, given the identical source of data).

BOUNDARIES

OF FINANCIAL

REPORTING

363

2. Business Change and the Deterioration of Financial Statement Usefulness


We contend that the increasing rate of change experienced by business enterprises, coupled with biased and delayed recognition of change by the accounting system, is the major reason for the documented decline in the usefulness of financial information. Empirical support for this contention is provided in this section. We first document the increasing rate of change affecting business enterprises and then explore the implications of business change for the usefulness of accounting information.
2.1 MEASURING BUSINESS CHANGE

While surveys of executives, investors, and policymakers generally support a perception that the business environment is changing at an ever-increasing rate (e.g., Deloitte & Touche [1995]), there is little empirical support for this view. We document the pattern of business change for our sample companies by ranking them on two indicators of value: book value of equity at fiscal year-end and market value of equity at year-end. We then classify the sample firms for each year and value indicator into ten equal-sized portfolios based on the rank of book value or market value. We measure the rate of business change by the frequency and magnitude of portfolio switches, namely, firms moving over time from one value portfolio to another. Specifically, we measure firm j's "absolute rank change" by its movement across portfolios from year t - 1 to year t. For example, if firm j is in book value portfolio 1 in 1977 and moved to portfolio 4 in 1978, its rank change measure is 3. For each year and value indicator, we calculate a yearly "mean absolute rank change" (MARC) which reflects the aggregate portfolio switches experienced by all the sample firms in that year.16 Our change measure will be low (zero at the limit) when portfolio membership is stable, whereas when firms bounce a lot from year to year across portfolios, the change measure will be high. The number of sample companies changes over the period examined as new firms become public or existing ones merge or go bankrupt. Our approach to measuring the rate of business change is similar to Stigler's way of estimating optimal firm size by observing over time shifts in the size distribution of firms within an industry (Stigler [1966, pp. 159-60]). If a specific size is optimal (yielding maximum economies of scale), firms should converge over time to that size. Our approach is also similar to the use of Markov Chains transition matrices to study social and occupational mobility issues (e.g., Kemeny and Snell [1967, pp. 191-200]). Table 4 presents the yearly "mean absolute rank change" (MARC) measures for the sample companies. Data for market value rankings are
16Results based on multiyear changes (e.g., rank changes over three to five years) are similar to those reported in table 4.

364

BARUCH LEV AND PAUL ZAROWIN TABLE 4 The IncreasingRate of Business Change Mean AbsoluteValuesof Yearly Rank Changes (MARC)ExperiencedbyFirms Classifiedinto TenPortfoliosbyMarketand Book Values Panel A: Yearly Measures of Change (MARC) Market Value Portfolios MARC MARC Year Measure Year Measure
1978 1979 1980 1981 1982 0.404 0.384 0.429 0.545 0.568

Book Value Portfolios MARC Year Measure


1978 1979 1980 1981 1982 0.179 0.181 0.240 0.276 0.294

1964 1965 1966 1967 1968 1969 1970 1971 1972 1973 1974 1975 1976 1977

0.309 0.308 0.418 0.416 0.487 0.434 0.499 0.443 0.432 1.113 0.547 0.490 0.422 0.385

1983 1984 1985 1986 1987 1988 1989 1990 1991 1992 1993 1994 1995

0.624 0.583 0.550 0.588 0.539 0.516 0.500 0.565 0.528 0.587 0.584 0.517 0.526

1983 1984 1985 1986 1987 1988 1989 1990 1991 1992 1993 1994 1995 1996

0.374 0.317 0.382 0.410 0.390 0.324 0.237 0.309 0.387 0.409 0.487 0.401 0.417 0.536

Panel B: Time Regressions


Regression: MARC (Indicator) t = a + b (Timet) + ct; t = 1964-96 (t-values in parentheses) a 1964-95 0.0026 b 0.0062 R2 0.48

Dependent

Variable

MARC (Market Value),

(0.03) MARC (Market Value), 1978-95 -0.0693 (-0.31) -0.8357


(-4.54)

(5.33) 0.0069 (2.63) 0.0138


(6.44)

0.25

MARC (Book Value), 1978-96

0.69

Sample firms are classified in each year into ten portfolios according to their market value of equity (MV) and alternatively by their book value of equity (BV). MARC (mean absolute rank change) indicates the average frequency of firms switching value portfolios from the past year to the current year, as well as the number of portfolios switched (i.e., magnitude of switch) bv each firm. MARC= / {=iDECit - DEC1t_1I} N1, where DECit and DECit-1 = decile rank (of book value or market value) for firm i in years t and t - 1, Nt= number of firms in year t, and I means summation over all firms in t. The observations for the market value portfolios are all firms on the CRSPDaily File with share price and number of shares at the end of years t and t - 1, an average of 4,000 firms per year. The observations for the book value portfolios are all firms on Compustat'sCurrentand Files with book value of equity at end of years t and t - 1, an average of 5,800 Research firms per year. MARCtis the yearly, mean absolute rank change in panel A. Time,is a year variable.

BOUNDARIES

OF FINANCIAL

REPORTING

365

derived from the CRSP database (1963-95) and those for book value rankings are derived from Compustat(1977-96). The generally increasing MARC measures in panel A for both market and book value classifications indicate that the frequency of firms switching across portfolio rankings has increased over the past 20-30 years. For market value rankings, the change measures increase from 0.3-0.4 in the 1960s to 0.5-0.6 in the 1990s. For book value rankings, the change measures increase almost monotonically from 0.2-0.3 in the late 1970s and early 1980s to 0.4-0.5 in the 1990s. These increases are statistically significant, as evidenced by the t-values of the three slope coefficients of the Timeregressions in panel B. For example, the Time coefficient of the book value change measures over 1978-96 is 0.014 (t = 6.44). Our evidence thus supports the perception of executives and investors concerning the increasing rate of change experienced by U.S. business enterprises. We argue that the increasing rate of business change over the past two or three decades and the deficient accounting treatment of change have contributed to the documented decline in the usefulness of financial information. Essentially, while the accounting system is primarily based on the reporting of discrete, transaction-based events, such as sales, purchases, and investments, the impact of change on business enterprises is rarely triggered by specific transactions.17 Change, internally (e.g., product development) or externally (e.g., deregulation) driven, often affects enterprise value long before revenue or expense transactions warrant an accounting record. Investors generally react to the impact of change on business enterprises in real time, hence the increasing disconnection between market and accounting values. For example, during the late 1980s and early 1990s, the regional telephone service in the United States has been deregulated, gradually transforming the industry from a monopolistic to competitive environment. Investors reacted to the deregulation, which increased the risk and decreased the expected revenues of the formerly monopolistic telephone companies, as it occurred, while the impact of deregulation on eventshas been minor for years.18 Only in 1994-95, accounting recordable roughly half a decade after the beginning of deregulation, did the regional telephone companies (Baby Bells) write off $26 billion in assets as a consequence of the deregulation. We illustrate the impact of change driven by telecommunications deregulation on the usefulness of earnings by documenting a decline in
17The exception to transaction-based accounting entries is end-of-period adjusting entries, such as those reflecting depreciation and doubtful receivables. 18 The significant and adverse reaction of investors to the impact of deregulation on telephone companies is evident by the stock performance of the regional telephone companies, which considerably lagged the market return. The average five-year (1991-95) cumulative return of the phone companies' stocks was 93.25%, while the market return (CRSPvalueweighted average) over the corresponding period was 119.59%.

366

BARUCH

LEV AND

PAUL ZAROWIN

the statistical association between stock returns of the regional teleand during dereguphone companies and their reported earnings before lation. We estimate for these companies the returns-earnings regression (1) for both the prederegulation period (1984-89) and the deregulation period (1990-96). The regressions are pooled time series and crosssection, with fixed effects for both time (year) and firms. The estimated R2s are: 0.93 for the prederegulation period versus 0.72 for the deregulation period (note that the fixed effects substantially increase R2). The estimated combined slope coefficients (ERCs) are 1.85 for the prederegulation period versus 0.68 for the deregulation period. The former combined slope coefficient (1.85) is significantly different from zero at the 0.02 level, while the latter coefficient (0.68) is insignificantly different from zero at the .10 level. Reported earnings of telephone companies have clearly become less value relevant to investors during and after the fast change period due to deregulation. Business change is primarily driven by increased competition and innovation. In contrast to the delayed reaction of the reporting system to deregulation, when change is driven by competition and innovation, the accounting system front-loads the costs and delays the recognition of benefits. For example, restructuring costs, such as those for employee training, production reengineering, or organizational redesign, are immediately expensed, while the benefits of restructuring, in the form of lower production costs and improved customer service, are recognized in later periods. Consequently, during restructuring, the financial statements reflect the cost of restructuring, but not its benefits, and are therefore largely disconnected from market values which reflect the expected benefits along with the costs.19 Similarly, the immediate expensing of investment in innovation (e.g., R&D) -the major change-driver in technology- and science-based companies-is both biased and inconsistent.20 Costs of innovation are recognized up front, while benefits are recorded in subsequent periods. To complicate things further, the accounting for intangibles is beset by inconsistencies. For example, a firm which develops an instrument for internal use will expense all development costs, but if the firm buys a similar instrument, it will be capitalized. Thus, change-drivers, such as deregulation, competition, and innovation, adversely affect the matching of costs with revenues, leading to a decrease in the value relevance of financial information. We next pro19Indeed, event studies (e.g., Francis, Hanna, and Vincent [1996]) indicate that sometimes investors' reaction to the restructuring charges is in fact positive. 20While it is generally believed that the expensing of intangibles is conservative, leading to lower reported profitability than under capitalization, for mature firms immediate expensing is in fact aggressive. Specifically, when the growth rate of investment in intangibles is lower than the firm's return on equity (ROE), the expensing of intangibles results in higher ROE and ROA than if the intangibles were capitalized. Intangibles' expensing also often inflates the rate of growth of reported earnings (see Lev, Sarath, and Sougiannis [1999] and Merck's example in Lev and Sougiannis [1996, appendix]).

BOUNDARIES

OF FINANCIAL

REPORTING

367

vide empirical support for the link between the increasing rate of business change and the decline in the informativeness of earnings. 2.2
CHANGE AND THE VALUE RELEVANCE OF EARNINGS

Our firm-specific change measure (the "absolute rank change") is based on the frequency and extent of over time movement of firms among portfolios formed by ranking on book value and market value of equity. To link this change measure to the documented temporal decrease in the informativeness of earnings, we first compute for each sample firm the across time absolute rank change, reflecting the number of times the firm switched book value portfolios during 1977-96, as well as the extent of such switches. To standardize the firm-specific measure, we scale it by the number of years the firm existed in the sample. For example, if firm j was in the top book value portfolio during 1977 to 1983, the second (next to top) portfolio in 1984-91, and the fifth portfolio from 1992 to 1996, its rank change indicator is 0.20 (one point for the single rank switch in 1984 plus three points for the three-rank switch-from portfolio 2 to 5-in 1992, divided by the 20 years of the firm in the sample). We classify the sample firms into two groups: stable and changing companies. The first group includes the No Change firms (about 1,000), namely, those that remained in the same portfolio during the entire sample period (1977-96). The second group-the Changefirms-includes the remaining sample (ranging from 3,000 in the early sample years to 5,500 in the mid-1990s). Alternatively, we classify firms into Low Change-firms with a firm-specific "absolute rank" change indicator < .10 (including, of course, the No Changefirms) and High Change-the remaining sample.2' Next, we examine the yearly cross-sectional returns-earnings regression (1) separately for the stable and changing firms. If change decreases the informativeness of earnings, the regression's R2 and combined slope coefficients (ERC) should be larger for stable firms (a stronger returnsearnings association) than for changing ones. Furthermore, given our evidence that the rate of change of business enterprises increased during the past 20 years, that increase clearly affected the changing firms more of than stable ones. Thus, we predict that the rate of decrease R2 and ERC over 1977-96 should be higher for changing firms than for stable ones. Results in table 5 support both expectations. Panel A of table 5 reports yearly estimates of R2s and combined slope coefficients for the four change classifications of firms, and panel B reports means and medians of the 19 yearly estimates. The data corroborate our first expectation: both the means and medians of the yearly R2 and ERC are larger for No Changefirms than for Changefirms (e.g., mean R2 of 0.124 versus 0.097 and mean ERCof 1.22 versus 1.02). Similarly, the
21Estimates based on other rank change cutoffs, such as 0.20, 0.30, yield results similar to those reported in table 5.

368

BARUCH LEV AND PAUL ZAROWIN TABLE 5 Business Change and Earnings Informativeness Estimatesfrom Annual Regressionsof StockReturns on the Level and Changeof Annual Earnings for Firms Classifiedby the Rate of Business Change

Panel A: Yearly Estimates of Regression (1)-Returns on Earnings No Change Change Low Change R2 R2 R2 Year ERC ERC ERC 1978 0.11 1.28 0.12 1.17 0.13 1.36 1979 0.16 1.96 0.09 1.02 0.15 1.85 0.79 1980 0.05 0.08 1.07 0.06 0.90 1981 0.26 1.69 0.11 1.03 0.28 1.81 1982 0.08 1.05 0.09 1.12 0.10 1.20 1983 0.09 1.11 0.07 1.01 0.07 1.02 1984 0.23 0.77 0.16 1.13 0.23 0.91 0.13 1985 0.93 0.17 1.08 1.05 0.14 1986 0.12 1.23 0.12 1.04 0.15 1.24 1987 0.03 0.65 0.04 0.07 0.87 0.75 1988 0.08 1.05 0.07 0.12 0.94 0.91 1989 0.14 1.14 0.10 0.96 0.11 1.23 1990 0.12 0.83 0.10 0.89 0.10 0.96 1991 0.08 0.99 0.09 0.96 0.11 1.11 1992 0.16 1.53 0.09 0.95 0.15 1.58 1993 1.68 0.12 0.08 1.49 1.06 0.11 1994 0.22 1.80 0.10 1.05 0.17 1.47 1995 0.10 1.22 0.06 1.01 0.12 1.43 1996 0.07 1.41 0.06 0.96 0.06 1.27

High Change R2 ERC 0.12 1.15 0.09 1.01 0.09 1.09 0.10 1.00 0.09 1.10 0.07 1.03 1.10 0.15 0.14 1.07 0.11 1.03 0.86 0.07 0.94 0.12 0.96 0.10 0.89 0.10 0.09 0.95 0.09 0.93 1.07 0.08 0.10 1.06 1.00 0.06 0.96 0.06

Panel B: Means and Medians of Annual R2 and ERCby Change Group (1977-96) R2 ERC Mean Median Mean Median Change Group 1.22 No Change 1.14 0.124 0.116 versus Change 0.097 0.095 1.02 1.02 p = 0.09 p = 0.03 Significance of Difference Low Change
versus

0.124 0.096 p = 0.06

0.114 0.093

1.24 1.01 p = 0.01

1.23 1.02

High Change Significance of Difference

R2 and ERCs of the Low Changefirms are larger than the association measures of the High Changefirms. To determine the statistical significance of the difference in means of change groups we regressed the R2 and ERCs of the combined Changeand No Changegroups (and also of the combined Low and High Changegroups), that is, 38 observations in each regression, on a 0-1 dummy variable reflecting membership in a change group. The significance levels of the dummy variables, all lower than 0.10, are reported in panel B of table 5. Thus, the rate of business change is negatively associated with the informativeness of earnings, as measured by the extent of the returns-earnings association. Panel C of table 5 presents estimates from regressions of the yearly R2 and ERCs (reported in panel A) on a Timevariable. The data confirm our second expectation that the temporal decline in the returns-earnings

BOUNDARIES OF FINANCIAL REPORTING


TA B L E 5 -continued

369

Panel C: Regressions of Yearly R2and ERC (from panel A) on Time (t-values in parentheses) R2 Intercept Time
No Change

R2 ERC
Change

0.204 (0.88) 0.514 (0.37) 0.256 (2.86) 1.636 (6.31) 0.330 (1.49) 1.180 (1.02) 0.246 (3.01) 1.585 (6.14)

-0.001 (-0.35) 0.008 (0.50) -0.002 (-1.78) -0.007 (-2.39) -0.002 (-0.93) 0.001 (0.05) -0.002 (-1.84) -0.007 (-2.23)

0.00 0.00

R2 ERC
Low Change

0.11 0.21

R2 ERC
High Change

0.00 0.00

R2 ERC

0.12 0.18

Firms classified as No Changedid not switch book value ranking during 1977-96, while Changefirms are the rest of the sample. Firms classified as Low Change have an "absolute rank change" indicator (defined in section 2.2) < .10, while High Changefirms are the rest of the sample. ERC= combined slope coefficients or "earnings response coefficient," namely, the sum of the estimated regression coefficients of level and change of earnings. The sample includes all firms with the required data on Compustat's Currentand ResearchFiles, an average of 4,000 and 700 firms per year for the Change and No Change groups, respectively, and an average of 3,700 and 1,000 firms per year for the High Change and Low Changegroups, respectively.

association is more pronounced for changing than for stable enterprises. The four coefficients of Timein the R2 and ERC regressions for both the No Changeand the Low Changegroups are not statistically significant (see t-values in parentheses), and the R2s of those four regressions equal zero, indicating essentially no deterioration over time in the returns-earnings association of stable companies. In contrast, the four Timecoefficients of both the Changeand High Changegroups are all negative and statistically significant, and the four R2s of these Time regressions range between 0.11-0.21, indicating that the association between returns and earnings of changing firms has declined over the past 20 years. The significance levels of the differences in Time coefficients between the No Changeand Changegroups (and the Low and High Changegroups) were determined by running a regression on the combined observations of the two groups with a dummy variable for group membership. P-levels of the dummy variable are: R2 regression (No Change versus Change) = 0.09, ERC regression (No Changeversus Change) = 0.03, R2 (Low versus High Change) = 0.06, and ERC (Low versus High Change) = 0.01.

370

BARUCH

LEV AND

PAUL ZAROWIN

To summarize, we have argued that the increasing rate of business change coupled with the ineffectiveness of the accounting system in reflecting the consequences of change contributed significantly to the temporal decline in the value relevance of accounting information. We have empirically established this link by providing evidence that the rate of change experienced by business enterprises increased over the 197796 period, and that the informativeness of earnings is negatively related to the rate of business change. We next consider an alternative explanation for the declining usefulness of earnings, namely, the increasing incidence of reported losses and nonrecurring items (the two are, of course, related). For example, Hayn [1995] reports that losses account for some of the observed decline in the slope coefficient of the returns-earnings regression (ERC) and Collins, Maydew, and Weiss [1997] attribute a shift in value relevance from earnings to book values to both the increasing significance of one-time items and the increasing frequency of losses. However, we believe that both reported losses and nonrecurring items are often the symptomsrather than the causes of the decline in informativeness of financial information. Specifically, it is the previously discussed failure of the accounting system to reflect change in a meaningful and timely manner that often leads to both reported losses and nonrecurring items. Examples include restructuring charges, where the reported loss is often an investment in future benefits (e.g., operating efficiencies), the chronic reported losses (to the mid-1990s) of cellular phone companies caused by the immediate expensing of customer acquisition costs (Amir and Lev [1996]), and the large losses and nonrecurring items reported by acquiring companies writing off all the purchased In-Process R&D (Deng and Lev [1998]).22 Thus, we argue, it is not losses or nonrecurring items that weaken the returns-earnings relation, rather it is the underlying failure of the current accounting system to meaningfully account for change, which is often manifested by losses and nonrecurring items. Given recent research attention to reported losses, we examined the role of such losses in our estimation of the association between the rate of business change and the decline in earnings' informativeness (table 5). Specifically, we added to the Timeregressions reported in panel C of table 5 the percentage of firms with negative EPS in each year. The regressions are thus: Estimated R2 (or ERCt) = a + b Timet + c (% Losses) + ?, t = 1978-96. (4) We estimated this regression for both the Low Change and High Change companies and found the coefficients of percentage losses (c in regression (4)) in each of the four regressions (R2 and ERCfor Low and High
22An example is IBM's reported loss of $538 million in the third quarter of 1995, due to the write-off of $1,840 million of acquired R&D-in-process (in the same quarter a year earlier, IBM reported positive earnings of $710 million). Obviously, the $538 million loss does not indicate any deterioration in IBM's operations.

BOUNDARIES

OF FINANCIAL

REPORTING

371

Change companies) to be statistically insignificant at the 0.10 level. For the Low Changefirms the estimated coefficients of percentage losses for R2 and ERC are 0.157 (t = 1.45) and -0.451 (t = -1.44), and for the High Changefirms the estimated coefficients of percentage losses for R2 and ERC are 0.132 (t = 1.42) and -0.462 (t = -1.60), respectively. Thus, the increasing incidence of reported losses in our sample of High Change firms does not alter our conclusion concerning the relationship between change and the decreasing value relevance of earnings.

3. Intangibles, Innovation, and Change


Intangible investments, R&D in particular, are generally considered as the major driver of business change, creating new products, franchises, and improved production processes. However, while some intangible investments trigger change, others are just aimed at preserving the status defined as "spending aimed at learning more quo. Thus, appliedresearch, about the technology process a firm is alreadyusing, or about a good that it is already producing"(Jovanovic and Nyarko [1995, p. 5, emphasis ours]), is generally intended to sustain an existing competitive position, not to defined as "spendchange the firm's operations. In contrast, basicresearch, ing directed towards processes not yet in use, or goods not yet produced" (Jovanovic and Nyarko [1995, p. 6, emphasis ours]) is clearly aimed at initiating change.23 Given these disparate intentions and consequences of R&D, the levelor intensity (R&D spending to sales) of R&D is not necessarily associated with change and the consequent loss of informativeness of financial data. From an accounting measurement perspective, too, the level of R&D expenditures does not necessarily affect the informativeness of earnings. Thus, if the rate of R&D spending is constant over time, reported earnings are invariant to the accounting treatment of R&D; earnings will be the same whether R&D is capitalized and amortized or immediately expensed. These statements are relevant to results in several recent studies (e.g., Collins, Maydew, and Weiss [1997] and Francis and Schipper [1999]) on the association between the intensity of intangibles and the R2 of the returns on earnings or the stock price on earnings + book value associations. Both papers report that firms intensively engaged in intangible investments were not characterized by a lower association between stock prices (or returns) and financial data than firms less intensive in intangibles.24
23The pharmaceutical industry provides an example of these two types of R&D. Basic research is aimed generally at developing New Molecular Entities (NMEs), which are entirely new drugs capable of drastically changing the firm's product mix and competitive position. Applied pharmaceutical R&D, the development of "me too" drugs, is aimed at modifying existing drugs or changing the route of administration, thereby essentially preserving the firm's competitive position. 24Collins, Maydew, and Weiss [1997] do, however, find that the increased importance of intangible-intensive firms is associated with a shift in value relevance from earnings to book values.

372

BARUCH

LEV AND

PAUL ZAROWIN

However, it is not a high level of intangible investment, indicated in those papers by a firm's membership in a high-tech industry or by a high intensity of R&D, that is expected to cause a decrease in the informativeness of financial information. In a steady-state R&D environment, the immediate expensing of R&D will result in the same earnings as those based on R&D capitalization; hence no loss of earnings informativeness can be ascribed to R&D in this case.25 It is only when the investment rate in intangibles changesover time that reported earnings based on immediate expensing will differ materially from economic earnings based on capitalization of intangibles. Accordingly, we conjecture that the expensing of R&D (the current GAAPpractice) by firms with increasinginvestment rate in R&D is associated with a decline in the informativeness of reported earnings. To examine this conjecture, we split our sample period, 1976-95, into three subperiods and compute for each sample firm the average R&D intensity (R&D to sales) in the "recent period" (1989-95) relative to the average R&D intensity in the "early period" (1976-83). Sample firms were then classified by the direction of change in R&D intensity into four categories: Low-Lowfirms, with R&D intensity of .01 or lower in both the early and recent periods; High-High firms, with R&D intensity exceeding .01 in both periods; Low-Highfirms, with R&D intensity below .01 in the early period and above .01 in the recent period; and High-Low firms, the converse of Low-High.26 We then reestimated the cross-sectional returns on earnings (level and change) regression (1) for each of the four groups of firms for every year, 1976-95. The average yearly regression estimates of R2 and the combined slope coefficients (ERCs), over the early sample period (1976-83) and the recent period (1989-95), are reported in table 6. The main diagonal of table 6 reports mean regression estimates for stable R&D companies: Low-Lowand High-High. The mean R2s of both groups decreased from the early (1976-83) to the recent (1989-95) period: from 0.137 to 0.099 for Low-Low, and from 0.156 to 0.126 for High-High companies. These decreases in average R2s for both groups, however, are statistically insignificant (at the .05 level). The average ERCs of the Low-Lowand High-High groups also decreased between the two periods: from 1.44 to 0.820 for Low-Low(t= -6.28) and from 1.94 to 1.14 (t = -3.45) for High-High. Consistent with the findings of other authors (e.g., Collins, Maydew, and Weiss [1997]), the R2s and ERCs of the Highintensive in R&D-are larger than those High subsample-companies
25 This steady-state R&D environment is not an unrealistic case, as researchers report that the firm-specific time series of R&D of most firms are remarkably stable (e.g., Helfat [1994]). 26 Results based on an R&D intensity cutoff of 0.5% (0.005) are similar to those reported in table 6.

BOUNDARIES TABLE

OF FINANCIAL 6

REPORTING

373

R&D Changeand Earnings Informativeness Estimatesfrom Returns-Earnings Regressions(1) for Firms Classified by theDirection of Changein TheirR&D Intensityfrom theEarly SamplePeriod (1976-83) to the RecentSamplePeriod (1989-95) Arrowsin Each Panel Indicate the Change in the Measure (R2 or ERC)from theEarly to the RecentPeriod Early Sample Period (1976-83) Recent Sample Period (1989-95) Low R&D
MED R&D Mean R2 0.000 0.000 0.016 0.003

Low R&D

High R&D

Mean ERC High R&D MED R&D Mean R2 Mean ERC

0.137 0.099 1.440 0.820 1,259 0.004 0.233 2.170 96 0.018 0.126 1.060

0.080 0.750 67 0.034 0.156 1.940

0.178 1.290

0.044 0.126 1.140 455

Low R&D are firms with an R&D intensity < 0.01 (1% of sales) and High R&D firms are those with R&D intensity 2 0.01. MED R&D = median R&D intensity (R&D over sales) in the early and recent periods of firms in the panel. Mean R2 = mean over 1976-83 and 1989-95 of adjusted R2 of the yearly returns on earnings cross-sectional regression (1). Mean ERC = mean of combined slope coefficients (al + a2 in (1)) over 1976-83 and 1989-95. # = number of firms.

of the Low-Low companies, confirming our earlier contention that high yet stable R&D spending does not induce a weak earnings-returns relation. The lower-left panel of table 6 presents results for Low-High companies, those characterized by an increasing rate of R&D expenditures. As indicated in the table, the median R&D intensity of these companies increased from 0.4% during 1976-83 to 1.8% in 1989-95. These firms experienced a sharp decline over the sample period in the returns-earnings R2 from 0.233 to 0.126 (t = -2.03) and in the combined ERC from 2.17 to 1.06 (t = -2.29). In contrast, High-Low firms, whose R&D intensity decreased from a median of 1.6% to 0.3%, experienced a borderline significant increasein the association between returns and earnings: the R2 of these firms increased from 0.080 to 0.178 (t = 1.54, p = .15) and the ERC increased from 0.75 to 1.29 (t = 1.61, p = .13). Thus, while the informativeness of earnings of all the sample companies decreased during the past 20 years, an increase in R&D intensity (Low-Highfirms) is associated with an abnormally steep decrease in earnings informativeness. In addition, a portion of the weakening of the returns-earnings association of the stable firms (Low-Low and HighHigh) may also be related to R&D increases, as the R&D intensity of both groups increased within our cutoff of 1% R&D intensity. For example, as indicated in table 6, the median R&D intensity of the HighHigh group increased from 3.4% to 4.4%. We conclude that while R&D

374

BARUCH

LEV AND

PAUL ZAROWIN

TABLE

R&D Increases and Earnings Informativeness Estimates from Regressions of Yearly R2 and ERCs, Derived from Cross-Sectional Returns-Earnings Regressions, on a Time Variable, 1977-95 Regressions on TIME: R2 (or ERCt) = a + b(Timet) + c (t-values in parentheses) R&D-Stable a ERC Regressions 3.53 (-7.09) 0.465 (3.54) b -0.030 (-5.20) -0.004 (-2.81) Firms R2 .59 R&D-Increasing a 7.187 (6.09) 0.703 (3.95) b -0.067 (-4.91) -0.007 (-3.35) Firms R2 .56

R2 Regressions

.28

.36

R&D-increasing firms have R&D intensity 2 .01 in the period 1989-95. R&D-stable firms are all other firms. The sample includes all firms with the required data on Compustat'sCurrentand Research Files, an average of 2,200 and 900 firms per year for the R&D-stable and R&D-increasing groups, respectively.

change is not the only reason for the temporal decline in the informativeness of earnings, it appears to be an important one. The results reported in table 6 are based on firms that existed throughout the period 1976-95, and some subsamples contain relatively few firms (e.g., there are 96 Low-High firms). To overcome the survivorship bias and increase sample sizes, we created a new sample classification: We define as R&Dincreasingfirms with an R&D intensity exceeding .01 (1% of sales) in the recent subperiod (1989-95), and all other firms as R&D-stable. Thus, all sample firms (not just surviving ones) must be in one of either the R&D-increasing or R&D-stable (or no R&D) groups. The firms classified as R&D-increasing (having an R&D intensity > .01 during 1989-95) indeed increased in R&D; their mean R&D intensity increased from 0.032 in 1976-83 to 0.049 in 1989-95. In contrast, the mean R&D intensity of the R&D-stable firms over the corresponding periods decreased from 0.0024 to 0.0021. We reestimated the returns-earnings regressions (1) for each year (1976-95) separately for the R&D-increasingand R&D-stablegroups, and regressed the resulting R2s and ERCs on Time. We report the regression estimates in table 7. It is evident that the estimated Time coefficients (b) of the R&D-increasinggroup are more negative than those of the R&Dstablegroup: -0.067 versus -0.030 for the ERC regressions and -0.007 versus -0.004 for the R2 regressions.27 We conclude that the weakening of the returns-earnings association is more pronounced for firms whose R&D intensity increased over the sample period than for stable R&D companies. Table 7 results are somewhat weaker than those reported in

27 The difference between the ERCs Time coefficients in table 7 (-0.067 and -0.030) is statistically significant at the 0.01 level (t = 4.9), while the difference in the R2 Time coefficients (-0.007 and -0.004) is not statistically significant (t = 0.7, p = .46).

BOUNDARIES

OF FINANCIAL

REPORTING

375

TABLE 8 Firms'Rate of Change and TheirR&D Intensity Estimatesfrom Regressionsof Yearly AverageR&D Intensity on Timefor Firms GroupedbyRate of Business Change (t-values in parentheses)
Regression: Mean R&D Intensityt = a + b (Timet) + st, t = 1978-96

Coefficient Estimates
Change Group Mean R&D Intensity 0.015 0.030 a -0.0174 (-1.38) -0.1490 (-12.15) -0.0138 b 0.0004 (2.55) 0.0021 (14.61) 0.0003
R2

No Change versus Change

0.23 0.92

Low Change

0.013 0.032

0.31 0.91

versus
High Change

(-1.57)
-0.1567 (-11.37)

(3.03)
0.0022 (13.71)

The classification of firms into the four change groups is described in section 2. The sample includes all firms with the required data on Compustat'sCurrentand Research Files, an average of 4,000 and 700 firms per year for the Change and No Changegroups, respectively, and an average of 3,700 and 1,000 firms per year for the High Changeand Low Changegroups, respectively. The sample is the same as that used for table 5.

table 6, because the classifications used in table 6 are more effective in capturing R&D change than in table 7. For example, the mean R&D intensity of the Low-Highfirms in table 6 (not reported) increased from 0.004 in 1976-83 to 0.031 in 1989-95, while the mean R&D-intensity of the R&D-increasing firms in table 7 only changed from 0.032 to 0.049 during the corresponding periods. To complete the linkages between the declining usefulness of earnings and the rate of business change, which is partially driven by R&D increases, we analyze now the association between the rate of business change and the change in R&D expenditures. Specifically, we show that fast-changing firms experienced a larger increase in R&D-intensity than stable companies. Thus, for the No Changeand Changegroups, and for the Low Change and High Change groups, as previously analyzed in table 5, we examine the annual average R&D intensity. We expect that the average R&D intensity of changing firms is higher than that of stable firms and, more important, that the rate of increase in R&D intensity of changing firms is higher than that of stable companies. The data in table 8 confirm both expectations. First, the mean R&D intensity (over the 1978-96 period) of the Changegroup is larger than the intensity of the No Changegroup (0.030 versus 0.015; t = 5.2; p = .01), and the mean R&D intensity of the High Changegroup is larger than that of the Low Changegroup (0.032 versus 0.013; t = 6.3, p = .01). To examine the second expectation, we regress for each of the four change groups

376

BARUCH

LEV AND

PAUL ZAROWIN

the yearly mean R&D intensity of the group on Time.Results (presented in the table 8) indicate that the rate of increase in R&D intensity during 1978-96 was substantially larger for changing firms than for stable ones: The Timecoefficients of the Changeand High Changegroups, 0.0021 and 0.0022, are 5-7 times larger than the Time coefficients of the No Change and Low Change groups, 0.0004 and 0.0003, respectively. To determine the significance of the differences in the estimated time coefficients, we regressed the mean R&D intensity on Time, combiningthe No Changeand Changegroups (and the Low and High Changegroups), using a 0-1 dummy for group membership. Both differences in Timecoefficients, between No Changeversus Changeand Low versus High Change,are significant at the 0.01 level. Note also the large differences in R2s in table 8: The Timevariable explains almost perfectly the temporal variation in R&D intensity for changing firms (R2: 0.92 and 0.91), while for stable companies, Time provides only a partial explanation for the temporal variance in R&D intensity (R2: 0.23 and 0.31). To summarize, we have provided evidence supporting the following phenomena and relationships: (i) the rate of change experienced by U.S. business enterprises has increased over the past 20 years; (ii) the increasing rate of business change is associated with a decline in the informativeness of earnings; (iii) an increase in R&D intensity is associated with a decline in earnings informativeness; and (iv) an increase in the rate of business change is associated with an increase in R&D intensity. This evidence, we believe, supports the view that the documented decline in the usefulness of financial information was mainly caused by the increasing pace of change affecting business enterprises and the inadequacy of the accounting system in reflecting the consequences of change. Among change-drivers, innovation, generally brought about by investment in R&D, is an important factor in the declining usefulness of financial information. We next discuss two proposals for enhancing the usefulness of financial reports.

4. Improving the Usefulness of Financial Information


We discuss two proposals aimed at enhancing the usefulness of financial a practice information. The first-capitalization of intangibles-extends currently used in limited circumstances, while the second-a systematic restatement of financial reports-calls for a substantial modification of current reporting practices. 4.1
THE CAPITALIZATION OF INTANGIBLE INVESTMENTS

We believe that the almost universal expensing of intangible investments in the United States is inconsistent with the FASB's conceptual framework (Statement of Financial Accounting ConceptsNo. 6) and with empirical evidence. The conceptual framework defines an asset as: "probable future economic benefit obtained or controlled by a particular entity

BOUNDARIES

OF FINANCIAL

REPORTING

377

as a result of past transactions or events.... assets ... may be intangible, and although not exchangeable they may be usable by the entity in producing or distributing other goods or services.... anything that is commonly used to produce goods or services, whether tangible or intangible and whether or not it has a market price or is otherwise exchangeable, also has future economic benefit" (FASB[1985b, paras. 25, 26, 173]). Surely, the recognition of intangible investments with attributable future benefits as assets is within the boundaries of GAAPObjections to capitalization center on the uncertainty associated with the benefits of intangibles: "The uncertainty [e.g., about R&D] is not about the intent to increase future economic benefits but about whether and, if so, to what extent they succeeded [sic] in doing so" (FASB[1985b, para. 175]). Given the uncertainty concerns, it makes sense to recognize intangible investments as assets when the uncertainty about benefits is considerably resolved. It is well known that as projects under development advance, from formulation of the initial idea through increasingly demanding feasibility tests (e.g., alpha and beta tests) to the final product, the uncertainty of commercial success continually decreases. Accordingly, a reasonable balance between relevance and reliability of information would suggest the capitalization of intangible investment when the project successfully passes a significant technological feasibility test, such as a working model for software or a clinical test for a drug. Surely, uncertainty about the future benefits of a clinically proven drug is not larger than the uncertainty associated with the expected benefits of a record or a movie under production whose capitalized costs are recognized as assets by current accounting practices (SFASNos. 50 and 53; see FASB [1981 a; 1981 b]). This approach to capitalization was taken in SFAS No. 86 (software for sale), the major exception in the United States to intangibles expensing. A similar approach is followed, subject to certain constraints, by the recently enacted international standard for intangibles (IASC [1998]). Accordingly, we propose the capitalization of all intangible investments with attributable benefits which have passed certain prespecified technological feasibility tests. We depart from the software capitalization standard (SFASNo. 86) by proposing that once capitalization commences (post feasibility test), all the project-related, previously expensed R&D should also be capitalized. Given that the uncertainty about the project's viability has been substantially reduced, we see no reason for a different accounting treatment of pre- and postfeasibility R&D. Note that our capitalization proposal, which is conditioned on the achievement of technological feasibility, differs substantially from a mechanical capitalization (accumulation) of all past expenditures on intangibles, which can easily be replicated by investors from successive income statements. The proposed capitalization allows management to convey important inside information about the progress and success of the development program. Indiscriminate capitalization of all past R&D expenditures does not provide such information.

378

BARUCH LEV AND PAUL ZAROWIN

The proposed capitalization of intangibles is consistent with recent empirical evidence interpreted within the "residual earnings" valuation framework for analyzing accounting principles issues (Dietrich et al. [1997]). This valuation framework equates an enterprise's intrinsic value to its current book value plus the present value of residual earnings (reported earnings minus a charge for equity capital). Accordingly, accounting standards which improve the alignment of reported book value with the firm's intrinsic value (usually proxied by market value) and/ or improve the prediction of earnings should be preferred over standards which do not. Empirical evidence supports the notion that the recognition of intangibles as assets may achieve one or both of the above criteria for a preferred accounting standard. For example, Lev and Sougiannis's [1996] finding that capitalized values of R&D are significantly associated with stock prices (after controlling for reported book values) implies that R&D capitalization will improve the alignment of book values with intrinsic values (proxied by stock prices). Similarly, Aboody and Lev's [1998] finding that reported capitalized values of software development costs are positively and significantly associated with market values, after controlling for reported book values and earnings, is consistent with the notion that capitalized software costs improve the alignment of book values with intrinsic values (proxied by market value). Furthermore, Aboody and Lev's finding that the annual values of capitalized software costs are associated with subsequent changes in earnings suggests that such capitalizations secprovide relevant information for the prediction of earnings-the ond desired element of a standard according to the residual earnings model.28 Amir and Lev's [1996] study of cellular companies, indicating that investors implicitly capitalize customer acquisition costs, also indicates that the capitalization of these costs will improve the alignment of book values with intrinsic values.29 Internationally, Abrahams and Sidhu [1998] finding that capitalized R&D values on Australian companies' balance sheets are significantly associated with market values, and Barth and Clinch [1998] finding that revaluations of intangibles by Australian companies are associated with market values, are also consistent with the claim that the capitalization of intangibles will improve the alignment of book values with intrinsic values. Finally, a simulation-based analysis (Healy, Myers, and Howe [1998]) demonstrates the general superiority of
28 The predictive ability of annual software capitalization with respect to future earnings is expected, since the capitalization of software development costs is predicated on the success of the development program (e.g., passing feasibility tests or developing a working pilot). Such developmental success should be associated with increases in subsequent sales and earnings. 29The implied capitalization of customer acquisition costs is indicated by a regression of stock returns on earnings before general expenses (which mainly include customer acquisition costs) and general expenses, which finds the latter variable to have a positive, large (relative to earnings), and statistically significant coefficient.

BOUNDARIES

OF FINANCIAL

REPORTING

379

intangibles' capitalization (successful efforts) over immediate expensing in providing meaningful performance data to investors. Furthermore, the analysis shows that even when the majority of firms manage earnings-a major concern with capitalization-intangibles' write-downs continue to be significantly related to economic returns. As with tangible assets, the amortization of the capitalized intangibles will be based on management's estimates of productive lives, guided by industry norms and research findings.30 The amortization rates may be revised as the actual benefits of intangibles materialize. A strict periodic impairment test along the lines of SFASNo. 121 (FASB[1995]) should be applied as a safeguard against overvaluation. How will the reporting deficiencies discussed in previous sections be alleviated by the proposed capitalization of intangibles? First, such capitalization will improve the periodic matching of costs and benefits, particularly for firms with high growth rates of intangible investment. This will lead to reported earnings which more meaningfully reflect enterprise performance than currently measured earnings. Second, the capitalized intangibles will be reported on corporate balance sheets, placing intangible assets on a common footing with tangible assets. The amortization and write-offs of these assets will convey valuable information about managers' assessment of the expected benefits of intangibles.31 Third, the capitalization of intangibles is a crucial step in providing a basis for evaluating the success of innovative activities. Data on the investment in intangibles classified by homogeneous product/activity groups and coupled with a breakdown of revenues and gross margins attributable to the intangibles will enable investors to assess the return on investment in research, product development, and brand enhancement. This important objective of capitalization-allowing for the assessment of return on investment-is often overlooked by those who object to capitalization on the grounds that most intangibles are not traded, or because the cost of intangibles often differs from current values (e.g., Scheutze [1993]). On the downside, capitalization of intangibles obviously increases the possibilities of earnings management. However, in contrast to other means of earnings management, such as the early recognition of revenues or an exaggerated restructuring charge, intangibles' capitalization is clearly and separately disclosed in the financial reports, allowing skeptical investors to easily reverse the capitalization. Thus, at best, intangibles' capitalization is a vehicle for managers to share with investors
For example, the U.S. Bureau of Economic Analysis capitalizes aggregate R&D expenditures in a satellite account to the national income and product accounts. This national R&D capital is amortized by 11% per year, roughly the midpoint of the range of amortization rates estimated by economists (Carson, Grimm, and Moylan [1994]). 31 Indeed, Aboody and Lev [1998] find that the amortization of capitalized software is negatively and significantly associated with stock returns. Write-offs were also found value relevant by Healy, Myers, and Howe [1998].
30

380

BARUCH

LEV AND

PAUL ZAROWIN

information about the progress and success of innovation-producing tivities. At worst, capitalization can be reversed, thereby returning financial reports to their current (full expensing) status. 4.2
RESTATED FINANCIAL REPORTS

acthe

Change, we have argued, and the inadequacy of its reflection by the current accounting system are mainly responsible for the deterioration in the usefulness of financial information. However, the consequences of change (e.g., the benefits of a corporate reorganization or of a significant drug development) are generally uncertain, and it is this uncertainty which is often invoked to justify the immediate expensing of changerelated investments. The capitalization of technologically feasible intangible investments proposed above provides a reasonable balance between relevance and reliability of financial information, but this procedure is restricted to products under development. Other change-drivers, such as corporate reorganization or industry deregulation, cannot be accounted for by the proposed capitalization. We accordingly propose a new accounting procedure-the systematic restatement of financial reportsto accommodate those change-drivers and other uncertainties affecting the quality of financial information. Consider, for example, a corporate restructuring where a significant investment is made in efficiency-enhancing and revenue-generating mechanisms, such as extensive employee training, reorganization of divisions and production lines, and acquisition of technology and know-how (e.g., in-process R&D). Current accounting rules require the immediate expensing of such outlays (e.g., restructuring charges), given the uncertainty of their benefits. Such an expensing, however, understates current earnings and book values and overstates subsequent earnings, if the planned efficiencies materialize. We propose that as the expected consequences of the reorganization materialize, both the current and previous financial statements will be restated to reflect the capitalization of the restructuring charges (i.e., reversing their previous expensing) and the amortization of the capitalized amount over the expected duration of of benefits. Such a restatement will correct both the understatement earnings in the restructuring period and the overstatement of earnings in subsequent periods. Consider once more the example of the telecommunications deregulation which, beginning in the late 1980s, opened the local telephone markets to competition. The regional telephone companies belatedly reacted to their loss of monopoly and to the new regulatory system which no longer assured asset values by writing off $26 billion of assets during 1994-95. Thus, throughout the early 1990s the earnings and book values of the telephone companies were overstated (reflecting inflated asset values), while the 1994-95 earnings were understated, given the massive write-offs. By our proposal, upon the write-off in 1995, a telephone company would have restated its 1990-94 financial reports to reflect a gradual write-off of asset values according to the realized impact of competition

BOUNDARIES

OF FINANCIAL

REPORTING

381

and consequent loss of asset values which has been experienced during that period. Thus, the 1995 report would be spared the nonrecurring massive loss, which in most cases is ignored by investors anyway, and the preceding reports would reflect the actual consequences of deregulation. The logic underlying the proposed restatements is that while financial statements purportedly report the consequences of past events, they are crucially dependent on assumptions about future outcomes (e.g., accounts receivable that will default, contingencies that will materialize, or the actual depreciation pattern of assets). As future events evolve and uncertainty is resolved, our understanding of the past is increasingly improved. Shortly after quarter-end, for example, our knowledge about the quarterly results is fuzzy, whereas two years, say, after quarter-end our uncertainty about the quarterly results is greatly reduced.32 Similarly, at launch, the expensing of R&D for a new drug may seem reasonable, yet when the drug receives FDA approval the past expensing is clearly inappropriate. Why then not restate past reports as uncertainty is resolved and one can better measure the past performance of the firm? We believe the systematic restatement of past reports is essential, given the contextual role of financial information (Finger, Lev, and Rose [1996]). Financial reports not only convey new information to investors, they also provide a rich history or context for interpreting current information and events. In fact, evidence presented in this and other studies indicates a deterioration in the amount of new information (timeliness) conveyed by key financial statement items, leaving the contextual function of financial reports to play an increasing role in investors' decisions. Though not extensively researched, some evidence indicates the importance of the contextual role of accounting. For example, Barth, Elliott, and Finn [1999] report that investors' reaction to an earnings surprise is conditioned on the sequence and signs of past surprises. Thus, an earnings increase following past increases is associated with a larger stock price change than an earnings increase following earnings decreases. Similarly, Lev, Radhakrishnan, and Seethamraju [1998] report that investors' reaction to an FDA drug approval depends, among other things, on the past operating success of the developing company. The evidence presented above that the returns-earnings R2s of firms with the full 20-year history are substantially larger than the R2s of firms with shorter historical records (tables 1 and 2) is also consistent with a contextual role of financial information. Finally, the findings of Petroni, Ryan, and Wahlen [1997] that revisions of reserve estimates of insurance companies extending as far back as ten years are significantly associated with investors' reaction to current information is consistent with both the contextual role of financial data and the value relevance of restatements.

32 Ijiri [1989, chap. 7] elaborates on this important idea, that understanding the past requires improved information about the future, and quotes the British mathematician Raymond Smullyan: "To know the past, one must first know the future."

382

BARUCH

LEV AND

PAUL ZAROWIN

If historical financial data affect the interpretation of new signals, then a continuous improvement of such history, in the form of a better matching of revenues with costs achieved by the proposed restatement of past reports, should improve investors' decisions. The restated data, reflecting the continuous resolution of uncertainty, will portray more realistic patterns of earnings, growth, and profitability (e.g., ROE) than the originally reported data. A natural reaction to the proposed restatement is that the revised information will no longer be relevant to decision makers and might even confuse them by presenting several different earnings numbers pertaining to a given accounting period. We doubt the validity of these concerns. First, a restatement of past reports is currently a GAAPrequirement in the case of acquisitions accounted for by the pooling method, without documented harm to investors. More importantly, and closer to our proposal, key macroeconomic variables, such as the gross domestic product (GDP) and the industrial production index, are routinely revised over several years after the initial estimates are released.33 For example, the first estimate of quarterly GDP is released toward the end of the month following the quarter. This "advance estimate" of GDPis based on judgmental assumptions and incomplete data on GDP components. The advance estimate of GDPis revised by the "preliminary GDPestimate" which is released toward the end of the second month following the quarter. The "final GDPestimate" for the quarter is released toward the end of the third month and is based on replacement of preliminary with comprehensive data and on changes in definitions and estimations. Following the final GDPestimate, the Bureau of Economic Analysis (BEA) schedules annual revisions of GDP,usually released in July, to improve the accuracy of GDPestimates as new information is obtained. The BEAalso revises the entire history of the quarterly data series every five years in what is known as a benchmark revision. Note the close resemblance of early estimates of macroeconomic data based on incomplete information to quarterly financial reports, which are similarly based on estimates and incomplete data. Regarding the concern with the usefulness of revised data and the confusion they can cause, a survey of the economic literature dealing with revisions of macroeconomic data (e.g., Mankiw and Shapiro [1986] and Diebold and Rudebusch [1991]) does not support such concerns. The evidence indicates that revisions are useful in conveying new information which cannot be forecasted by the early estimates.
33For example, on July 31, 1998, the Bureau of Economic Analysis (BEA) of the U.S. Department of Commerce announced that the seasonally adjusted estimate of real Gross Domestic Product (GDP) growth for the second quarter of 1998 was an annualized 1.4%. In addition, the July press release contained revised estimates for the real GDPseries (and components) from the first quarter of 1995 until the first quarter of 1998. The revision showed an increase in the estimated average year-over-year real GDPgrowth from 2.9% to 3.3% for the period from 1995 to 1997.

BOUNDARIES

OF FINANCIAL

REPORTING

383

5. Conclusions
We have documented in this study a systematic decline in the usefulness of financial information to investors over the past 20 years, as manifested by a weakening association between capital market values and key financial variables-earnings, cash flows, and book values. We increashave identified a major reason for the usefulness decline-the ing rate and impact of business change and the inadequate accounting linked change empiritreatment of change and its consequences-and cally to loss of informativeness of financial data. Of the various changedrivers, we have focused on intangible investments, thereby completing the linkage: intangibles-business change-loss of value relevance of financial information. The social consequences of the decline in the usefulness of financial information are yet to be fully examined. If investors were able to obtain the information increasingly missing from financial reports from alternative sources, at no added cost, then the social consequences of a decline in accounting usefulness may not be serious, except for accountants. Preliminary evidence, however, is inconsistent with a smooth, costless information substitution. Thus, for example, Barth, Kasznik, and McNichols [1998] report that analysts expend more resources in the analysis of intangible-intensive companies. Aboody and Lev [1998] find that gains to officers from insider trading in R&D-intensive companies are substantially larger than insider gains in firms without R&D. Insider gains come, of course, at the expense of other investors and are probably related to the poor information on R&D available in financial statements. And Boone and Raman [1997] report that unexpected changes in R&D are associated with an increase in the size of bid-ask spreads and price volatility. Wider bid-ask spreads, a reaction to an increase in information asymmetry, imply larger transaction costs to investors and, in turn, an increased cost of capital to the firm. These findings suggest that the reporting inadequacies documented above may adversely affect investors' and firms' welfare. Given these concerns with reporting deficiencies, we have advanced two proposals that may enhance the usefulness of financial information-an extended capitalization of intangible investments and a systematic restatement of past financial reports.
REFERENCES D., AND B. LEV. "The Value Relevance of Intangibles: The Case of Software Capitalization." Journal of AccountingResearch36 (Supplement 1998): 161-91. ABRAHAMS, T., AND B. SIDHU. "The Role of R&D Capitalisations in Firm Valuation and Performance Measurement." AustralianJournal of Management(1998): 169-83. AMIR, E., AND B. LEV. "Value-Relevance of Nonfinancial Information: The Wireless Communications Industry."Journal of Accounting and Economics22 (1996): 3-30. BARTH, M., AND G. CLINCH. "Revalued Financial, Tangible, and Intangible Assets: Associations with Share Prices and Non-Market-Based Value Estimates." Journal of Accounting Research36 (Supplement 1998): 199-233.
ABOODY,

384
BARTH,

BARUCH

LEV AND

PAUL ZAROWIN

M.; J. ELLIOTT; AND M. FINN. "Market Rewards Associated with Patterns of Increasing Earnings."Journal of AccountingResearch37 (Autumn 1999): 387- 413. BARTH, M.; R. KASZNIK; and M. McNICHOLS. "Analyst Coverage and Intangible Assets." Working paper, Stanford University, 1998. BOONE, J. P., AND K. K. RAMAN. "Unrecognized R&D Assets and the Market Microstructure." Working paper, University of Texas, 1997. BOWEN, R.; D. BURGSTAHLER; AND L. DALEY. "The Incremental Information Content of Accrual versus Cash Flows." The AccountingReview62 (1987): 723- 47. BROWN,S.; K. Lo; AND T. Lys. "Use of R2 in Accounting Research: Measuring Changes in Value Relevance Over the Last Four Decades." Working paper, Northwestern University, 1998. CARSON, C.; B. GRIMM; AND C. MOYLAN. "A Satellite Account for Research and Development." Surveyof CurrentBusiness (November 1994): 37-71. CHANG, J. "The Decline in Value Relevance of Earnings and Book Values." Working paper, Harvard University, 1998. CHENG, C.; W. HOPWOOD; AND J. McKEOWN. "Nonlinearity and Specification Problems in Unexpected Earnings Response Regression Model." The Accounting Review 67 (1992): 579-98. COLLINS, D.; E. MAYDEW; AND I. WEISS. "Changes in the Value-Relevance of Earnings and Book Values Over the Past Forty Years."Journal of Accounting and Economics(December 1997): 39-67. DELOITrE & TOUCHE. Survey American of BusinessLeaders. New York:Deloitte & Touche, 1995. DENG, Z., AND B. LEV."The Valuation of Acquired R&D-in-Process." Working paper, New York University, 1998. DIEBOLD, E, AND G. RUDEBUSCH. "Forecasting Output with the Composite Leasing Index: A Real-Time Analysis."Journal of the American StatisticalAssociation86 (1991): 603-10. DIETRICH, R.; R. FREEMAN;T. HARRIS; K. PALEPU; D. LARCKER; S. PENMAN; AND K. SCHIPPER. "EvaluatingFinancial Reporting Standards."Working paper, University of Chicago, 1997. ELY, K., AND G. WAYMIRE. "Accounting Standard Setting Organizations and Earnings Relevance: Longitudinal Evidence from NYSE Common Stocks, 1927-93." Journal of AccountingResearch (Autumn 1999): 293-317. 37 FINANCIAL ACCOUNTING No. STANDARDS BOARD. Statementof Financial Accounting Concepts 1: Objectives Financial Reportingby Business Enterprises. of Stamford, Conn.: FASB, 1978. . Statementof Financial Accounting Standards No. 50: Financial Reportingin the Record and Music Industry. Stamford, Conn.: FASB, 1981 a. . Statementof Financial Accounting Standards No. 53: Financial Reporting by Producers and Distributorsof Motion PictureFilms. Stamford, Conn.: FASB, 1981 b. . Statementof Financial Accounting StandardsNo. 86: Accountingfor the Cost of Computer Marketed.Stamford, Conn.: FASB, 1985a. Softwareto Be Sold, Leased or Otherwise _____. Statement of Financial Accounting ConceptsNo. 6: Elements of Financial Statements. Stamford, Conn.: FASB, 1985b. . Statementof Financial Accounting Standards No. 121: Accountingfor the Impairmentof Long-LivedAssets and for Long-LivedAssets to Be Disposed Of Stamford, Conn.: FASB, 1995. FINGER, C.; B. LEV; AND A. ROSE. "The Contextual Role of Financial Reports." Working paper, New York University, 1996. FRANCIS, J., AND K. SCHIPPER. "Have Financial Statements Lost Their Relevance?" Journal of AccountingResearch37 (Autumn 1999): 319-52. FRANCIS, J.; D. HANNA; AND L. VINCENT. "Causes and Effects of Discretionary Asset WriteOffs."Journal of Accounting Research34 (Supplement 1996): 117-34. HAYN, C. "The Information Content of Losses." Journal of Accounting and Economics 20 (1995): 125-54. HEALY, P.; S. MYERS; AND S. HowE. "R&D Accounting and the Relevance-Objectivity Tradeoff: A Simulation Using Data from the Pharmaceutical Industry." Working paper, Harvard University, 1998.

BOUNDARIES
HELFAT,

OF FINANCIAL

REPORTING

385

C. "Firm Specificity in Corporate Applied R&D." Organization Science 5 (1994): 173-84. Bookkeeping: Exploring the Dynamic Structureof IjiRi, Y MomentumAccounting and Triple-Entry American Accounting Association, Studies in Accounting ReAccounting Measurements. search, no. 31. Sarasota, Fla.: American Accounting Assn., 1989. INTERNATIONAL ACCOUNTING STANDARDS COMMITTEE. International Accounting Standard No. 38: IntangibleAssets.London: IASC, 1998. JOVANOVIC, AND Y. NYARKo."Research and Productivity." Working paper, University of B., Pennsylvania, 1995. KEMENY, J., AND L. SNELL. Finite Markov Chains. New York: Van Nostrand, 1967. LANG, M. "Time-Varying Stock Price Response to Earnings Induced by Uncertainty about the Time-Series Process of Earnings." Journal of Accounting Research29 (Autumn 1991): 229-57. LEv, B. "On the Usefulness of Earnings and Earnings Research: Lessons and Directions from Two Decades of Empirical Research."Journal of AccountingResearch27 (Supplement 1989): 153-92. LEv, B., AND T. SOUGIANNIS. "The Capitalization, Amortization, and Value-Relevance of R&D." Journal of Accounting and Economics21 (1996): 107-38. LEv, B., AND S. R. THIAGARAJAN. "Fundamental Information Analysis."Journal of Accounting Research31 (Autumn 1993): 190-215. LEv, B.; S. RADHAKRISHNAN; AND C. SEETHAMRAJU. "FDADrug Approvals and the Formation of Investors' Beliefs." Working paper, New York University, 1998. LEv, B.; B. SARATH; AND T SOUGIANNIS. "Reporting Biases Caused by R&D Expensing and Their Consequences." Working paper, New York University, 1999. Liu, J.,AND J.THOMAS. "Stock Returns and Accounting Earnings." Working paper, Columbia University, 1998. LIvNAT, J.,AND P. ZAROWIN. "The Incremental Information Content of Cash-Flow Components." Journal of Accounting and Economics13 (1990): 25-46. MANKIW, AND M. SHAPIRO. "News or Noise: An Analysis of GNP Revisions." Survey of G., CurrentBusiness (May 1986): 20-25. OHLSON, J. "Earnings, Book Values and Dividends in Security Valuation." Contemporary AccountingResearch11 (1995): 661-87. OU, J., AND S. PENMAN. "Financial Statement Analysis and the Prediction of Stock Returns."Journal of Accounting and Economics(November 1989): 295-329. PETRONI, K.; S. RYAN; AND J. WAHLEN. "The Risks and Value Relevance of Revisions of Accrual Estimates: Evidence from Property-Casualty Insurers' Loss Reserves Development Disclosure." Working paper, Michigan State University, 1997. RAMESH, K., AND R. THIAGARAJAN. Inter-Temporal Decline in Earnings Response Coefficients." Working paper, Northwestern University, 1995. SCHEUTZE, W. "What Is an Asset?" Accounting Horizons 7 (1993): 66-70. STEWART, T. Intellectual Capital. New York: Doubleday, 1997. STIGLER, G. The Theoryof Price.New York: MacMillan, 1966.