The relative importance of earnings and other information in the valuation of rd intensive firms

... examine the relative roles o f earnings and non -earnings information in valuing R&D intensive firms In this model, the trade off between book value and expected future abnormal earnings makes the. .. other than the past history of earnings into their forecasts of future earnings If there were an increase in the importance o f other information to firm valuation then the coefficients in a regression... permission of the copyright owner Further reproduction prohibited without permission I i THE RELATIVE IMPORTANCE OF EARNINGS AND OTHER INFORMATION IN THE VALUATION OF R&D INTENSIVE FIRMS ! !

INFORMATION TO USERS

This manuscript has been reproduced from the microfilm master UMI films the text directly from the original or copy submitted Thus, some thesis and dissertation copies are in typewriter face, while others may be from any type of computer printer

The quality of this reproduction is dependent upon the quality of the copy submitted Broken or indistinct print, colored or poor quality illustrations and photographs, print bleedthrough, substandard margins, and improper alignment can adversely affect reproduction

In the unlikely event that the author did not send UMI a complete manuscript and there are missing pages, these will be noted Also, if unauthorized copyright material had to be removed, a note will indicate the deletion

Oversize materials (e.g., maps, drawings, charts) are reproduced by sectioning the original, beginning at the upper left-hand comer and continuing from left to right in equal sections with small overlaps

Photographs included in the original manuscript have been reproduced xerographically in this copy Higher quality 6‘ x 9" black and white photographic prints are available for any photographs or illustrations appearing

in this copy for an additional charge Contact UMI directly to order

Bell & Howell Information and Learning

300 North Zeeb Road, Ann Arbor, Ml 48106-1346 USA

800-521-0600

Reproduced with permission of the copyright owner Further reproduction prohibited without permission.

Reproduced with permission of the copyright owner Further reproduction prohibited without permission.

MBA, University o f Colorado at Denver, 1996

MS Finance, University o f Colorado at Denver, 1996

B.S.B.A., Bryant College, 1989

A thesis submitted to the Faculty o f the Graduate School o f the University o f Colorado in partial fulfillment

o f the requirement for the degree o f

Doctor o f Philosophy Department of Accounting

2000

Reproduced with permission of the copyright owner Further reproduction prohibited without permission.

UMI Num ber 9979354

_ _ <S>

UMIUMI Microform9979354 Copyright 2000 by Bell & Howell Information and Learning Company All rights reserved This microform edition is protected against unauthorized copying under Title 17, United States Code.

Bell & Howell Information and Learning Company

300 North Zeeb Road P.O Box 1346 Ann Arbor, Ml 48106-1346

Reproduced with permission of the copyright owner Further reproduction prohibited without permission.

This thesis entitled:

The Relative Importance of Earnings and Other Information

in the Valuation o f R&D Intensive Firms

written by Denise Jones has been approved for the Department o f Accounting

The final copy o f this thesis has been examined by the signatories, and we find that both the content and the form meet acceptable presentation standards o f scholarly work in

the above mentioned discipline

Jones, Denise (Ph.D., Accounting)The Relative Importance o f Earnings and Other Information in the

Valuation o f R&D Intensive FirmsThesis directed by Professor Barry Lewis

Expected future abnormal earnings depend on past earnings and other information not contained in past earnings, such as changes in the competitive environment o f the firm or new product introductions Since accounting earnings are generally more reliable than other information, it is important to understand the role each o f these play in valuing different types o f firms Using the residual income valuation model, I hypothesize that the magnitude o f the valuation coefficient on earnings is lower for R&D intensive firms Empirical results do not support this hypothesis However, I find that the magnitude o f the valuation coefficient on other information is higher for R&D intensive firms

Other information gets incorporated into firm value through its impact on future expectations o f earnings Market participants, such as financial analysts, forecast fiiture earnings using past earnings and other information voluntarily provided by firms I examine the voluntary disclosure o f other information that firms make in their annual reporting to shareholders and the SEC and throughout the year to financial analysts I separate the voluntary disclosures into two types: general

disclosures and R&D related disclosures R&D related disclosures include information on research projects in process and completed projects that have resulted

in a new product General disclosures include all other disclosures such as information on past operating results, competitive environment, employees, product markets, etc I hypothesize and find that an increase in the amount o f R&D related disclosures improves the accuracy o f analysts’ earnings forecasts However, an increase in the amount of general disclosures appears to have no effect on forecast accuracy

iii

Reproduced with permission of the copyright owner Further reproduction prohibited without permission.

1 Introduction

2 Literature R eview

2 1 Evidence on the Market Treatment o f Research and Development E xpenditures

2.2 D isclosure

3 Model and Hypothesis D evelopm ent

3 1 Role o f Earnings and Other Information in Valuation 3.2 Voluntary Disclosure o f Other Inform ation

4 Research Design and Empirical Results for the Relative Roles o f Earnings and Other Inform ation

4.1 Time Series Properties o f Abnormal Earnings and Other Inform ation

4.2 Empirical Version o f EBD M o d e l

4.3 Sample and Variable M easurem ent

4.4 R esults

5 Research Design and Empirical Results for the Effect o f Voluntary D isclosure

5.1 Sample Selection

5.2 Empirical M o d el

5 3 Measurement o f D isclosure

5.4 R esults

6 C onclusion

Bibliography Appendix A Change in Valuation Coefficients with Respect to Differences in Persistence P aram eters

B General Disclosure S c o re

C R&D Disclosure S c o re

D Companies in S am ple

v

1

6 6

12

18 18 25

27 27 33 33 39

42 42 44 46 47 52

54

59 60 64

66

Reproduced with permission of the copyright owner Further reproduction prohibited without permission.

Table

I Calculation o f Persistence Parameters for the Sample Partitioned

into Quintiles Based on R&D Intensity 31

2 Sample Determination and Com position 35

3 Descriptive Statistics 36

4 Regression o f Market Value on Dividends, Earnings, and

Other Information 41

5 Number o f Firms by Industry by R&D Intensity in 1997 43

6 Descriptive Statistics o f Firms with Disclosure D a ta 49

7 Correlation between Analyst Forecast Error, Disclosure Score,

and Control V ariables 50

8 Regression o f Analyst Forecast Error on Disclosure Score and

Control V ariables 51

vi

Reproduced with permission of the copyright owner Further reproduction prohibited without permission.

1 Introduction

This dissertation addresses two related research questions First, do the relative roles o f earnings and other information in firm valuation differ for R&D intensive firms? Second, to what extent do firms’ disclosure policies mitigate difficulties in valuing R&D intensive firms?

Prior research suggests that the market treats R&D expenditures as an asset (Lev and Sougiannis, 1996; Hall, 1993; Bublitz and Ettredge, 1989; Hirschey and Weygandt, 1985) Therefore, expensing R&D investments immediately may lead to omission o f an important asset on the balance sheet The residual income valuation model (Feltham and Ohlson, 1995; Ohlson, 1995; Peasnell, 1982) separates the value

o f the firm into two components, book value plus the present value o f expected future abnormal earnings In this model, failure to record assets as part o f book value will not affect the value o f the firm, as long as future abnormal earnings can be estimated accurately Unfortunately, the high degree o f uncertainty about future economic benefits from R&D activities that prevents capitalization o f R&D assets on the balance sheet (see SFAS No 2, paragraphs 48-50) also makes it difficult to estimate future abnormal earnings for R&D intensive firms Kothari, Laguerre and Leone(1999) provide evidence on this issue, reporting that future earnings variability related

to R&D expenditures is three times the future earnings variability related to capital expenditures Therefore, difficulty in predicting future abnormal earnings could lead

to larger valuation errors for firms whose R&D activities comprise a substantial portion o f the business and which, therefore, have a significant R&D expense

l

Reproduced with permission of the copyright owner Further reproduction prohibited without permission.

Using assumptions about the time-series process o f abnormal earnings, the residual income valuation model can be expressed in terms o f current earnings, book value, dividends, operating assets and other information (Ohlson, 1998; Feltham and Ohlson 1995) In this model, multiples o f the earnings and other information

i

j variables represent expected future abnormal earnings Earnings are capitalized intoi

! firm value to the extent they are expected to persist into the future Other information

I not contained in past earnings, such as new product introductions, is incorporated into

i

1

j firm value through its impact on expected future eamings Following Dechow,Hutton and Sloan (1999) and Ohlson (1998), I use analysts’ expectations about future eamings to proxy for other information Specifically, I define other information as the difference between one-year ahead eamings forecasts and forecasts based on the time-series o f abnormal eamings

As a first step in understanding differences in how abnormal eamings contribute to firm value for R&D intensive and non-R&D intensive firms, I address whether the magnitude o f the valuation coefficients on eamings and other information differs for R&D intensive firms To determine if the magnitudes o f the coefficients

on eamings and other information should differ for R&D intensive firms, I look at the theoretical values o f the coefficients The coefficients depend on the persistence o f abnormal eamings and the persistence o f other information I hypothesize that the persistence o f abnormal earnings decreases with R&D intensity Many smaller R&D intensive firms are in a rapid growth stage and have high variability in their earnings For example, periods o f prosperity after a new product introduction are often

followed by less successful periods, as the new product technology becomes

2

I

Reproduced with permission of the copyright owner Further reproduction prohibited without permission.

outdated Even stable firms face pressures to constantly introduce new products to keep ahead o f changing technology Consistent with my hypothesis, I find that R&D intensive firms have less persistent abnormal earnings Whether the persistence of other information differs for R&D intensive firms is less clear High persistence of other information indicates that the impact o f the new information will persist for

| several periods into the future Other information can take many different forms and

I can impact eamings positively or negatively For example, the cost o f a strike shouldI

impact eamings negatively and should have a short term effect only In contrast, a new product introduction should positively impact eamings and could affect eamings for several periods into the future I find no systematic differences in the persistence

j of other information for R&D intensive and non-R&D intensive firms

I

I

The theoretical value o f the coefficient on eamings is positively related to the persistence o f abnormal earnings and inversely related to the persistence o f other information Given the result that the persistence o f abnormal eamings is lower for

j R&D intensive firms, I hypothesize that the eamings o f R&D intensive firms are

j valued less by the market than the eamings o f firms with no R&D activities Testresults do not support this hypothesis However, I do find that the coefficient on other information is significantly higher for R&D intensive firms For R&D intensive firms, the market places a higher weight on other information, presumably because the abnormal eamings are less persistent and alternative information sources are needed to value the firm

Investigating the relative importance o f eamings and other information in firm valuation is important because these two variables combined determine expectations

3

Reproduced with permission of the copyright owner Further reproduction prohibited without permission.

about future abnormal earnings However, eamings and other information can not be used with equal ease in predicting future abnormal eamings Mechanical time-series models can be developed to predict how current eamings translate into future

eamings (and hence future abnormal eamings) The effect o f other information on future eamings is less clear One reason for this is that there are many types o f new information, such as new products, changing competitive environment, change in management, etc., and each market participant values the information differently Additionally, firms provide this information to market participants voluntarily, and the specific information provided varies by firm

The first half o f the dissertation establishes the importance o f other information, not contained in the basic financial statements, for the valuation o f R&D intensive firms The second research question I consider is whether the voluntary disclosure o f this other information mitigates difficulties in firm valuation Other information gets incorporated into firm value through its impact on future

expectations o f earnings Market participants, such as financial analysts, forecast future eamings using past eamings and other information voluntarily provided by firms One way o f determining the impact that voluntary disclosures have on market participants' ability to accurately value the firm is to examine the impact of voluntary disclosures on eamings forecast prediction errors

I collect data on the voluntary disclosures made by 144 firms over a one year time period I examine disclosures from two sources: annual reports to shareholders and the SEC and summaries o f financial analysts’ conversations with management during both conference calls and other times I examine two types o f voluntary

4

Reproduced with permission of the copyright owner Further reproduction prohibited without permission.

disclosures: general disclosures and R&D related disclosures R&D related disclosures include information on current research projects and the output from those research projects in the form o f new products General disclosures are all other types

of disclosures including information on the manufacturing and sales o f current products, competitive environment, major contractual relationships, changes in current operating results, and forward looking information After controlling for variables known to affect analyst forecast accuracy, I hypothesize and find that an increase in the voluntary disclosure o f R&D related information results in an improvement in analyst eamings forecast accuracy In addition, I find that general disclosures have no impact on analyst eamings forecast accuracy

The remainder o f the dissertation is organized as follows Section 2 reviews the existing literature on research and development expenditures and voluntary disclosure Section 3 develops the model and hypotheses Section 4 describes the research design and reports the results o f empirical tests o f the roles of eamings and other information in firm valuation Section 5 describes the research design and reports the results of empirical tests o f the impact o f voluntary disclosure on eamings forecast accuracy Section 6 concludes the paper

5

Reproduced with permission of the copyright owner Further reproduction prohibited without permission.

2 Literature Review

This dissertation will contribute to two streams o f literature: research on the valuation o f firms with significant research and development outlays and research on voluntary disclosure The relevant papers from each are discussed below

2.1 Evidence on the M arket Treatment o f Research and Development Expenditures

In 1974 the Financial Accounting Standards Board (FASB) issued Statement

o f Financial Accounting Standards No 2 (FAS 2) requiring the immediate and full expensing o f all research and development outlays Since that time, several studies have documented that, on average, the market treats R&D outlays as an asset Early studies found a positive relationship between R&D expenditures and the market value

o f the firm Hirschey and Weygandt (1985) find that R&D and advertising expenditures are positively related to Tobin’s Q (market value o f the firm divided by the replacement cost o f tangible assets) Hall (1983) examines the relationship between the market value o f the firm and current R&D expenditures and R&D capital R&D capital is estimated by assuming 100% capitalization o f R&D expenditures and a 15% yearly amortization rate The sample is all firms in the National Bureau o f Economic Research database from the period 1976 to 1991 The advantage of this database is that the variables are inflation adjusted Hall finds a positive relationship between the market value o f the firm and both the R&D expenditures and the R&D capital However, the value the market places on R&D expenditures decreased over the period 1986 to 1991

6

Reproduced with permission o f the copyright owner Further reproduction prohibited without permission.

j Extending this prior work, Lev and Sougiannis (1996) use the relationshipbetween current R&D expenditures and future operating income to calculate the average future benefit o f a current R&D outlay They use this to estimate yearly, industry specific R&D amortization rates and to compute the portion o f successful R&D outlays that should have been capitalized They report that, on average, book

value is understated by 22% and eamings are understated by 21% In addition, thesei

! firm specific adjustments to eamings and book values for R&D capitalization are

]

positively related to stock prices Similarly, Deng and Lev (1998) study the relationship between stock prices and the fair market value o f R&D projects in process that is estimated and disclosed when a firm is acquired Their sample consists o f 375 cases where technology in process was part o f an acquisition during the period 1985 to 1996 As expected they report a positive association between stock prices and the disclosed value o f R&D in process

An alternative market based technique is to look at stock returns Bublitz and Ettredge (1989) examine the relationship between annual unexpected market returns and annual changes in eamings broken down into changes in sales, advertising expenses, R&D expenses and other expenses They hypothesize that if advertising and R&D expenditures benefit future periods then there is a correlated omitted variable that will bias the coefficient on advertising and R&D toward zero As predicted they find that the coefficient on R&D expense is not different from zero

In one o f the few studies examining variation in the stock market’s valuation

o f R&D expenditures, Lev and Zarowin (1998) use a time series approach to calculate

a firm specific price response to research and development expenditures They report

Reproduced with permission of the copyright owner Further reproduction prohibited without permission.

that this R&D response coefficient is positively related to the expected future payoff from the R&D investment (estimated from a regression o f current eamings on lagged R&D expenditures) and positively related to the risk o f the firm’s overall R&D activities (payoff period of R&D benefits and volatility o f R&D outlays).

Studies have also shown that the market uses information besides the direct R&D outlay in determining an asset value for innovative activities Using financial statement footnote disclosures on R&D limited partnerships, Shevlin (1991) estimates the value to the firm of the option to exploit technology developed by the limited partnership He finds that the market positively values both in-house R&D and the calculated value o f the limited partnership R&D project In addition to financial statement disclosures, other public information appears to be used by the market Hirschey, Richardson and Scholz (1998) examine patent data for 403 high technology companies over the period 1989 to 1995 They regress the market value o f the firm

on book value, eamings, R&D expenditures, and four patent measures: number of patents, patent impact (ratio o f patent citations to the average number o f citations), link to other research (references cited on patent application), and technology cycle time (time elapsed from the previous generation o f patents in the same area) They report that, consistent with prior studies, R&D expenditures are positively valued by the market In addition, the number o f patents, patent impact and link to other research are also positively valued by the market Similarly, Hall, Jaffe and Trajtenberg (1998) find a positive relationship between citation weighted patent counts and Tobin’s-Q (market value o f the firm divided by the replacement cost of

8

Reproduced with permission of the copyright owner Further reproduction prohibited without permission.

tangible assets) for a sample o f over 4,800 US manufacturing firms over the past 30years.

High growth, technology based industries often have periods o f losses and low eamings due to the immediate expensing o f R&D and other start up costs In

addition, many companies such as internet or communications companies have sustained periods o f losses Several studies have addressed whether traditional financial statement measures such as eamings and book values are used to value these firms or whether the market uses other non-financial measures Amir and Lev (1996) report that, for a small sample o f independent cellular phone companies, neither book values nor eamings explains stock price However, when population in the service area (a measure o f growth potential) and penetration rate (ratio o f subscribers to the population) are added to the regression, eamings becomes positive and significant and both population and penetration ratio are positively related to market value Recent studies have examined how internet stocks are valued Trueman, Wong, and Zhang(2000) examine a sample o f 56 internet companies They find gross profit to be positively associated with stock prices but not net income In addition, they report that two measures o f internet usage, number o f unique visitors and number o f pages viewed, are positively related to stock price In a similar study o f 86 internet firms, Rajgopal, Kotha and Venkatachalam (2000) report a positive relationship between market value and the average monthly unique visitors as a percentage o f the total web population

Although R&D outlays lead to future economic benefits and have been shown

to be relevant to the market, there still could exist a problem if management had to

9

Reproduced with permission of the copyright owner Further reproduction prohibited without permission.

estimate an amount o f R&D capital The management o f a firm has incentives to manage the firms eamings for reasons such as to increase compensation, make the firm look more favorable before stock offerings, meet regulatory requirements, etc (see, e.g., Teoh, Wong and Rao, 1997; Jones, 1991; Healy, 1985) If R&D

capitalization were mandatory, then it would create an opportunity to use the capitalization process as an eamings management tool Ely and Waymire (1999) investigate the value relevance o f intangible assets capitalized by firms in 1927, a period when management had complete discretion in the capitalization, amortization, and disclosure policies o f the firm Unlike the studies previously discussed, covering the most recent time period, they find no relationship between capitalized intangibles and share prices In addition, the relationship between eamings and share prices decreases with the level o f capitalized intangibles, possibly because investors perceive that managers overstate eamings through intangibles capitalization

In addition to the possibility o f eamings management, a high degree o f uncertainty about future benefits from current R&D projects could hamper management’s ability to estimate an amount o f R&D to be capitalized Kothari, Laguerre, and Leone (1999) show that the variability o f future eamings due to R&D investments is greater than the variability o f future eamings due to capital

expenditures and therefore, R&D investments generate more uncertain future benefits

Several studies have focused on the implications o f alternative methods o f accounting for research and development activities Loudder and Behn (1995) examine a sample o f firms during the period that FAS 2 was implemented, 1973-

10

Reproduced with permission of the copyright owner Further reproduction prohibited without permission.

1975 They find that, prior to FAS 2, the response of stock returns to the firm’s eamings was higher for firms that capitalized R&D After the adoption o f FAS 2, there was a decline in the eamings response coefficient for firm’s previously capitalizing R&D but no decline for firms that always expensed R&D The authors conclude that the eamings o f capitalizing firms were more useful than the eamings o f expensing firms Supporting this with a more recent sample, Chambers, Jennings and Thompson (1998) calculate “as i f ’ earnings and book values under a policy o f 100% capitalization o f R&D outlays and amortization over industry specific periods They show that there is a small increase in the ability o f earnings and book values to explain share prices when the capitalization policy is followed In addition, when the components o f eamings and book values are examined, the R&D asset and R&D expense from a policy o f 100% capitalization are reflected in prices similarly to property, plant, and equipment and depreciation.

Healy, Myers and Howe (1997) use simulated data from the pharmaceutical industry to construct financial statements under three methods o f reporting for R&D outlays: cash method (full expense), full cost capitalization, and successful efforts capitalization They find that the successful efforts method o f capitalizing R&D provides more information about economic returns and values When discretion in the w rite down o f the R&D asset is introduced, the relation between economic returns and R&D write-downs declines significantly but overall value relevance is still higher under the successful efforts method

To determine an asset value for R&D capital, market participants must first understand how current R&D outlays translate into future cash flows or eamings for

11

Reproduced with permission of the copyright owner Further reproduction prohibited without permission.

the firm This dissertation will contribute to this literature by examining whether the market extrapolates current eamings into future eamings or whether there is

additional information that the market requires to estimate future eamings This is important because one benefit o f capitalizing R&D expenditures is that it provides information on the amount o f R&D expenditures that management considers spent on successful projects However, this one capitalized amount does not tell the market which particular research programs were successful and the capitalization is subject

to manipulation by management As a first step in understanding how R&D firms are valued it is important to understand whether it is the R&D outlays themselves that inform the market or whether it is alternative information such as new product introductions

2.2 Disclosure

Early theoretical models o f disclosure show that when traders have a rational expectation about management’s motivation to withhold unfavorable information about the firm then a manager is forced to follow a policy o f full disclosure

Otherwise the traders will assume the worst about the firm and discount the value of the firm In contrast to the predictions o f these early models, there is wide variation

in the amount and types of voluntary disclosures made by firms Firms have different incentives for how to choose their overall level o f disclosure as well as what specific items to voluntarily disclose Verrecchia (1983) shows that the existence o f

proprietary information extends the range o f possible explanations about why management is withholding information and therefore the withheld information is not

12

Reproduced with permission o f the copyright owner Further reproduction prohibited without permission.

unambiguously considered bad news A manager will only disclose proprietary information when the increase in the firm value from the information exceeds the proprietary costs o f its disclosure.

Proprietary costs can take many forms For example, knowledge o f certain operating costs or profits from different operating segments could lead to labor union renegotiations Additionally, knowledge o f the initial development stages o f a new technology could allow an existing competitor to begin development and reduce the firm’s advantage o f an early entry o f the product to the market place Information about the profitability o f an industry could also lead to the entry of new competitors into the industry Harris (1998) studies 929 firms reporting information on separate business segments during the period 1987 to 1991 She reports that in less

competitive industries (industries with high market share concentrations and a slow rate o f abnormal profit adjustment) managers are less likely to report operations as a separate industry segment, suggesting a desire to protect abnormal profits and higher market share in these industries

Possible reasons for a firm to make voluntary disclosures are to reduce legal costs by pre-empting a negative stock price response to bad news, reduce information asymmetry between managers and shareholders, reduce the cost of capital, increase analyst following, increase the holdings o f institutional investors, and maximize firm value prior to a security offering For example, Healy, Palepu and Sweeney (1995) examine 90 firms with sustained increases in one measure o f disclosure

informativeness, FAF reports The FAF (now AIMR) makes a yearly assessment o f a firms overall effectiveness in communicating with investors in three categories,

13

Reproduced with permission of the copyright owner Further reproduction prohibited without permission.

annual published information, quarterly and other published information and investor relations Healy, Palepu and Sweeney find that managers increase disclosure, as rated

in the FAF reports, when their firm’s earnings growth is undercapitalized by investors Undercapitalization occurs when the response o f stock returns to changes

in eamings over a fifteen month period is lower for a firm than other firms in the same industry

Several studies have addressed the issue o f whether firms are more forthcoming with their disclosures when they require external financing Frankel, McNichols and Wilson (1995) study whether disclosure o f numeric or qualitative eamings forecasts is related to external financing decisions They report that firms that access capital markets for debt or equity financing are more likely to issue management eamings forecasts However, firms are not more likely to issue a management eamings forecast right before an offering Similarly, Clarkson, Kao and Richardson (1994) examine voluntary disclosures made in the MD&A section o f the annual report from 1989 to 1991 o f firms listed on the Toronto Stock Exchange.They report that there is a positive association between the extent of external financing activities in the previous year and the propensity to disclose forward looking information on the expected change in operating results However, this relationship holds for firms with good news to report only Similar to Harris, they also find that the probability o f disclosure is decreasing in the threat o f competitor entry

Lang and Lundholm (1993) examine the relationship between certain firm characteristics and the FAF disclosure rating Similar to the other studies, they find a

14

Reproduced with permission of the copyright owner Further reproduction prohibited without permission.

positive relationship between FAF disclosure rating and the occurrence of a debt or equity offering in the current year or previous two years In addition, they find that larger firms have a higher disclosure score In a similar study, Clarkson, Kao and Richardson (1999) asked the Toronto Society o f Financial Analysts to rate the disclosure quality o f the MD&A section of the 1991 and 1992 annual reports o f 55 firms listed on the Toronto Stock Exchange They report that MD&A disclosure quality is positively related to the amount o f equity issued in the subsequent year Similar to Lang and Lundholm, they also find that larger firms have a higher disclosure rating.

These studies show that specific firms have a number o f reasons to disclose or withhold information about the firms operations This is particularly true o f high technology firms that have incentives to withhold proprietary information from competitors about the success of their R&D programs and also have incentives to disclose information to gain access to capital to fund their programs Because each firm is faced with a different opportunity set and different financing needs, there should be high variation in the amount and types of disclosures each firm makes about its research and development activities

There is little research on the benefits o f voluntary disclosure Primarily due

to the difficulty in measuring disclosure policy and its related benefits and costs It is difficult to find a proxy for a disclosure policy that is both comprehensive and

available for a wide variety o f firms

Botosan (1997) creates an overall disclosure index for disclosures made in the annual report This index is used to rank the disclosures o f 122 firms in the

15

Reproduced with permission of the copyright owner Further reproduction prohibited without permission.

machinery industry Botosan finds that a higher disclosure ranking is associated with

a lower cost o f capital for firms with a low analyst following No evidence o f an association between disclosure ranking and the cost o f capital is found for firms with

a high analyst following Botosan attributes this to the disclosure index being limited

to the annual report and therefore may not be a good proxy o f a firm's overall level of disclosure for firms with a high analyst following Supporting the Botosan study with

a broader sample, Sengupta (1998) finds a negative relationship between the cost of debt and the FAF disclosure ratings o f 103 firms in 15 different industries

Two recent studies have examined the properties o f analysts’ eamings forecasts and a proxy for disclosure quality Barron, Kile and O ’Keefe (1999) study MD&A disclosure quality as measured by the SEC The SEC reviewed the MD&A section o f the annual report o f 650 firms during the period 1987 to 1989 and assigned

a compliance rating to each firm Barron, Kile and O ’Keefe find that firms with a higher SEC compliance rating have lower analyst eamings forecast errors Lang and Lundholm (1996) examine the relations between the FAF overall communication score and properties o f analysts’ eamings forecasts They find that firms with a higher FAF score have more accurate eamings forecasts

There are limitations to using the FAF scores as a measure o f disclosure

First, the disclosure scores are subjective and reflect analyst perception o f a firm’s disclosures only For example, Bamber and Cheon (1998) show that, in the FAF report, analysts rate the communication policies o f firms that issue management forecasts in meetings with analysts more favorably than they rate the communication policies o f firms issuing management forecasts in press releases Second, the firms

16

Reproduced with permission of the copyright owner Further reproduction prohibited without permission.

Reproduced with permission of the copyright owner Further reproduction prohibited without permission.

3 Model and Hypothesis Development

3.1 Role o f Earnings and Other Information in Valuation

Recent studies have documented a decline in the value relevance of earnings over time (Lev and Zarowin, 1998; Collins, Maydew and Weiss, 1997; Francis and Schipper, 1996) Francis and Schipper and Collins, Maydew and Weiss document a shift over the past 40-43 years in value relevance from earnings to book values and Collins, Maydew and Weiss attribute part o f this shift to a growing proportion o f firms in intangible industries However, Francis and Schipper find no difference between high and low technology firms Brown, Lo and Lys (1999) control for scale effects and find that the combined value relevance o f both earnings and book values has declined over time

Lev and Zarowin (1998) find that the value relevance o f both earnings and book values has decreased over the last twenty years and present evidence that this relates to an increase in the rate o f business change in the past 20 years, which is linked to R&D intensity They also report that an increase in R&D intensity for specific firms is related to a decrease in earnings informativeness, evidenced by a decrease in both the response o f stock prices to earnings (ERC) and the R2 in a regression o f stock returns on earnings and change in earnings Similarly, Liu and Thomas (1999) find that high technology firms, defined as firms in the computer, semi-conductor and bio-technology industries, have a lower ERC than other firms when the previous periods earnings are used to measure expected earnings However, when analysts’ forecasts are used to measure expected earnings, the ERC is higher for the high technology firms One explanation for the ERC increase when analyst

18

Reproduced with permission of the copyright owner Further reproduction prohibited without permission.

forecasts are used is that analysts incorporate information other than the past history

o f earnings into their forecasts o f future earnings If there were an increase in the importance o f other information to firm valuation then the coefficients in a regression

of either stock prices or returns on accounting data could change

Ohlson (1995) and Feltham and Ohlson (1995) propose a model where price can be expressed as the sum o f book value and the present value o f future abnormal

where: MVt = market value at time t

bt = book value at time t

r = discount factor

xt = earnings at time t

x, - r*bt-i = xat = abnormal earnings at time t

One problem with implementing the model for empirical work is that earnings and book value must be estimated for an infinite number o f periods into the future Feltham and Ohlson (1995) simplify the model by imposing a time-series stochastic process on abnormal earnings and assuming that future abnormal earnings are a function o f persistence in abnormal operating earnings, growth in operating earnings, conservatism in the accounting for operating assets, and other information Abnormal earnings, operating assets and other information are assumed to follow a simple AR(1) process:

where: xt = operating earnings

bt = book value

r = discount factor oat = operating assets

vu = other information about future abnormal earnings not contained in

current abnormal earnings v2t = other information about future operating assets not contained in current

operating assets

Bn, S2t, 631 = unpredictable, mean zero disturbance terms

co 11 = persistence o f abnormal earnings, 0 < con < 1

o 12 = correction for conservative accounting for operating assets, coi2 ^ 0

CO22 = growth in operating assets, 1 < o22 < (1 + 0

7 = persistence o f other information, j y j < 1

The expectation o f future abnormal earnings for firm i in period t can be expressed as a function of the current abnormal earnings expected to persist into the future, a correction for the conservatism in the accounting for operating assets, and alternative information about the future prospects o f the firm not contained in current abnormal earnings (such as long-term sales contracts, new product introductions, other growth opportunities, union negotiations, etc.) The expectation o f future operating assets can be expressed as a function o f current operating assets, which are expected to grow, and other information that would cause changes in future operating assets such as a restructuring or acquisition o f a new line o f business

The expectation o f future other information can be expressed as the portion of the current other information that is expected to persist into the future For example,

if a new product is introduced then the future revenue from that product would increase future earnings The other information is the increase to future earnings related to the new product If the new product is a computer that is expected to generate revenue for four years then the other information related to the new computer would persist for four years into the future and this would be reflected in a

20

Reproduced with permission of the copyright owner Further reproduction prohibited without permission.

high y If the other information is related to a strike that caused a temporary increase

in salary expense then this information would not be expected to persist into the future and it would be reflected in a low y

Applying the stochastic process followed by abnormal earnings to the valuation model results in a linear information model (see Feltham and Ohlson 1995):

estimated in the model Ohlson (1998) and Dechow, Hutton and Sloan (1999) propose that the other information can be inferred through its impact on expectations There is no empirical data on the expectation of future operating assets Therefore no empirical proxy is developed for other information about future operating assets (v*) However, financial analysts forecast future earnings and this can be used to develop a proxy for the other information about future abnormal earnings ( V [ t) Analysts use current earnings as well as other information to make forecasts o f future earnings Other information is the portion o f the forecast of future earnings not related to current earnings and can be defined as:

21

Reproduced with permission of the copyright owner Further reproduction prohibited without permission.

where: ft-i = one year ahead earnings forecast

r, b, x, cdu are as defined above

Combining (3) and (4) gives:

Given clean surplus accounting:

bt = bt.[ + xt - dt bt-! = bt - xt + dt (6)

i

Simplifying and combining terms allows the value o f the firm to be expressed

in terms o f earnings, book values, dividends, other information and operating assets (see Ohlson 1998):

The coefficients on earnings, book value, dividends, and other information are all a function of the persistence o f abnormal earnings (©i i), the persistence o f other information (y), and the discount factor (r) The coefficient on operating assets is a function of the correction for conservative accounting (0)12), growth in operating assets (0)22), and the persistence o f abnormal earnings (©u) Differences in the persistence parameters for R&D intensive firms leads to differences in the coefficients o f R&D intensive firms and non-R&D firms The nature o f intangible

(8) _ (1 - y ) ( \ - c o u ) ( \ + r )

intensive firms may lead to a less persistent earnings stream than the earnings o f firms with little R&D activity One reason is that these earnings are likely to be more variable than a pure manufacturing company, particularly for start-up companies or

i

j companies in a rapid growth stage Another reason is that product life cycles have

| shortened and products become obsolete more quickly than in the past (see Wallman,

i

1995) Because o f this, even if a firm that relies heavily on R&D is in a stable growth

j stage, its current earnings might not reflect future firm performance as well as otheri

j information about the progress o f R&D programs This is reflected in the lowerj

response o f stock returns to earnings o f R&D firms documented by Liu and Thomas (1999) This leads to hypothesis one (stated in alternative form):

H I: The persistence o f abnormal earnings is lower for R&D intensive firms

Whether the persistence o f other information differs for R&D intensive firms

is less clear The persistence parameter (y) will be higher when the impact o f the new information persists into multiple periods For example, information about a

permanent restructuring or a new product introduction would be expected to persist for several years into the future Other information could have a positive or negative impact on earnings and the number o f periods the impact is expected to last could vary from one period for a temporary employee strike to infinite periods for a major restructuring Due to the nature o f R&D intensive firms, the persistence o f other information should be at least the same as firms with no R&D activities, if not higher

Because the coefficients o f the EBD valuation model variables are all functions o f the persistence parameters and the discount factor, hypotheses about the valuation coefficients follow directly from hypotheses about the persistence

23

I

Reproduced with permission of the copyright owner Further reproduction prohibited without permission.

parameters In Appendix A, I show the first derivatives of the coefficients on book value, earnings, dividends, operating assets, and other information with respect to the persistence parameters ©u and y The first derivative o f the coefficient on earnings (P2) with respect to ©n and y shows that the impact of earnings on firm value increases with the persistence o f abnormal earnings and decreases with the persistence o f other information Given that the persistence o f other information is the same or higher for R&D intensive firms and the persistence o f abnormal earnings

is lower, then the coefficient on earnings o f high R&D intensive firms will be lower This leads to hypothesis two (stated in alternative form):

H2: The earnings o f R&D intensive firms are valued less by the market than the earnings o f non-R&D intensive firms

The coefficient on book value (Pi) is inversely related to both ©i 1 and y The coefficients on dividends (p3) and other information (P5) are positively related to both

©ii and y It is unclear how the coefficients o f these variables will differ for R&D intensive firms if the persistence o f abnormal earnings decreases and at the same time the persistence o f other information increases Therefore, no predictions are made about the magnitude o f these coefficients

The coefficient on operating assets (p4) is positively related to the persistence

of abnormal earnings (©11) This would indicate that a lower persistence o f abnormal earnings for R&D intensive firms would lead to a decrease in the coefficient on operating assets However, the most important parameter for this coefficient is the correction for conservative accounting (©12)- Since R&D outlays are immediately expensed, then the correction for conservative accounting should be higher for R&D

24

Reproduced with permission of the copyright owner Further reproduction prohibited without permission.

intensive firms that have a greater portion o f their market value that is not captured on the balance sheet Due to these offsetting influences, no prediction is made for this coefficient.

3.2 I oluntary Disclosure o f Other Information

As the importance o f other information in firm valuation increases, the probability o f errors in firm valuation increases This is because other information is incorporated into firm valuation through its impact on the forecast o f future earnings Other information has two drawbacks relative to the time series o f earnings in

forecasting future earnings o f a firm First, the amount and types o f other information provided to market participants by firms varies widely Firms have different

incentives about what to voluntarily disclose and studies have shown that the amounts and types o f voluntary disclosures they make differ (see, e.g., Frankel, Johnson and Skinner, 1999; Bamberand Cheon, 1998; Harris, 1998; Frankel, McNichols and Wilson, 1995; Clarkson, Kao and Richardson, 1994; Lang and Lundholm, 1993; Verrecchia, 1983) This is particularly true of high technology firms, some o f which have incentives to withhold proprietary information from competitors about the success o f their R&D programs, and some o f which have incentives to disclose information to gain access to capital to fund their programs Second, unlike using a time series model to forecast future earnings, determining when and how other information will impact future earnings is very subjective

Studies have shown that firm’ disclosure practices affect analyst forecast accuracy Lang and Lundholm (1996) examine the relations between the Financial

25

Reproduced with permission of the copyright owner Further reproduction prohibited without permission.

Analyst Federation (FAF, now AIMR) overall communication score and properties o f analysts’ earnings forecasts The FAF makes a yearly assessment of a firms overall effectiveness in communicating with investors in three categories, annual published information, quarterly and other published information, and investor relations Lang and Lundholm find that firms with a higher FAF score have more accurate earnings forecasts Similarly, Barron, Kile and O ’Keefe (1999) find that high MD&A quality,

as measured by the SEC, is associated with less error in analysts’ earnings forecasts Given these studies, it is likely that a firm’s policy on making R&D related

disclosures affects the magnitude o f the analyst forecast error This leads to hypothesis three (stated in alternative form):

H3: The absolute value o f the analyst forecast error is negatively related to the level o f R&D related disclosures

26

Reproduced with permission of the copyright owner Further reproduction prohibited without permission.

4 Research Design and Empirical Results for the Relative Roles of Earnings

and Other Information

4.1 Time-Series Properties o f Abnormal Earnings and Other Information

Hypothesis one predicts that the persistence o f abnormal earnings is lower forR&D intensive firms Following Dechow, Hutton and Sloan (1999), I estimate thepersistence parameters using a pooled time-series and cross-sectional regression o fthe following models:

Other information: V,j-l =Yo + Y\V,.: (B)

Abnormal earnings for firm i in period t is defined as x tat = x tJ - rbt t and

other information is defined as vt J = ( f , , - rbt [) — , ( r , , - rbu _t ) Earnings (x)

and book value (b) are obtained from Compustat Book value (bt) is the book value at the end o f fiscal year t Earnings (xt) are earnings from continuing operations before special items, after tax, as designated by Compustat for year t Earnings are defined

as earnings before special items to best match the earnings that analysts forecast (Philbrick and Ricks, 1991)

Following Myers (1999) and Frankel and Lee (1998) the discount rate (r) is allowed to vary over time and by industry Firms are assigned to each industry based

on SIC code and are assigned the industry cost o f equity each year The industry cost

o f equity is calculated as:

rc(j4) = rt(t) + rpremG)where: re(j,t) is the estimated cost o f equity for industry j in year t

n(t) is the yearly t-bill return in year ttpremCj) is the risk premium for industry j as estimated by Fama and

French(1997)

27

Reproduced with permission of the copyright owner Further reproduction prohibited without permission.

The expectation of future earnings (f) is the analyst forecast o f earnings for period t+1 made at time t In order to make sure that analysts had access to the same microeconomic and industry data at the time that the forecast was made for all firms, the sample will be restricted to December 31 year-end firms The analysts’ forecasts are obtained from the I/B/E/S database The analyst forecast for each firm in year t is the mean of all forecasts made during January and February o f year t+1 for the year ending December t+1 This forecast is used for several reasons The mean is used because when only recent forecasts are used then the mean or median reduces idiosyncratic individual error and improves accuracy (O’Brien, 1988) Forecasts made during January and February are used because most firms announce year t earnings by the end of February and therefore forecasts during this period will be updated for year t earnings and should be recently revised as there is more revision activity after interim earnings announcements than before (Stickel, 1989; Brown, Foster and Noreen, 1985) Extending the forecast period beyond February gives the analyst a timing advantage as firms begin to preannounce the first quarter of year t+1

earnings

I first estimate the abnormal earnings persistence parameter (con) using the simple time series o f abnormal earnings model (A) I estimate this model using a pooled time-series, cross-section using all data available from 1980 to 1998 Because the persistence parameter is used in the calculation o f the other information variable, the sample o f firms is limited to firms that have the required Compustat, CRSP and I/B/E/S data Because the model is estimated on a pooled time-series, cross-section

of firms, each firm is not required to have data available for the entire sample period,

28

Reproduced with permission of the copyright owner Further reproduction prohibited without permission.

therefore there is no survivorship bias To control for heteroscedasticity, abnormal earnings are scaled by the market value o f equity Since the persistence parameters are expected to differ for firms with varying degrees o f R&D intensity, the sample is split into quintiles each year based on R&D intensity (R&D expenditures divided by sales) and the persistence parameter is estimated separately for each quintile The first quintile contains all firms with zero R&D expenditures.

I use the estimate of the abnormal earnings persistence parameter (oi i) to calculate the other information variable (v) I then estimate the simple time series of other information model (B) This model is estimated using a pooled time-series, cross section over the period 1984 to 1997, the period when the analyst forecast data are available To control for heteroscedasticity, the variables are scaled by market value o f equity

Table I, panel A shows that the abnormal earnings persistence parameter (con) is higher for firms with no R&D activities, supporting hypothesis one It is 0.70 for firms in the first quintile, mainly firms with no R&D expenditures, and 0.53 and

0 53 for firms in the fourth and fifth quintiles, high R&D intensive firms Table 1, panel B shows that the other information persistence parameter (y) does not vary in any predictable manner by R&D intensity It is 0.26 for firms in the first quintile, 0.47 for firms in the second quintile, 0.36 for firms in the third quintile, 0.16 for firms

in the fourth quintile, and 0.35 for firms in the fifth quintiles Dechow, Hutton and Sloan (1999) estimated the same models and found similar results For their sample, they report that e>i i is 62 and y is 32 I estimate the models separately for each quintile based on R&D intensity Each year I assign firms to a quintile prior to

29

Reproduced with permission of the copyright owner Further reproduction prohibited without permission.

calculating the abnormal earnings and other information variables If a firm has missing data in any year then it is dropped from the sample in that year As can be seen by the number o f observations, firms in the fifth quintile have more missing data than firms in the other quintiles This is likely due to a large proportion o f high R&D intensive firms entering the sample in recent years and missing the necessary prior year data.

30

Reproduced with permission of the copyright owner Further reproduction prohibited without permission.