R&D INVESTMENTS, BOND RATINGS AND BOND RISK PREMIUMS
A THESIS
SUBMITTED TO THE FACULTY OF THE GRADUATE SCHOOL OF THE UNIVERSITY OF MINNESOTA
BY
CHARLES SHI
IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF
DOCTORAL OF PHILOSOPHY
Judy Rayburn, Adviser
August 2000
Trang 2UMI Number: 9980338
s2 UMI
UMI Microform9980338
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
Trang 3UNIVERSITY OF MINNESOTA
This is to certify that I have examined this copy of a doctoral thesis by
Charles shi
and have found that it is complete and satisfactory in all respects, and that any and all revisions required by the final
examining committee have been made —
C m4 CS
CC,
: Judy Rayburn
Signature of Faculty Adviser
Date
Trang 4ACKNOWLEDGEMENTS
I gratefully acknowledge invaluable guidance and critiques from Judy Rayburn (adviser)
and Pervin Shroff Thanks also go to my other committee members: Chandra Kanodia and Matt Mitchell This paper benefited from comments of and discussions with David Aboody, Chun Chang, Charles Lee, Baruch Lev, Dawn Hukai, Anjit Mukherji, Stephen Penman, Bal Radhakrishna, Theodore Sougiannis, Andrew Winton, Peter Woodlock, and
Trang 5Abstract
The debate on whether corporate R&D investments should be capitalized as asscts or expensed as incurred begs the question: which effect dominates — the future benetits
from R&D investments or their riskiness? Extant R&D literature focusing on the relation between R&D variables and equity metrics is plagued with the inability of researchers to
address this question This is because both the benefits and riskiness of R&D have the
same directional impacts on the equity valuation of levered firms (Merton 1973, 194)
This study adds another dimension to the literature by assessing the combined effects of the future benefits and riskiness of R&D in the context of the bond market
Option pricing theory stipulates that the mean (expected future benefits) and the
variance (riskiness) of R&D investments play opposite roles in the pricing of bonds Relying on this theoretical framework, | document significant positive associations of
R&D variables with bond default risk and bond risk premium This suggests that, for creditors, the nsk of R&D appears to dominate the future benefits of R&D In other
words, creditors view R&D investments more like risk proxies than assets Since the bond market has remained to be the firms’ most significant external financing channel
(Anderson et al., 1994), and R&D investments have become increasingly important to the U.S economy, this paper generates new evidence for debates on the accounting treatment
Trang 6Table of Contents
Acknowledgments Abstract
Table of Contents
Part 1: Overview of R&D Accounting and Literature Review
Chapter |: Accounting for Research and Development Costs: An Overview Chapter 2: Issues in Accounting for Software Development Costs
Chapter 3: Literature Review and Current Debate
Part Il: R&D Investments, Bond Ratings and Bond Risk Premiums
Chapter 4: Introduction
Chapter 5: Contribution of This Study Chapter 6: Research Methodology
Chapter 7: Sample Selection and Descriptive Statistics
Chapter 8: Empirical Results Chapter 9: Conclusion
Part [11: Summary and Future Research Chapter 10: Summary and Future Research
Trang 7Appendix I: Summary of R&D Accounting Around the World
Appendix II: Effect of Capitalization on Variability of Software Development
Tables: Table 1: Table 2: Table 3: Table 4: Table 5: Table 6: Table 7: Table 8: Table 9:
Expenditures and Reported Earnings
Sample Composition and Averages of Some Key Variables
Descriptive Statistics for Selected Variables
Pearson Correlations
Results of SUR Estimation: Annual R&D Expenditures
Results of SUR Estimation: Multi-year R&D Constructs
Results of SUR Estimation: Adjusted Model
Results of SUR Estimation: Incremental Effects of Multi-year R&D Constructs over Annual R&D
Results of SUR Estimation: LS Measure
Differential Associations: Long vs Short Useful Life Samples Results of SUR Estimation: LS Measure
Controlling for Other Risk Factors
Trang 8Part I: Overview of R&D Accounting and Literature Review
Chapter 1: Accounting for Research and Development Costs: An Overview Accounting for Research and Development Costs
Expenditures devoted to developing a new product or process or to improving an
existing product or process are typically classified as research and development (R&D)
The Financial Accounting Standards Board (FASB) issued Statement of Financial Accounting Standards (SFAS) No 2 in October 1974 The Statement mandates that firms
expense R&D expenditures immediately as they are incurred, and that the aggregate amount of R&D costs’ be disclosed
The disclosure requirement is an important component of SFAS No 2 Many
firms previously did not report R&D expenditures as a separate line item in their financial
statements (Intermediate Accounting, 5" edition, Chasteen et al.), presumably fearing
that the disclosure of the information may reveal crucial proprietary knowledge to their competitors and jeopardize their competitiveness The FASB, supported by respondents
to the Discussion Memorandum, added the disclosure requirement to the Statement based on the belief that the R&D outlays are informative for assessing the firms’ future performance
Prior to the issuance of SFAS No 2, there were diverse practices of accounting
and reporting on R&D Due to the absence of regulatory guidance, firms enjoyed a great
deal of discretion on how to account for their R&D outlays Some firms chose to capitalize a certain portion of the R&D expenditures and to amortize them over some
Trang 9
arbitrary period (no more than 40 years), while others opted to expense all R&D outlays as incurred According to the Disclosure Journal of Index of Corporate Events (annual edition, May 1973-April 1974), 560 publicly traded firms capitalized their R&D costs
Two factors prompted the FASB to reduce the flexibility allowed in reporting the firms’ R&D activities First, the existing diversity of practices created difficulties for cross-sectional comparative analysis of firms’ performance The FASB believed that a uniform accounting treatment of the R&D outlays would better serve the necds of financial statement users (SFAS No 2, paragraphs 54-55) Second, as R&D became an increasingly important cost component for many firms, the accounting choice of capitalization vs expensing could have a substantial impact on the firms’ financial statements (Vigeland, 1981)
Four alternative accounting methods were considered by the FASB: (1) expensing-all-R&D as incurred, (2) capitalizing-all-R&D as incurred, (3) selective capitalization capitalizing a portion of R&D costs if certain conditions are satisfied and
expensing all other costs, and (4) accumulation of all costs in a special category until the existence of future benefits could be established
The fact that firms invested a significant amount of their resources in R&D
programs suggested that the R&D activities, at least at an aggregate level, provided the
firms with future benefits such as increased revenue streams, high profitability and
market share gains However, there was typically a substantial degree of uncertainty regarding future benefits of individual R&D projects Also, the FASB, at that ume, saw
Trang 10typically involved many projects at different stages of completion These projects had
varying degrees of uncertainty about their success rates Hence, it would be difficult to
come up with any meaningful, and verifiable (auditable) amortization schedules if R&D were capitalized on a firm-wide basis These considerations led to the FASB’s disfavor of
the capitalizing-all-R&D option
The alternative proposals of selective capitalization and accumulation of costs in a
special category were rejected because the implementation of both proposals would require the identification of conditions, such as technological feasibility, that allow the
commencement of R&D capitalization The FASB did not believe that the establishment
of these conditions could be objectively and comparably made by all firms (SFAS No 3 p 55)
SFAS No 2 does not apply to R&D activities conducted under a contractual
arrangement or R&D costs specifically reimbursable under the terms of a contract In other words, if a firm performs R&D activities both on their own account and under a
government contract, the firm may capitalize and amortize the costs associated with the
contract but must expense the R&D costs related to its own business
The expensing-all-R&D approach required by SFAS No 2 was viewed by many
as a simple, practical, and conservative way of handling the uncertainty and measurement difficulties embedded in R&D activities More than two decades have passed since SFAS
No 2 was issued The Statement remains the only authoritative pronouncement
Trang 11An International Perspective
Expensing all R&D expenditures in the year they are incurred is an uncommon
practice around the world According to a survey compiled by Coopers & Lybrand (1993 see Appendix 1), only the United States, Germany and Mexico, 3 out of 33 listed countries, require immediate expensing of R&D expenditures The high degree of uncertainty involved in estimating the amount and timing of future benefits of R&D
investments, and the lack of a causal association between individual R&D projects and
the respective future benefits form the grounds for the use of immediate expense
recognition
The capitalization approach, an alternative to the expensing rule, is more widely used in the rest of the world Proponents of the capitalization treatment maintain that firms should be allowed to capitalize a portion of R&D costs, if a R&D project is assessed to be highly likely to lead to the successful introduction of a new product or process, and the future benefits flowing from the project can be reasonably estimated Proponents further argue that capitalization has two main advantages over expensing: (1) it provides a better matching between the expense and future revenue; and (2) the related
disclosure of capitalization and subsequent amortization may convey to financial statement users additional value-relevant information, such as the expected success rates
of R&D investments and estimated economic lives of capitalized R&D assets (Eccher
1997; Chambers et al 1998)
As indicated in Appendix I, 30 out of 33 surveyed countries permit capitalization
of varying portions of R&D expenditures if certain conditions are satisfied The
Trang 12commercial success of a product or a process under development The requirements are
less stringent for some countries like Italy and Brazil, where the accounting choice is
largely at a firm’s discretion as long as reasonable justification is provided In contrast
Canada, New Zealand, the United Kingdom, and International Accounting Standard (IAS) 9, Research and Development Costs, take the position that research expenditures
should be subject to immediate expensing, while development costs are allowed to be
capitalized conditional on the satisfaction of some pre-specified conditions The rationale is that it is the development process that eventually turns the viable research knowledge into a tangible product or process Since these countries’ criteria for capitalization are very similar to those contained in IAS 9, it is instructive to take a close look at [AS 9
Under [AS 9, all of the following conditions must be met in order for a firm to capitalize
the development costs:
(1) The product or process is clearly defined and the costs attributable to the product or process can be separately identified and measured reliably;
(2) the technical feasibility of the product or process can be demonstrated; (3) the enterprise intends to produce and market, or use, the product or process: (4) a market for the product or process exits or, if it is to be used internally rather than
sold, its usefulness to the enterprise, can be demonstrated; and
(5) adequate resources exist, or their availability can be demonstrated, to complete
the project and market or use the product or process {Paragraph 17, IAS 9]
The primary emphasis of IAS 9 centers around whether a project can pass the technological feasibility test and whether the firm has intentions and resources to complete the project and to make it commercially successful Given the substantial uncertainty inherent in R&D projects and their product markets, the implementation of the rule remains, to a large degree, a discretionary choice
Trang 13Chapter 2 Issues in Accounting for Software Development Costs An Exception to SFAS No 2
The software industry was in its infancy when SFAS No 2 was issued in October 1974 The FASB took the position that development of computer software should be
treated as a R&D activity and hence should be expensed as incurred according to SFAS No 2 The majority of software firms, following the issuance of SFAS No 2, expensed
software development costs in the year they incurred
However, many software companies questioned the applicability of the Standard to software creation activities They argued that not all of software development costs fell into the category of R&D expenditures In response to a growing number of complaints from the software industry, the FASB issued FASB Interpretation No 6, Applicability of FASB Statement No.2 to Computer Software in February 1975, and subsequently FASB Technical Bulletin No 79-2, Computer Software Costs in 1978 Interpretations of the two supplements were such that not all costs involved in software creation process are necessarily R&D This left the door open for software companies to capitalize a certain
portion of their development costs
As the software industry grew rapidly, more and more companies chose to capitalize their software development costs During the early 1980s, as many as 12 to 20 percent of software companies capitalized some portion of their software creation costs
(McGee 1988) Even though the FASB issued several pronouncements, such as FASB
Trang 14comparability of software companies’ performance In an attempt to curtail the diversity in practice, the SEC announced a moratorium on the capitalization of software
development costs in April 1983, which prohibited public companies from capitalizing software creation costs unless they had disclosed their practice previously The
moratorium remained effective until the issuance of SFAS No 86 in August 1985
The Accounting Standards Division’s Task Force of the American Institute of
Certified Public Accountants (AICPA) met to discuss accounting for the development
and sale of computer software in late 1982 Less than three years later, the FASB adopted SFAS No 86, Accounting for the Costs of Computer Software to be Sold, Leased, or Otherwise Marketed The Statement requires that costs involved in software creation be dichotomized according to the technological feasibility of the project Technological feasibility is typically established upon completion of a detailed program design or
working model All costs prior to the technological feasibility stage must be expensed as
incurred as R&D Software development costs incurred subsequent to the establishment of technological feasibility, labeled as software production costs in SFAS No 86, must be capitalized The capitalized software costs are also required to be revalued against their
net realizable value” on a product-by-product basis at each quarterly balance sheet date
and must be written down to net realizable value if the book value of the capitalized costs exceeds the net realizable value Amortization should be carried out on a product-by- product basis using the greater of revenue-based amortization or straight-line amortization over the remaining estimated economic life of the software
Trang 15
SFAS No 86 excluded the development costs of software projects for internal use
because it was not perceived as a significant issue in financial reporting at that time
(Statement of Position 98-1, paragraph 1) As software developed for internal use became more common, the AICPA stepped in and issued Statement of Position (SOP) 98-1 Accounting for the Costs of Computer Software Developed or Obtained for Internal Use in March 1998 SOP 98-1 mandates the capitalization of development costs for internal-
use software in the application development stage; furthermore, it requires all software
costs incurred in the preliminary project stage be expensed as incurred The capitalized costs should typically be amortized on a straight-line basis The impairment test is required in accordance with SFAS No 121, Accounting for the Impairment of Long-Lived Assets and for Long-Lived Assets to be Disposed Of
Firm Characteristics of Capitalizers vs Expensers
SFAS No 86 allows flexibility in its implementation A software firm is required
to capitalize development costs only when technological feasibility has been established
and the expected net realizable value of a software product is assessed at higher than the book value of the capitalized costs Since significant subjectivity is involved in evaluating these capitalization requirements, firms have considerable discretion in determining whether to capitalize or expense software development costs
Trang 16capitalization usually outweighs the income-deceasing amortization To the extent that
managers may be motivated to choose an accounting method to maximize reported
accounting numbers,” firms have incentives to favor capitalization over expensing On the other hand, capitalizing soft intangibles is generally viewed as a less conservative accounting practice As a result, larger firms would incur higher political costs and are more likely to suffer a reputation loss if they chose the capitalization treatment (Daley and Vigeland, 1983) Concurring with the political cost hypothesis, anecdotal
observations indicate that software heavyweights such as Microsoft, Novell and Borland have consistently expensed their software development costs
Since software firms may self-select their software reporting methods, it is
interesting to determine firm attributes that distinguish software capitalizers from expensers Drawing from the literature on the debate of the SFAS No 86, Aboody and
Lev (1998) evaluated the ability of six firm-specific attributes to discriminate between software capitalizers and expensers: firm size, software development intensity (the ratio of annual software development costs to sales), profitability, profitability change
leverage, and a systematic risk factor Using 163 sample firms during 1987 to 1995, Aboody and Lev (1998) found that firm size, profitability, and software development
intensity are significantly associated with firms’ accounting choice In particular, smaller, less profitable firms with higher software development intensity are more likely to capitalize software development costs in their sample
Trang 17
Software Firms’ Change of Attitude Toward SFAS No 86
The issuance of SFAS No 86 in 1985 was a result of two factors: overwhelming
support for the capitalization treatment by software companies, and, intensive lobbying
efforts by the then software trade group of - ADAPSO (The Association of Data Processing Service Organizations) The managerial motives to fight for the right to capitalize software development expenditures are quite transparent Capitalization typically results in more attractive financial measures, e.g., higher net income, return on equity, and total assets These apparent benefits from capitalization attracted as many as
12 to 20 percent of software companies to capitalize some portion of their software creation expenditures during the early 1980s, before SFAS No 86 was issued (McGee
1988) After SFAS No 86 became effective in 1986, a majority of the software companies switched to capitalizing a portion of development costs (Aboody and Lev
1998) However, the support for capitalization of software costs has diminished over
time Ironically, the Software Publishers Association, that lobbied for the issuance of
SFAS No 86, filed a petition to the FASB seeking abolition of the Statement in March
1996 What factors could motivate the dramatic reversal of the software firms’ attitude
toward the capitalization rule they supported in the early 1980s?
Aboody and Lev (1998) provide a motive for software producers’ change of
attitude Using software companies’ data from a recent nice-year span (1987-95), they provide evidence that capitalization failed to continuously boost net income and return on
equity by the mid-1990s The income-decreasing amortization eventually exceeded the
income-enhancing capitalization as firms matured (which took place on average in 1995
the motivation for management choice between alternative R&D accounting methods, see Horwitz and Kolodny 1980; Daley and Vigeland 1983)
Trang 18based on their sample) Hence, those companies lost their original incentives to continue
capitalizing software development expenditures
I provide one other plausible factor to explain software firms’ change of
preference; the adoption of SFAS No 86 seems to result in an increase in earnings
variability on average The benefits of capitalization (higher reported earnings and returns
on equity) diminish as the software industry matures while costs of adopting the rule,
such as a more volatile earnings stream, still persist When the costs eventually exceed
the declining benefits, firms might have an incentive to abandon their support for the
capitalization rule
To investigate the impact of capitalization of software development costs on earnings variability, | compare the unsigned coefficient of variation of reported earnings under capitalization with that under the as-if expensing rule over a four-year horizon
subsequent to the adoption year of SFAS No 86.* This approach isolates the impact of
other confounding factors on earnings variability Disclosure requirements imposed by SFAS No 86 make it feasible for researchers to undo the capitalization process For
example, as-if earnings under expensing can be calculated as reported earnings under capitalization minus annual capitalized software costs plus annual amortized software
costs Since software development expenses are the only earnings component affected by the Standard, it is instructive to compare the variability of this particular earnings item
under capitalizing vs expensing treatments Findings based on 43 capitalizing firms are
presented in Appendix II
Trang 19
Panel A of Appendix II shows the effects of capitalization on the variability of
software development expenditures Software expenses under capitalization are equal to
the software development expenditures expensed plus amortization plus occasional write-
offs Software expenses under as-if expensing are the total annual software development outlays (the sum of software development expenditures expensed and capitalized) All variables are normalized by net annual sales The change in variability is measured as the
difference between the dispersion in software expenditures under capitalization and that
under as-if expensing Percentile statistics show that 69% of coefficient of variation differences are greater that zero, implying that, for my sample, software development
expenses under capitalization are generally more variable than under as-if expensing In addition, Wilcoxon Signed-Ranks test shows that the median difference in earnings variability is significant at the level of 0.0009
The effects of software capitalization on variability of reported earnings per share are summarized in Panel B of Appendix II Since scaling by the weighted number of shares may not form a sufficient control for the size effect, the change in variability of earnings is measured as the ratio of dispersion of reported earnings per share under capitalization to adjusted earnings under the as-if expensing rule (this ratio analysis is also used in Francis 1990) Strikingly similar to the results for the change in variability of software expenditures, 62% of earnings dispersion ratios are greater than |, suggesting that most of 43 capitalizing firms’ reported earnings are more volatile than adjusted earnings under expensing Furthermore, the Wilcoxon statistic is significant at the 0.0001
Trang 20Taken together, an analysis of 43 capitalizers indicates that capitalization seems to
result in higher earnings variability The increased earnings variability may reflect the
dynamic, risky nature of software development Through capitalization, these risks and uncertainties involved in software development ventures are introduced into reported
earnings, which may consequently lead to higher earnings variability
Trang 21Chapter 3: Literature Review and Current Debate
A growing body of empirical accounting research has been devoted accounting
treatments of R&D expenditures since the issuance of SFAS No 2 in 1974 Early studies
examined the economic consequences of the expensing rule A significant portion of the literature, however, has focused on the value-relevance of R&D, that is, a firm's R&D
investments provide future benefits, and hence are reflected in the stock price Unt!
recently, few efforts were made to examine the reliability of R&D accounting The following review of prior research is organized along the aforementioned three lines of
inquiry: economics consequences, value-relevance and reliability
Economic Consequences of SFAS No 2
Following the issuance of SFAS No 2, research was aimed primarily at the impact of the Statement on managerial R&D investment decisions The question was investigated indirectly by (1) testing the capital market reaction surrounding the releases
of the Statement-related events (e.g, Dukes, 1976; Vigeland, 1981, Wasley and
Linsmeier, 1992), (2) surveying potentially affected firms’ managers about their perceived effects of the Statement on R&D investments (e.g., Horwitz and Kolodny, 1980), or directly, (3) by examining changes in the levels of R&D outlays surrounding the adoption of SFAS No 2 for firms which previously capitalized R&D (e.g., Elliott et al., 1980; Horwitz and Kolodny, 1980; and Dukes et al., 1984)
The findings of these studies are mixed Horwitz and Kolodny (1980) reported a
Trang 22managerial R&D decisions were detected (Elliott et al., 1980; Vigeland, 1981) Later
attempts to reconcile the discrepancies in the prior studies indicate that the divergent
findings are driven, at least partially, by systematic differences in sample firm size between OTC and listed firms (Dukes et al., 1984; Wasley and Linsmeier, 1992)
Why are small firms’ R&D activities more likely to be adversely affected by the
Statement? When SFAS No 2 was issued in 1974, less than 20% of publicly traded firms
were capitalizing R&D and most of them were small (Barron’s, November 18, 1974)
Several factors have been advanced to explain why the capitalization method was more commonly used by small firms First, small, technology-oriented firms typically grow at a much faster rate than large firms, and their R&D is a major operating expenditure In
general, capitalization results in more favorable reported earnings which last beyond
initial adoption years Furthermore, many firms partially compensate managers based on
reported accounting variables such as earnings As a result, the managers have an
incentive to make an accounting choice which would maximize their bonus rewards Second, when compared with small non-technological firms or large firms, small, high- technology firms have less internal funds and rely more heavily on external financing at their early development stages (National Bureau of Standards [1976, p.18]}) Lower book
value of equity and earnings resulting from a mandatory switch to the expensing-all R&D rule may lead to a higher probability of violation of existing debt covenants or may increase the cost of raising new capital due to higher perceived risk by the market
participants Third, institutional investors tend to have less interest in small firms.” As a
result, a high portion of equity sources of small firms may come from “unsophisticated”
* According to surveys revealed by National Analysts Federation [1977], there was little institutional interest in companies with less than $50-100 million market capitalization
Trang 23“unaffiliated” investors It is probable that these investors take the reported numbers at their face value and their evaluation of the firms are influenced by unfavorable impact on
reported numbers.”
To summarize, the extant literature seems to suggest that SFAS No 2 has little impact on large firms’ R&D investments while there is evidence indicating that the R&D levels of small, technological firms may be impacted by the Statement
Value-Relevance of R&D
A central premise underlying the expensing rule is what the FASB believed, at
that time, to be the lack of a direct causal relationship between R&D expenditures and
specific expected benefits (SFAS No 2, paragraph 14).’ This claim has prompted a stream of subsequent research to examine the relationship A common approach ts to
relate R&D information to stock market valuation Hirschey and Weygandt (1985) report
a positive association between a firm’s R&D outlays and its market-to-book ratio Cockburn and Griliches (1988), Hall (1993), Chambers et al (1998), and Lev and
Zarowin (1998) also document that the equity market incorporates R&D into firm valuation Researchers interpret the positive associations between R&D constructs and
equity market metrics as evidence that R&D activities generate future benefits in the
aggregate
Woolridge (1988) and Chan et al (1990) employ a different approach to assess the value-relevance of R&D Using an event study methodology, they find that stock
° Horwitz and Kolodny (1980) surveyed chief financial officers of 168 deferral and 212 expensing firms Approximately 93 percent of the deferral firms as well as 75 percent of expensing firms believed that SFAS No 2 would have a negative effect on the small investors’ evaluation of affected firms
” The FASB was also concerned that the future benefit of R&D are too uncertain and unpredictable to be reliably measured (SFAS No 2, paragraphs 39-40)
Trang 24prices respond positively to the announcements of increased R&D expenditures, suggesting that the stock market views R&D as value-enhancing investments Bublitz and Ettredge (1989) report similar evidence by examining the association between cumulative abnormal stock returns and unexpected changes in advertising and R&D spending Their findings suggest that R&D provides long-lived benefits while benefits from advertising
are short-lived
Instead of inferring that R&D benefits from stock prices/returns, Lev and Sougiannis (1996) extract the value of R&D from subsequent reported earnings They
then, adjust reported earnings and net book values by an as-if capitalization treatment They show that information contained in the adjustments is value-relevant to the cquity
market by association of stock prices and returns with the adjusted accounting numbers The potential informational benefits of capitalized R&D expenditures are also reported
by Chambers et al (1998) Applying hypothetical amortization schemes uniformly across the sample years, Chambers et al (1998) document that adjusted earnings and book values are more strongly associated with stock prices than the reported counterparts under the expensing rule They, then, argue that even arbitrary capitalization treatments, having
an advantage of allowing no management's discretion, seem to still improve the “usefulness” of accounting numbers for valuation purposes
The methodologies used in value-relevance studies have two limitations First, the association of security prices/retums with estimates of R&D assets or adjusted accounting variables is also consistent with alternative explanations For example, high
value of R&D estimates may also proxy for the high uncertainty and riskiness inherent in
Trang 25Second, the realized security prices are endogenously determined by the information conveyed by the reported numbers One cannot observe what prices would have been if R&D had been capitalized Therefore, implications for alternative reporting practices
drawn from the observed prices under the expensing rule may be problematic
This concern can be mitigated by investigating the software industry where
software firms, under SFAS No 86, are required to capitalize software development costs incurred subsequent to the establishment of technological feasibility Aboody and Lev
(1998) find that capitalized software costs are incorporated in stock prices and returns, hence providing direct evidence that software capitalization is value-relevant to cquity holders
Reliability of R&D Benefits
In contrast to ample studies on the value-relevance of R&D, research regarding the reliability of R&D outcomes is very scarce Kothari et al (1998) gauge the uncertainty/reliability of future benefits from R&D outlays relative to that from property plant and equipment (PP&E) Regressing the standard deviation of future carninys a
proxy for the uncertainty of R&D benefits, against annual R&D expenditures and PP&E they document that the coefficient on R&D is about three times as large as that on PP&E The evidence is consistent with their hypothesis that the future benefits resulting from
R&D activities are more uncertain than that from tangible investments
Trang 26models—the expensing-all, the full cost, and the successful efforts methods They show that the successful efforts method appears to win the horse race in providing value-
relevant information when there is no managerial manipulation Then, they allow management to exercise discretion by randomly postponing writing down the R&D assets for unsuccessful projects, and find that the successful efforts scheme still has superior information to the other alternatives if the frequency of delays is modest, even though the
information advantage is dampened by management’s opportunistic actions They conclude that the successful efforts method has the potential to enhance the
informativeness of accounting reporting as opposed to the expensing rule Current Debate
FASB issued the expensing rule in 1974 on the grounds that there was a high
degree of uncertainty about the future benefits of R&D, that there was no direct causal
relationship between R&D expenditures and increased future benefits in terms of sales
profits and share of industry sales, and that comments and feedback submitted by security analysts and bankers indicated that capitalization was not useful in assessing a firm's
earnings potential (SFAS No 2, paragraphs 39-41, 50, 54) Consequently, the FASB concluded that expensing was a better way to reflect the uncertain and risky nature of
R&D activities,
Since SFAS No 2 was issued, the U.S has witnessed a substantial transformation from a traditional manufacturing to an information-based digitalized economy Intangibles have played an increasingly important role in the corporate value creation process Numerous studies assert that the failure of current accounting systems to capture the growing amount of increasingly important intangibles has, in part, contributed to the
Trang 27alleged declining usefulness of financial reporting (Rimerman, 1990; Elliott and Jacobson, 1991; Jenkins, 1994; Stewart, 1995; Wallman, 1995; Chang, 1998; Lev and
Zarowin 1998) Moreover, in contrast to the FASB’s belief, subsequent academic
research has provided bountiful evidence that R&D investments are consistently associated with stock prices and retums in the aggregate, suggesting that the benefits of
R&D are long-lived
The debate is not about whether intangibles assets exist, but about how to better measure and disclose them, and about the pros and cons of reporting them even though
they may never be ideally measured Proponents of the expensing rule argue that R&D capital cannot be measured in a meaningful way due to high uncertainty and risk inherent
in the expected benefits of R&D Security analysts like the Association for Investment Management and Research (1993) also contend that the capitalized R&D costs would bear little information relevant to the valuation process due to the lack of correlation
between the R&D costs and subsequent future benefits CFOs of some companies predict
that reporting unreliable and noisy estimates of the intangibles in the financial statements might result in unintended and harmful results, e.g., misled investors and distracted managers (CFO, February 1999, page 30)
Critics of the existing expensing treatment are primarily academics, and, to certain extent, regulators Such critics contend that current accounting models do poorly
in capturing the intangibles, although intangibles constitute a rapidly increasing share of corporate value Thus, it is argued, the value-relevance of existing measurement systems
is deteriorating over time There is a growing need for new methods which better reflect
Trang 28
the value of the intangibles These calls for better reporting systems have led to actions by regulators and policy makers For example, the SEC held a symposium solely devoted
to financial accounting and reporting on intangible assets in 1996; in September 1998, the International Accounting Standards Committee (IASC) passed a new standard on intangibles (IAS 38), allowing firms to capitalize a wide range of internally developed
intangibles; the Canadian Institute of Chartered Accountants (CICA) has launched the
Canadian Performance Reporting Initiative, and one of its five key projects rcyards intellectual capital management focusing on “experimenting with practices and techniques for managing and measuring intellectual capital”
Although a consensus is emerging that new accounting measurements are needed
in response to the new economy, the process of finding better ways of measuring evaluating and reporting on critical intangible sources of wealth is still at an early stage of development As a result, there have been calls for more research A Brookings
Institute task force, co-chaired by Steve Wallman, a former SEC commissioner, has been convened to tackle broad aspects of intangibles related issues, and make policy
recommendations to reduce or remedy current measuring and reporting deficiencies on
Trang 29Part II: R&D Investments, Bond Ratings and Bond Risk Premiums Chapter 4: Introduction
The debate on whether R&D expenditures should be capitalized as assets or
expensed as incurred reflects the trade-off between two asset recognition criteria.’ On one hand, consensuses have emerged that R&D expenditures on aggregate provide firms with
future benefits, such as introduction of new products and increased revenues That ts R&D investments result in value-creating economic assets, which should be recognized in the financial statements On the other hand, the risky and unpredictable nature of R&D
outcomes makes it extremely difficult to quantify the future benefits of R&D.'° Reporting unreliable and noisy estimates of R&D intangibles on the balance sheet would mislead
investors and creditors
The on-going debate begs the question: which effect dominates - the future benefits from R&D investments or their riskiness? '' Extant R&D literature investigating the relation between R&D variables and equity metrics is plagued with the inability of
researchers to address this question This is because both the benefits and riskiness of R&D have the sume directional impacts on the equity valuation of levered firms (Merton 1973, 1974) In other words, the risk and uncertainty of R&D, along with the future
* The Statement of Financial Accounting Concepts No 6, Elements of Financial Statements (1985), asserts that a firm will recognize a transaction as an asset only if (1) the transaction will generate future benefits for the firm, and (2) the future benefits can be quantified with a reasonable degree of precision
'® As the case for tangible investments, capitalization of R&D may be made on a cost basis However, subsequent amortization and revaluation of capitalized intangibles still require the reasonably precise assessment of the outlook of R&D projects For example, the knowledge of a future revenue stream from a pipeline drug would be critical to form a meaningful and informative amortization schedule Moreover the values of R&D intangibles, subject to various technological shocks, often change dramatically over ume while financial statements are not updated in a timely fashion
Trang 30benefits of R&D, also contribute to the widely documented positive association between R&D measures and stock prices/returns
This paper adds another dimension to the literature by studying the combined
effects of the future benefits and riskiness of R&D in the context of the bond market The
advantage of taking the perspective of the bondholder instead of equity holder is that, in
contrast to the same directional effects on equity valuation, the future benefits and
riskiness of R&D play opposite roles in the pricing of bonds (Merton 1973, 1974) Therefore, the bond market provides a unique setting enabling me to investigate whether R&D investments are expected to generate future ner benefits and have a more asset-like
nature or whether the risk associated with future benefits is so high that R&D is less asset-like in nature and more useful as a measure of risk
Option pricing theory stipulates that the mean and the variance of future cash
flows from investments have opposite impacts on the value of the firm's debt (Merton
1973, 1974) That is, an increase in the mean of future cash flows arising from the firm's
investments increases the value of its bonds by reducing the probability of default, while an increase in the variance of future cash flows decreases the value of its bonds by increasing the probability of default Relying on this theory, I examine the association
between bond risk measures and R&D investments to determine whether the mean effect
Trang 31would imply a stronger variance effect that swamps the mean effect of future benefits
from R&D investments
The sample used in this study consists of 132 new issues of industrial bonds by R&D-intensive firms The bonds were issued between 1991 and 1994 and were rated by Moody’s investment services Two commonly used bond risk measures, Moody's bond ratings and bond risk premium are used as dependent variables Risk premium is
calculated as the difference between a bond’s market yield-to-matunity and the yield on a U.S Treasury bond of comparable maturity on the issuance date The advantage of using
the bond risk premium is that the measure directly controls for the impact of the term structure of interest rates on raw yield to maturity This is particularly important because the sample bonds were issued over four years and the yield to maturity varies with the level of interest rates
This study finds a significant positive correlation of the firm’s annual R&D expenditures with bond default risk (proxied by bond ratings) and risk premium, controlling for other common bond risk determinants This finding suggests that, from the
creditor’s point of view, the adverse effect of high volatility and uncertainty (the variance
effect) of the firm’s R&D activities outweighs the favorable impact of the firm value increments (the mean effect) In other words, even though the expected mean valuc of the firm’s R&D outlays may be positive, it does not overcome the huge variance of the future cash flows Therefore, for creditors, R&D expenditures may reflect less assct-like characteristics but more risk attributes proxying for the excess variance effect over the
Trang 32To further test whether the finding depends on the choice of R&D measures, | usc
three other R&D constructs commonly used in the literature'* to estimate R&D “assets”
Specifically, variables are constructed based on the sum of the past five-year R&D expenditures, five-year straight-line depreciation, and industry-specific amortization
schedules, as reported by Lev and Sougiannis (1996, in Table 3, pp 121) All R&D constructs have significantly positive coefficients, indicating that the findings are robust with respect to the alternative measures
I further attempt to identify the R&D variable that is the best risk proxy by comparing the four R&D constructs While the three multi-year measures have comparable explanatory power, they demonstrate incremental power over the naive
variable (annual R&D expenditures) in explaining bond risk premiums This suggests that
these measures are better R&D risk proxies, probably because they are constructed from multiple years of observations, and incorporate the time series variation of R&D expenditures; hence they are more reflective of the firm’s R&D activitics
I also examine whether the Lev-Sougiannis measure (LS measure hereafter) provides more information regarding a firm’s R&D activities than the other proxies In contrast to the other “one-size-fits-all” constructs, the LS measure allows the
amortization schedules to vary across industries, and hence is likely to better capture the
risky and uncertain nature of R&D To test whether the LS measure has discriminative
power of R&D risk, the sample is partitioned by the estimated length of the R&D useful life (partition criteria are elaborated in section III) The partitioned analysis indicates that
'? One may argue that using annual R&D expenditures implicitly assumes that R&D outlays are expensed as incurred, hence the variable may not fully reflect all future benefits accruing from R&D investments and consequently may underestimate the mean effect
Trang 33there are significantly differential associations between the bond risk measures and the LS measure across the two samples, implying that R&D investments with longer uscful
lives are associated with higher uncertainty and risk The ability of the LS measure to discriminate varying degrees of R&D risk across the industries indicates that the amortization schedules derived by Lev and Sougiannis (1996) are informative and therefore lend the LS measure an edge over the other competing proxies
This study sheds light on our understanding of the trade-off between the future benefits (relevance) and risk (reliability) related to the financial reporting of R&D expenditures Under the Statement of Financial Accounting Standard (SFAS) No 3, R&D outlays are required to be expensed as incurred Critics of the expensing rule contend that there is a growing body of evidence that R&D on average generates future benefits (E.g., Hirschey and Weygandt 1985; Cockbum and Griliches 1988; Hall 1993;
Lev and Sougiannis 1996) Hence recognizing the value-creating R&D investments as assets will enhance the value-relevance of the financial statements On the other hand
proponents of the expensing rule argue that the outcomes of R&D activities are too
volatile to warrant a capitalization treatment The results in this study show that, from the
perspective of the bond market, the uncertainty of R&D appears to overwhelm the potential increase in the mean of future cash flows from R&D investments This piece of
Trang 34in the management’s discussion and analysis to aid the users of financial statements in
assessing the value of R&D activities.'°
The findings of this study may be of interest to policy makers Regulators and policy makers recognize the need for a better method to reflect the risks and value of intangibles.'* Given the dominant importance of debt financing'* and R&D investments to the U.S economy, this paper generates new evidence for debates on the accounting
treatment of intangibles
'’ For example, information regarding the expected useful lives of main products, as typically disclosed by software companies, will help to assess the expected benefits and risk of investments in software products
'* For example, SEC held a symposium solely devoted to financial reporting on intangible assets in 1996
FASB recently reversed an earlier position which would allow the capitalization of purchased R&D trom business combinations and stated “we will probably consider R&D in its entirety at some future date when we have the resources necessary to pursue the issues" (FASB news release, July 28, 1999)
'S Among the three primary extemal financing channels - the issuance of bonds, common stocks, and preferred stocks, the bond market has remained to be the most important source for the firm to raise its external capital For example, the issuance of bond constitutes 87.1 percent of total three offerings in the
1938 - 1941 period, and the share of bond financing continues to be above 80 percent in the 1990 - ¡993
Trang 35Chapter 5; Contribution of This Study
Existing value-relevance studies draw inferences from the association between
R&D variables and stock market metrics Option pricing theory states that both the mean and variance of future cash flows from R&D investments impact equity value in the same
direction (Merton 1973, 1974) As a result, the widely documented positive relationship
between R&D constructs and stock prices/retums is attributable to both the mean and
variance effects Stated differently, the risk and uncertainty of R&D also contribute to an
increase in stock prices and returns.'® Recognition of the contribution of R&D risk provides new implications for interpreting the empirical results For example, one cannot
unambiguously infer from the positive relationship that R&D outlays generate net positive expected future cash flows, i.e., net positive mean effect.'’ Second, the positive
association does not necessarily mean that the equity market treats R&D as if it were an
asset because R&D as a risk factor is also consistent with the findings
This paper contributes to the literature by shedding light on a fundamental question underlying the on-going debate on R&D accounting: which effect dominates the future benefits or the riskiness of R&D? The ambiguity in interpretations of the future benefits and riskiness of R&D results from the fact that both mean and variance effects have the same directional impacts on equity valuation In contrast, the two factors
'® Intuition for why the variance of the future cash flow distribution increases equity value stems trom the combination of two factors First, the manager/owners, who have residual claims, benefit from large payoffs if the investments pan out well Second, owners have limited liability and hence may not have to be responsible for the whole loss if the projects turn sour Therefore, the owners are tempted to take on risky projects at the expense of debt holders, resulting in risk shifting (wealth transfer) from the equity holders
(debt holders) to the debt holders (equity holders)
Trang 36play opposite roles in bond valuation Therefore, the bond market provides a unique setting to estimate the combined effects | document the positive association between
R&D variables and the bond risk measures, suggesting that the risk and uncertainty of R&D outweigh bond value increment arising from the mean effect, and hence R&D constructs are viewed more like risk proxies than assets by creditors Furthermore, the
positive coefficients on R&D variables may be also due in part to the perception thal
Trang 37Chapter 6: Research Methodology
Prior bond studies typically regress bond yields or default risk measures on the
variables of interest and a list of other determinants found important in explaining bond
yields and default risk as control variables (see Reiter 1990 for the literature review) To
investigate the link of R&D constructs to the risk of default and the risk premium, the
following system of two linear equations are examined:
Default Risk ,.; = f (R&D Constructs ,, Other Bond Determinant Set 1) (1)
Risk Premium ,.,; = f (R&D Constructs ,, Other Bond Determinant Set 2) — (2)
The disturbances in equations (1) and (2) contain common unspecified factors For example, macroeconomic shocks such as a government monetary policy change
affect both default risk and risk premium The advantage of estimating the above two
equations simultaneously is that it accommodates the possibility that the error terms of the two equations are contemporaneously correlated and improves the efficiency of
parameter estimates The results from Lagrange multiplier tests, discussed in section V show a significant correlation between the error terms of the two equations, therefore, the
analyses are carried out using the seemingly unrelated regressions (SUR) model in place
of OLS to improve estimation efficiency
Dependent Variables
The dependent variable for equation 1, default risk, is proxied by indicator
variables for bond ratings, with integer values | through 5 representing the ratings of Aaa, Aa, A, Baa, and Ba below.'® Following the classic bond paper by Fisher (1959) and
'* [ lumped bonds with the ratings of Ba and B into category 5 because there are only 11 issues with B ratings Finer decomposition, assigning 5 and 6 to Ba and B issues respectively, does not change the results qualitatively
Trang 38recent studies such as Reiter (1991), I gauge the risk premium (PREM, measured in basis
points) as the difference between the yield to maturity on an industrial bond and the daily averages of the constant maturity yield on U.S Treasury Bond of comparable maturity on the issuance date If an industrial bond cannot be matched with a Treasury bond with the exact same maturity, a benchmark Treasury bond yield is constructed by linearly
interpolating the yields of two adjacent Treasury bonds with maturity closest to the
corporate bond maturity.'” Data on bond ratings and yield to maturity on corporate bonds
are collected from Moody’s Bond Survey Information on the constant maturity yield on
U.S Treasury bonds is from the Federal Reserve Board of Governors’ Statistical Release
Subscripts t+land t indicate that dependent variables are one year ahead of independent
variables (accounting variables) because bond premiums and ratings are affected most by publicly available past accounting information (Ederington and Yawitz 1986)."
R&D Constructs
The variable of interest in the current study is R&D I employ four different R&D measures: annual R&D expenditures (RD1), R&D estimates constructed from summation of past five-year R&D expenditures (RD2), five-year straight-line (RD3) and industry- specific amortization schemes (RD4) The first measure is a “flow” variable while the
other three are “stock” variables The sample for this study consists of the same five
R&D-intensive industries as those used by Lev and Sougiannis (1996), hence the
industry-specific amortization schedules are extracted from Table 3 of their paper In particular, the fourth construct for firm / at year ¢, RD4 ; , , is estimated as follows:
" For example, if a corporate bond has an 8 year maturity, a bench mark yield is computed by linearly interpolating the yields of Treasury bonds with maturity of 7 and 10 years
Trang 39N-1 k
RD4,, => RD,,_,1- }-d,) (3)
ka0 /=0
where N is the economic useful life of R&D ranging from 5 years in the industry of scientific instruments to 9 years in the chemicals and pharmaceutics industry An amortization scheme is represented by a sequence of d ,, j is from 0 to N-1 Therefore, RD4 is the sum of the unamortized portion of past annual R&D expenditures; it is the
amount of expenditures that are expected to produce current and future earnings over the useful life of the R&D intangibles
Annual amortization of the R&D intangibles is derived accordingly as follows:
RDAMT4,, = yd, RDS,
;z0 (4)
RD3 and RDAMT3 are similarly computed except that they are based on the uniform five year straight-line method which means that N equals 5 and d , is set to be 0.2
for all j, where j is from 0 to 4 To control for size effect, all RD variables are scaled by year-end market value of equity
Other Bond Risk Determinants
The selection of control variables is based on the literature on bond premium and ratings In general, the determinants of bond premium include: (1) default risk variables proxied by accounting information, (2) issue characteristics of maturity, call provision convertibility and subordination status, and (3) macroeconomic conditions such as the effect of the business cycle Particularly, the control variables for the risk premium
equation are defined as below:
Trang 40DE = Long term debt-to-equity ratio The higher this ratio, the higher default risk and
bond premium
PROFIT = Net income to net sales Profitability is expected to be negatively correlated with default risk
TIMES = Income before interest expense divided by interest expense Times interest
earned ratio is expected to be inversely associated with default risk and bond premium
LOGASSET = Log of total assets Firm size proxics for default risk and bond marketability Larger firms are expected to have lower default risk and
hence lower risk premium
Issue Characteristics:
LOGMAT = Log of years to maturity The longer years to maturity, the higher interest
rate risk exposure Hence, it is expected to be positively correlated with risk premium
LOGSIZE = Log of issue size It can be viewed as a measure of marketability, and is
expected to be inversely correlated with risk premium That is because, everything else being equal, the larger the size of a bond issue, the
more frequently one should expect the bond to be traded The more frequently a bond exchanges hands, the less uncertain the bond price is
On the other hand, the larger a bond issue size, the higher the debt burden, hence, the higher the probability of default Therefore, the impact of issue size is ambiguous
CALL = Years to first call over years to maturity Call provision exposes bondholders to interest risk Lower call ratio implies a higher level of call protection to
the issuer and hence higher interest risk exposure to bondholders This variable is expected to be negatively associated with risk premium
CONV = | for convertible bonds and 0 otherwise Other things equal, convertible bonds result in lower risk premium
SUBO = | for subordinated bonds and 0 otherwise Subordination status is expected to be associated with higher bond premium.”'
Macroeconomic Conditions:
ECYC = Average yield on Moody’s Aaa bonds for the month of issue less average
* This variable is dropped from PREM (risk premium) equations because it happens to be highly correlated