TH IOU

 

   

Michigzzi Sate

Tu§ : v.__ :"'.'_“
L'sti'v Cage)?

 
   
   

This is to certify that the
thesis entitled

AN EVALUATION OF
THE INFORMATION CONTENT OF
STATEMENT OF FINANCIAL ACCOUNTING STANDARDS NO. 2

presented by

Kathleen Elizabeth Sinning

has been accepted towards fulfillment
of the requirements for

Ph.D. Business - Acctg.

degree in

 

 

Major professor

Date /"//’ /7§’

0-7639

 

 

OVERDUE FINES ARE 25¢ PER DAY
PER ITEM

Return to book drop to remove
this checkout from‘your record.

 

 

 

C) Copyright by

Kathleen Elizabeth Sinning
1978

AN EVALUATION OF
THE INFORMATION CONTENT OF
STATEMENT OF FINANCIAL ACCOUNTING STANDARDS No. 2

BY

Kathleen Elizabeth Sinning

A DISSERTATION

Submitted to
Michigan State University
in partial fulfillment of the requirements
for the degree of

DOCTOR OF PHILOSOPHY

Department of Accounting and Financial Administration

1978

\

ABSTRACT

AN EVALUATION OF
THE INFORMATION CONTENT OF
STATEMENT OF FINANCIAL ACCOUNTING STANDARDS NO. 2

by

Kathleen Elizabeth Sinning

The objective of this study was to assess the information content
of the accounting and reporting requirements of Statement of Financial
Accounting Standards No. 2. Statement No. 2 requires all research and
development costs to be expensed and disclosed in the year incurred.

The Financial Accounting Standards Board based its expensing deci-
sion on the assertion that there is a high degree of uncertainty
about the future benefits of individual research and development
projects. The Board contended that the relationship between research
and development expenditures and the resultant future benefits is so
uncertain that alternative accounting procedures such as capitalization
would not provide investors with useful information for evaluating
the earnings potential of an enterprise. The FASB believed that the
immediate expensing of all research and development costs would better
enable investors to predict the risk and return of an investment.

The information content of Statement No. 2 was assessed by
.analyzing the relative stability of the systematic risk of affected

firms around the time the statement was issued. If Statement No. 2

Kathleen Elizabeth Sinning
provided new information to evaluate the risk and return of a security,
a shift in the systematic risk of those firms required to change their
accounting treatment of research and development expenditures would be
expected.

Two portfolios of firms were included in the study. The treatment
portfolio was comprised of firms that had capitalized all or part of
their R and D expenditures prior to Statement No. 2 and switched to
expensing those costs as a result of the statement. Firms that had
expensed and disclosed all of their research and development expendi—
tures prior to Statement No. 2 were included in the control portfolio.
A significant adjustment in the systematic risk of the control port-
folio would not be expected because the accounting and reporting pro-
cedures of those firms did not change as a result of the statement.

Ordinary least squares procedures and the market model were used
to estimate the systematic risk of the treatment and control portfolios.
An analysis of covariance procedure developed by Harvey (1976) was
employed to test the stability of each portfolio's systematic risk
during the period surrounding the issuance and adoption of Statement
No. 2.

To determine if there were significant differences in the risk
behaviors ofthe treatment and control groups, a portfolio was created
‘which consisted of monthly differences in the returns of the treatment
and control portfolios. An analysis of covariance procedure was used
to evaluate the constancy of the monthly betas calculated for the
"difference portfolio." If there were no significant differences in

‘the risk behaviors ofthe treatment and control groups, the beta

Kathleen Elizabeth Sinning
coefficient of the "difference portfolio" would be relatively stable.

The results of this analysis suggest that the accounting and
reporting requirements of Statement No. 2 possessed no information
content. Although the systematic risk of the treatment portfolio
(change firms) decreased dramatically around the time the Exposure
Draft was issued, the systematic risk of the control portfolio (non-
change firms) also decreased at that time. The statistical procedures
conducted on the "difference portfolio" indicated that the risk changes
of the two groups were not significantly different.

These results may imply one of two conditions: (1) capitalization
or selective capitalization of research and development expenditures
conveyed sufficient information to enable the market to evaluate the
risk of a security or (2) the market obtained information about firms'
research and development costs from alternative accounting and non-
accounting sources. The accounting and reporting policies of State-
ment No. 2 may have had no effect on investors' perceptions of

systematic risk because they provided redundant information.

ACKNOWLEDGMENTS

I would like to thank my committee members, Drs. Ronald Marshall,
Kelly Price, and Richard Simonds for their assistance with this research.
I am especially grateful to Dr. Simonds for his valuable advice and
guidance throughout all phases of the study.

I wish to express my appreciation to Deloitte, Haskins and Sells,
the Michigan Accountancy Foundation, and the Department of Accounting
and Financial Administration at Michigan State University for their
financial support during my doctoral program.

Thanks are due Mrs. Jo McKenzie for typing the preliminary drafts
as well as the final copy of the thesis and Dr. Christine Williams, who
provided me with a quiet office.

Finally, I would like to thank Hans Dykxhoorn for his unrelenting

encouragement, support and cooperation.

iii

TABLE OF CONTENTS

LIST OF TABIES O O O O O O O O O O O O O O O O O C O O O 0

LIST OF FIGURES O O O O O O O O O O O O O O O O I O O O 0

Chapter

I.

II.

III.

IV.

INTRODUCTION. . . . . . . . . . . . . . . . . .
Research Objective. . . . . . . . . . . . . .
Criticism of Statement No. 2 . . . . . . . .
Effects of Statement No. 2 . . . . . . . . .

LITERATURE REVIEW . . . . . . . . . . . . . . .

The Information Content of Annual Earnings
NMberso I O O O O O O O O O O O O O O O O O

The Information Content of Accounting Changes .

Information Content and Changes in Systematic
Risk 0 O O O O O I O O O O O O O O O O O O 0

RESEARCH DESIGN AND METHODOLOGY . . . . . . . .

Sample Selection. . . . . . . . . . . . . . .
Estimates of Systematic Risk. . . . . . . . .
Data Sources. . . . . . . . . . . . . . . . .
Statistical Testing Procedures. . . . . . . .
Difference Analysis . . . . . . . . . . . . .

RESULTS OF THE ANALYSIS . . . . . . . . . . .

Test for Constancy of the Risk Parameters . .

Analysis of Observed Beta Decline . . . . . .

Comparison of the Treatment and Control
Group Risk Changes . . . . . . . . . . . . .

SUMMARY AND CONCLUSIONS . . . . . . . . . . . .
Conclusions . . . . . . . . . . . . . . . . .
Implications of the Results . . . . . . . . .

Limitations of the Study. . . . . . . . . . .
Suggestions for Future Research . . . . . . .

iv

Page
vi

vii

0004‘

19
19
22
29
33
35
42
44
45
50
56

S6
65

73
82
84
85

88
89

Page

APPENDICES
A. Treatment Sample. . . . . . . . . . . . . . . . . . . . 91
B. Control Sample. . . . . . . . . . . . . . . . . . . . . 92

LIST OF REFERENCES. . . . . . . . . . . . . . . . . . . . . . . . 93

LIST OF TABLES

Page

12

15
39
41
41

42

Table
1 Examples of R and D Footnote Disclosures Prior
to Statement No. 2 . . . . . . . . . . . . . .
2 Examples of Footnote Disclosures of Changes in
Accounting for Research and Development Costs .
3 Industry Membership of Sample Firms . . . . .
4 1973 Asset Size of Treatment and Control Firms. .
5 Effect of Selection Criteria on Sample Size . . .
6 1974 Research Intensity of Sample Firms . . . .
7 Treatment Portfolio Moving Beta Coefficients
Estimated Using 35 Months of Return Data
and the Market Mbdel. . . . . . . . . . . . . .
8 Control Portfolio Moving Beta Coefficients
Estimated Using 35 Months of Return Data
and the Market Mbdel. . . . . . . . . . . . . .
9 Recursive F-Statistics Computed for the
Treatment and Control Portfolios. . . . . . . .
10 Mbving Beta Coefficients Re-estimated Using
the Constrained Model and 35 Mbnths of
Return Data 0 C O O O O O O O O I O I O O O O O
11 Recursive F-Statistics for the Treatment and
Control Portfolios Beta Coefficients Estimated
Using the Constrained Model . . . . . . . . . .
12 Percentage Changes in the Beta Coefficients
of the Treatment and Control Portfolios . . . .
13 Recursive F-Statistics Computed for the

"Difference Portfolio" . . . . . . . . . . . .

vi

Figure

10

11

12

LIST OF FIGURES

Time Continuum of Critical Events Related
to Statement No. 2 . . . . . . . . . . . . . . . . .

Illustration of the Sequence of F-Tests in
the Harvey Analysis of Covariance Procedure . . . . . .

Hypothetical Beta Behavior if Statement No. 2
Possessed No Information Content. . . . . .

Hypothetical Beta Behavior if Statement No. 2
Possessed Information Content . . . . . . . .

iMoving Portfolio Betas Estimated Using 35 Months
of Return Data and the Market Model . . . . . . . . . .

Ordering of the 77 Monthly Return Observations
into 7 Groups . . . . . . . . . . . . . . . . . . . . .

Monthly Returns of the Treatment Portfolio
Plotted Against the Returns on the Market
for August 1970 to December 1976. . . . . .

Monthly Returns of the Control Portfolio
Plotted Against the Returns on the Market
for August 1970 to December 1976. . . . . .

Moving Portfolio Betas Estimated Using 20 Months
of Return Data and the Market Mbdel . . . . . . . . . .

Percentage Changes in the Beta Coefficients of
the Treatment and Control Portfolios
(Beta Coefficients Estimated Using 35 Months
of Return Data and the Market Model). . . . . . . . . .

Comparison of the Critical Values of the
F-Distribution and the Test Statistics
for the "Difference Portfolio" . . . . . . . . . . . .

"Difference Portfolio" Beta Coefficients Estimated

Using 35 Months of Return Differences and the
Constrained Mbdel . . . . . . . . . . . . . . . . . . .

vii

49

51

51

59

6O

68

69

7O

75

78

79

CHAPTER ONE

INTRODUCTION

In October 1974, Statement of Financial Accounting Standards No.

2, "Accounting for Research and Development Costs," was issued. State-
ment No. 2 established broad guidelines for determining which activities
should be classified as research and development (R and D) and identi-
fied the elements of costs that should be included in R and D. State-
ment No. 2 requires all research and development expenditures as defined
by the statement to be charged to expense when incurred.

In defining the activities and costs that were to be recognized as
part of research and development, the Financial Accounting Standards
Board's (FASB) goal was to achieve a reasonable degree of comparability
among enterprises. Prior to Statement No. 2, considerable variation
existed with respect to the kinds of costs that were identified as
research and development even under substantially similar circumstances.

The Board defined research as:

planned search or critical investigation aimed at
discovery of new knowledge with the hope that such knowl-
edge will be useful in developing a new product or service

... or new process or technique ... or in bringing about a
significant improvement to an existing product or process.

 

1"Statement of Financial Accounting Standards No. 2 - Accounting
:for Research and Development Costs," Journal of Accountancy (December
1974), p. 82.

 

2

Development was defined as the translation of research findings or
other knowledge into a plan or design for a new product or process or
for a significant improvement to an existing product or process. Devel-
opment includes the formulation, design, and testing of product alterna-
tives, the construction of models and prototypes, and the operation of
pilot plants. As the definitions indicate, the FASB made a clear dis-
tinction between the activities that constitute research and those that
are considered to be development.

In determining which method or methods were to be used to account
for research and development costs, the Board considered four alterna-
tives:

(1) charge all costs to expense when incurred;

(2) capitalize all costs when incurred;

(3) capitalize costs when incurred if specified conditions

are fulfilled and charge all other costs to expense;
(4) accumulate all costs in a special category until the
existence of future benefits can be determined.
Prior to Statement No. 2, expensing, capitalization and selective cap-
italization were used extensively to account for research and develop—
ment costs.

The FASB decided against accumulating all R and D costs in a
special category distinct from assets and expenses because it believed
that the use of a special category would change the nature of the basic
financial statements and would complicate the computation of ratios and
other financial data. Selective capitalization was rejected as an
alternative because the Board felt that no set of conditions could be

established for selective capitalization that would be objectively and

3
comparably applied by all enterprises.

Proponents of capitalizing all research and development costs
believed that the accounting treatment for research and development
costs should be determined by considering all of the R and D activities
of an enterprise in the aggregate. They maintained that if there is a
high probability of future benefits from an enterprise's total research
and development program, the entire cost of those activities should be
capitalized regardless of the certainty of future benefits of indi-
vidual projects.

The Board asserted that it is not appropriate to consider accounting
for research and development activities on a total-enterprise basis
because the expectation of future benefits is not evaluated in relation
to broad categories of expenditures but in relation to individual or
related projects. The FASB, therefore, rejected the alternative of
capitilizing all R and D costs.

The FASB favored expensing all research and development costs be-
cause it believed that there was a high degree of uncertainty about the
future economic benefits of research and development projects. The
Board felt that expensing all research and development costs would pro-
vide investors with useful information for assessing the earnings poten-
tial of an enterprise and would better enable them to predict the risk
and return of an investment in the firm.

Statement No. 2 requires all research and development expenditures
to be charged to expense when incurred. Disclosure must be made in the
financial statements of the total amount of research and development
costs charged to expense in each period for which an income statement

is presented. The requirement must be applied retroactively by prior

period adjustment.2

Research Objective

The objective of this study is to assess the information content
of the accounting and reporting requirements of Statement No. 2. Since
the FASB asserted that expensing and disclosing all R and D costs would
provide investors with useful information to predict the risk and return
of an investment in a firm, this study will attempt to determine if
Statement No. 2 provided any information which had an effect on
investors' assessments of the riskiness of firms affected.

The information content of Statement No. 2 will be evaluated by
determining whether or not the systematic risk of firms required to
change their accounting treatment of R and D shifted significantly at
or around the time the statement was issued. If the statement provided
investors with information to assess the risk and return of securities,
a shift in systematic risk would be expected. The research design and
statistical methods employed to assess the information content will be
discussed in Chapter Three.

Gonedes and Dopuch (1974) believe that assessments of effects of
.accounting statements are important because recommendations concerning
laecounting procedures are often justified in terms of the effects of the
prescribed procedures. They state:

Since assertions about effects are important parts

of the justifications offered for recommendations and

prescriptions, we can assess the strength of these

justifications by evaluating the theoretical or empirical

support for the assertions about effects... In short,
assessments of the effects of alternative accounting

21bid., p. 83.

5

procedures and regulations can be useful to accounting

policy-making bodies in making their decisions and to

their constituencies in evaluating those decisions. (P. 80)

Assessing the information content of Statement No. 2 is of interest
not only because providing useful information is the objective of the
statement but because the theoretical arguments used by the FASB to sup-
port its expensing decision have been impugned.3 In addition, respon-
dents to the Discussion Memorandum and Exposure Draft on research and
development expressed concern that the regulation would have an adverse
impact on the ability of firms to raise debt or equity capital for
future research and development projects. A firm's ability to raise
capital may have been affected if Statement No. 2 led to a reevaluation
of the firm's risk. This study may therefore provide some preliminary
evidence as to whether firms' ability to finance future R and D was
affected by Statement No. 2.

The Financial Accounting Standards Board has recently established
a formal review procedure for statements that have been in effect for
at least two years. The initial reviews will cover Statements No. 1
‘through No. 12. The results of this study may provide some input for a
:review of Statement No. 2. If this study finds that expensing research
sand development costs provides information to financial statement users,
:Lt may offer some justification for a practice that many feel is a

1>ackward step in the development of appropriate accounting standards.

3See Bierman and Dukes (1975), Corbin (1975), and Johnson (1976).

6
Criticism of Statement No. 2

The major justification offered by the FASB for expensing research
and development expenditures was what it termed a high degree of uncer-
tainty about the future benefits of individual research and development
projects. The Board cited two studies which found the failure rate of
research and development projects to be very high.4

The FASB has been criticized for its lack of scholarly research in
concluding that the uncertainties of company sponsored R and D are for—
midable (Bierman and Dukes, 1975; Schiff and Fabricant, 1974). A study
by Gee (1971) which classified R and D projects by objective found that
about 70 percent of research and development expenditures are for sup-
port of existing business while the remainder is for exploratory research
and high risk business development. Research and development in support
of existing business is undertaken to maintain and improve the profit-
ability of existing business. Gee found that projects in this category
are selected that have the most immediate effect on whatever economic
criterion is of greatest interest to the corporation. There is rela-
tively low uncertainty about the outcome of these expenditures.

Gee's findings are supported by statistics of the National Science
Foundation which estimated that three-quarters of company sponsored
research and development consisted of development in 1974. Only a
small portion of research and development expenditures was for explora-
‘tory and high risk R and D (Schiff and Fabricant, 1974).

Schiff and Fabricant believe that the 1968 Booz-Allen and Hamilton

eatudy cited by the FASB as evidence of the uncertainty and high failure

4See Booz-Allen and Hamilton, Inc. (1968) and Gerlach and
Vhainwright (1968).

7
rate of R and D projects does not provide such evidence at all. They
argue that if more than half of all R and D expenditures is addIESSEd
to the support of existing business and if the success rate for those
expenditures is over 50 percent (62 percent for all industry groups in
the Booz-Allen and Hamilton study), then it is difficult to accept the
argument that the outcome of R and D expenditures is uncertain.

Mansfield (1969) found that the bulk of R and D projects carried
out by firms in his sample was relatively safe from a technical point
of view. He states that:

Models or policies based on the popular supposi-

tion that the bulk of the research and development in

the industrial laboratory is very risky ... are likely

to be misconceived and misleading. (p. 66)

I appears that the Board's argument of uncertainty of future bene-
fits as a primary reason for requiring the expensing of R and D costs
is unsubstantiated.

In Statement of Financial Accounting Standards No. 2, the Board
cites three studies5 which "failed to find a significant correlation
‘between research and development expenditures and increased future
‘benefits as measured by subsequent sales, earnings, or share of industry
asales." (p. 16) The Board states that because there is no direct or
«even indirect basis for relating costs to revenues, the immediate
recognition principle of expense recognition should apply.

One of the three studies cited to support this conclusion was by
AKlex J. Milburn. Milburn's study was based on an earlier study by

~Iohnson (1967) which tested and failed to find a significant correlation

5See Johnson (1967), Milburn (1971), and Newman (1968).

8

between research and development costs as a percentage of total income
and net income as a percentage of net worth. Using Johnson's sample of
ten major drug companies, Milburn tested the hypothesis that changes

in the ratio of drug manufacturers' research and development expendi-
tures to the total drug industry's research and development expendi-
tures would lead to corresponding changes in the company's future share
of industry sales. Milburn did not find a statistically significant
correlation.

The Board concluded from these studies that "a causal relationship
between expenditures and specific future revenue can seldom be demon-
strated, even with the benefit of hindsight." (p. 16)

In a letter of comment to the FASB's Exposure Draft on accounting
for research and development costs, Milburn questioned whether the
empirical evidence cited by the Board was sufficient to support the
causal relationship conclusion. He pointed out that because the results
of his and Johnson's studies were based on such a small sample of firms,
the correlation would have to have been very high to be statistically
ssignificant. He also indicated that potentially important intervening
\rariables may have obscured observable correlations.

Milburn noted that differences may exist between the relationship
c>f basic research expenditures and development expenditures to observ-
aible future benefits. Because his study and Johnson's study used aggre-
ggated financial data, the differences could not be determined.

Milburn and other respondents to the Exposure Draft expressed dis—
appointment that the Board would conclude that no correlation can be
demonstrated between research and development efforts and future bene-

ffiits given the scant and ambiguous empirical evidence it cited. The

9
studies cited by the Board which were unable to detect a significant
relationship are not sufficient evidence to conclude that such a rela-
tionship does not exist. Bierman and Dukes (1974) adduce studies by
Scherer, Bailey and Angilley which provide support for the hypothesis
that research and development expenditures do produce benefits for the
firm.

The FASB's lack of causal relationship argument does not support
its expensing decision.

The FASB asserted that expensing research and development costs
would provide useful financial information for investment and credit
decisions. It indicated that two of the basic elements in the decision
models of many financial users are risk and expected return. The Board
claimed that capitalization of research and development costs would not
be useful in assessing the earnings potential of an enterprise because
of the high degree of uncertainty associated with those costs. It con-
cluded, therefore, that an investor's ability to predict the return on
an investment and the variability of that return would not be enhanced
‘by capitalization.

Respondents to the Exposure Draft on research and development costs
expressed concern that uniform treatment of all research and develop-
tnent expenditures under all conditions would result in less informative
:reports and less equitable treatment of companies than if more discrimin-
ziting treatment were applied.6 Statement No. 2 makes essentially dif-
ferent activities, higher risk basic research and lower risk develop-

lnent, appear to be alike. Requiring all R and D costs to be expensed,

6See Letters of Comment No. 124 from Ernst and Ernst and No. 153
from Haskins and Sells.

10
regardless of the nature of those costs, may lead to the appearance of
similarities between companies when differences should be emphasized.
Expensing all research and development expenditures may not be useful
in evaluating the risk and return of securities because it ignores
economic realities and treats all R and D expenditures as valueless.

It is interesting that in requiring the expensing of all R and D
costs, the FASB is doing exactly what it had earlier termed inappro-
priate. It is considering accounting for research and development
expenditures on an aggregate basis. The Board rejected the alternative
of capitalizing all R and D costs because it maintained that future
benefits should not be evaluated in relation to broad categories of
expenditures but on an individual or related project basis. By
requiring mandatory expensing, the Board is stating, in effect, that
the future benefits of all R and D projects are highly uncertain. This,
obviously, is not so. It is also interesting that after defining and
making the distinction between research and development, the FASB dis-
regarded the basic differences between the two activities and required
identical treatment.

The immediate expensing of all R and D costs may result in inequit-
able treatment of companies that have discernible future benefits asso-
ciated with their R and D projects, for example, companies undertaking
planned development of products or processes for which a market exists.
The financial positions and results of operations of these companies
will be distorted by expensing outlays that should be treated as
assets. Capitalizing and subsequently amortizing these research and
development expenditures would result in financial reporting that pro-

vides investors with a better understanding of the companies' activities.

11

By denying the option to capitalize, Statement No. 2 suppresses
information that otherwise would be provided. When companies capi-
talized research and development costs prior to Statement No. 2, they
very often disclosed in notes to the financial statements the types of
costs they capitalized and the basis of their decisions to capitalize.
Table 1 presents examples of these footnote disclosures.

Companies with significant amounts of capitalized research and
development costs also disclosed the amounts capitalized and amortized
in the lO-K reports filed annually with the Securities and Exchange
Commission (SEC). For financial statements filed with the SEC on or
after December 31, 1973, any item in the category of "other assets and
deferred charges" that is in excess of 5 percent of total assets must
be disclosed separately. The policy for deferral and amortization
must be disclosed and if the amounts deferred and amortized are
material, they must be presented in schedule form.7

With the issuance of Statement No. 2, the only information about
R and D expenditures that must be disclosed in the financial statements
is the t2£al_amount of research and development costs incurred and
expensed. Some contend that this requirement has lowered rather than
raised the quality and quantity of information available to financial
statement users.

A study by Dukes (1974) which tested the association between
earnings generated by expensing research and development costs and

security prices found that security price behavior is more closely

 

7See Securities and Exchange Commission, Regulation S-X, Rules
5-12, 12-08, and 12-16.

12
TABLE 1

Examples of R and D Footnote Disclosures
Prior to Statement No. 2

 

Example 1:

Product development costs relating to new products and new models
of existing products are deferred and amortized over a period not to
exceed three years commencing with the start of production.

Basic research prior to development is expensed as incurred.

Example 2:

The Company expenses research and development costs as incurred,
except as follows:

Research and development costs applicable to specific product
lines are capitalized only to the extent that they can be realized
from firm orders.

Product costs applicable to audio-visual learning techniques are
capitalized as incurred and amortized as units are sold, within a
maximum period of two years.

Example 3:

The Company defers new research and development costs. Deferred
costs include patent costs, salaries of scientists and engineers, and
certain equipment, materials, and supplies. Other expenses incurred
subsequent to the start of production are charged against income as
incurred. The deferred costs are amortized over a period of five years
following the start of production.

Deferred costs are reviewed semi-annually and upon the occurrence
of any significant event relative to the development program to deter-
‘mine if any write-offs are necessary.

Example 4:

The Company has a program of long-range research and development.
Costs pertaining to this program are deferred and amortized over a
three year period beginning in the succeeding year. Amortization for
.1971 and 1970 amounted to $1,287,760 and $371,200, respectively.

JExample 5:

Research and development costs for new product lines have been
czapitalized for financial statement purposes and are being amortized
c>ver three to five years. These costs are expensed as incurred for

imicome tax reporting.
A schedule of additions and related amortization for the last

three years is as follows:

 

13
related to earnings adjusted for the capitalization of R and D costs
than earnings computed with research and development costs expensed.
Dukes' findings are consistent with the hypothesis that reported
earnings are "systematically adjusted before they are impounded into
security prices, that the adjustment is largely industry specific,
and that the adjustment appears to be directly related to the research
intensity of the industry." (p. 126)

Dukes concluded that accounting policy makers should attempt to
ensure that sufficient data are provided to financial statement users
to allow them to generate their own expectations regarding future pay-
offs of research and development expenditures. In light of Dukes'
findings, the FASB may have provided investors with more relevant data
if they had permitted both capitalization and expensing but had
required disclosure of the total R & D expenditures as well as of the

amounts expensed and capitalized.

Effects of Statement No. 2

This research will attempt to determine if the expensing and dis-
closure requirements of Statement No. 2 provided any information which
had an effect on investors' assessments of the riskiness of firms
affected. Sunder (1973) notes that changes in the market's assessment
(of the risk of a stock may occur due to changes in the economic status
<3f the firm or changes in the information system relating the firm to
investors.

Changing from capitalizing research and development costs to
expensing those costs may have had little effect on the economic

status of firms that have had relatively stable R and D expenditures

14
for a number of years. Earnings would be the same whether the costs
were immediately expensed or capitalized and subsequently amortized.
Cash flows would be unaffected assuming that the R and D expenditures
had been expensed for tax purposes prior to the change.

The firm's balance sheet would be altered since the capitalized
R and D costs would be written off against retained earnings. This,
in turn, would affect the firm's debt-to-equity ratio. Lev (1974)
states that within a given risk class, the higher the firm's leverage
(debt-to-equity ratio), the higher the systematic risk. Therefore, a
change to expensing R and D costs could affect systematic risk depending
upon how significantly the debt to equity ratio was increased.8 Table
2 presents examples of the footnote disclosures of the change in
accounting for R and D.

A change from capitalizing to expensing could seriously affect
the economic status of firms with widely fluctuating research and
development expenditures or young, growth firms with large R and D
expenditures. Mandatory expensing would reduce net income and might
even result in a loss. At the same time, the firm's assets would be
reduced and its debt-to-equity ratio would increase. This could have
an impact on the firm's credit and market standing and affect its

.ability to raise debt or equity capital. With a higher cost of capital

 

8Systematic risk is defined as the covariance of the security's
treturn with the market$s return divided by the variance of the market's
:return, Bim.= cov (Ri,Rm) . The higher the debt-to-equity ratio or

 

02(Rg)
relative share of fixed interest charges, the higher the risk asso-
ciated with the common stocks.

15
TABLE 2

Examples of Footnote Disclosures of Changes in
Accounting for Research and Development Costs

 

Example 1:

In accordance with a recent ruling of the Financial Accounting
Standards Board, research and development expenses were excluded as
an element of cost used in pricing inventory. This has been done on
a retroactive basis and consequently prior years have been restated.
Inventory and taxes have been decreased by $20,853,000 and $9,692,000,
respectively at December 31, 1973. The effect of this change on 1974
and 1973 income was not significant.

Example 2:

During 1974 the Company adopted the policy of charging all product
research and development expenses ... to costs of goods sold as in-
curred. The effect of this change on annual net earnings was not

material.
Financial data for 1973 has been adjusted by the following amounts

and restated to reflect this change in accounting policy.

Earnings before taxes $(l,100,000)

Taxes on earnings (500,000)

Net earnings (600,000)

Net earnings per share (.01)

Retained earnings, January 1 (600,000)
Example 3:

In accordance with the Financial Accounting Standards Board State-
ment No. 2, issued in October 1974, the consolidated statements for
1974 reflect all research and development as expensed when incurred,
and prior years statements have been restated. ...

The cumulative effect of this restatement at 1 January 1974,
after applicable income taxes, was a reduction of $252,145,179 in
earnings retained for growth, which reduced shareholders' equity from
$941,158,160 to $689,012,981. This change in accounting in 1974
decreased research and development expense by $18,794,000 and increased
Inet earnings, after applicable income taxes, by $7,163,984 ($.18 a
:Bhare). Comparable amounts for 1973 were $12,380,591 and $3,781,759
($.10 a share), respectively.

Example 4:

In accordance with a pronouncement of the Financial Accounting
Standards Board, the Company revised its accounting policy in 1974
trlth respect to development costs related to specific production
programs. The consolidated financial statements have been restated
tx: remove TriStar development costs from inventory ($487.8 million at
December 30, 1973), and to charge such costs to earnings in the periods
incurred.

16

TABLE 2
(Continued)

 

The amount of such costs charged to earnings in 1974 was $18 million.
In connection with the change in accounting, gross profit ... which
previously had not been recognized has been recorded retroactively ...
A reconciliation of net earnings (loss) with amounts previously

reported for the four years ended
millions of dollars):

Net earnings (loss) - as
previously reported

Development costs incurred
Gross profit recognized
Related income tax effect

Net earnings (loss) as restated

As a result of the change in

December 30, 1973 follows (in

1973 1972 1971 1970

$16.8 $16.2 $15.4 $(86.3)
(20.7) (52.9) (92.7) (141.6)
20.7 9.9 -- --

1.4 19.6 37.9 40.1

$18.2 $(7.2) $(39.4) $(l87.8)

 

 

 

 

 

 

accounting, consolidated retained

earnings at December 31, 1972 has been reduced by $281 million.

 

l7

and fluctuating returns, the market's assessment of the firm's risk
might change.

Some firms may have to reduce their level of R and D activity as
a result of Statement No. 2 if they are unable to obtain financing for
the projects or their cost of capital increases. The market may per-
ceive these firms as riskier since a decrease in R and D activity may
result in fewer new or improved products or processes being developed.

Even if a firm's economic status were unaffected by Statement No.
2, its systematic risk may have changed because the information system
relating the firm to its environment may have changed. Prior to State-
ment No. 2, firms that capitalized all or part of their R and D expendi-
tures disclosed that accounting policy in notes to the financial state-
ments. The amortization policy was often disclosed, but the amount
capitalized and amortized annually was usually not presented. For
some firms, this information was made more complete by the SEC's dis-
closure requirements concerning the capitalization and amortization of
deferred assets and costs and expenses in excess of one percent of total
sales and revenues. Investors of those firms were able to determine
the total R and D costs and the amounts expensed and capitalized and
thus generate their own expectations concerning the future outcomes
of those expenditures.

Investors of firms that did not disclose the capitalized portion
(If their research and development expenditures may not have been able
to determine the firms' total R and D expenditures unless alternative
sources of information were available. These investors may have found
that the expensing and disclosure requirements of Statement No. 2 pro-

vided a more objective basis for evaluating the future earnings

18
performance of an enterprise. If this occurred, a revaluing of the
firms' stock in line with the stockholders' revised expectations would
have taken place with resultant changes in systematic risk. Adjustments
in beta would not have occurred if Statement No. 2 provided no infor-
mation or information that was already available from alternative
sources.
The formal research hypothesis is stated as follows:
Ho: The systematic risk of firms did not change signifi—
cantly as a result of the accounting and reporting
policies of Statement No. 2.
H1: The systematic risk of firms changed significantly

as a result of the accounting and reporting policies
of Statement No. 2.

CHAPTER TWO
LITERATURE REVIEW

In recent years, a number of market-based studies have been under-
taken to assess the information content of accounting numbers, alterna-
tive accounting procedures, and changes in accounting techniques.
Research has also been conducted to determine if accounting information
is used by the market to estimate the systematic risk of a firm. This
present analysis of the information effect of Statement No. 2 is related
to current studies which have investigated the impact of certain

accounting requirements on the riskiness of firms.

The Information Content of Annual Earnings Numbers

One of the earliest studies of information content was performed
by Ball and Brown (1968). Ball and Brown's objective was to assess
empirically the relative importance of the annual earnings number.
They believed that an observed revision of stock prices associated with
the release of the income report would provide evidence that the infor-
mation reflected in income numbers is useful.

To test for information content, Ball and Brown constructed two
alternative models of the market's expectation of income, a linear

regression model and a naive martingale model.1 Earnings forecast

 

1Amartingalemodel implies that the best estimate of the current
period's income number is the previous period's income number
(E(Incomet) - Income t-l)'

19

20
errors were generated by comparing the estimated earnings with actual
earnings. Firms were grouped into two portfolios, one with positive
forecast errors (actual earnings exceeded expected earnings) and the
other with negative forecast errors. Using the Abnormal Performance
Index,2 monthly abnormal returns were computed for each portfolio for
eleven months prior to the earnings announcements and six months
afterwards.

Ball and Brown hypothesized that if the annual income numbers had
information content, there would be an association between the sign of
the forecast error and the sign of the abnormal return. The results
indicated that a positive relationship existed. When actual income
differed from expected income, the market tended to react in the same
direction.

Ball and Brown concluded that the annual earnings numbers do have
information content, but that most of the information contained in
reported income (85 to 90 percent) is anticipated by the market by the
time the annual report is released. The market uses other sources of
information when they are more timely than annual net income.

Ball and Brown's research is relevant to the current study in that
if the new accounting and reporting policy is found to have no infor-
mation content, it may be because the market is getting information

concerning research and development expenditures from other sources.

 

2The Abnormal Performance Index (API) is a measure of the
contemporaneous association between accounting data and security
prices.

3Alternative sources of information include trade journals,
security analysts' forecasts, industry forecasts, prospectuses,
etc.

21

A study by Foster (1973) provided evidence that the market uses
sources of information other than annual or interim income reports.
Foster found that both individual investors and the aggregate market
consider pre-audited estimates of earnings per share by company offi-
cials to have information content.

The reaction of individual investors to the earnings per share
estimates was evaluated by observing trading volume during the week
announcements of earnings per share estimates were made in the Wall

Street Journal. Foster found that there was a 51 percent increase in

 

the weekly average of the daily percentage of shares traded during the
week the estimates were made relative to the sixteen weeks surrounding
the estimates. There was only a one percent increase in trading volume
of these same firms when the preliminary earnings reports were released
following the estimates of earnings per share.

For a control group of firms for which no officials' estimates of
earnings per share were publicized, there was a 47 percent increase in
trading volume in the preliminary earnings announcement week. Foster
interpreted the results as being consistent with individual investors
perceiving the annual earnings per share number to have information
content but reacting to the earliest source of that number, the company
official's estimate.

To examine the aggregate market reaction to earnings per share

estimates by company officials, Foster used a price analysis similar

 

4A.number of studies have determined that quarterly earnings
announcements possess information content. May (1971) and Kiger
C1972) found significant price changes during the weeks that quar-
terly earnings announc-ents were made.

22

to that employed by Ball and Brown (1968). In assessing the ability of
trading strategies based on those estimates to generate abnormal returns,
Foster determined that there was opportunity to earn abnormal returns
up until the time of the announcements. Once the estimates were made
public, strategies using information in the earnings per share estimates
did not yield abnormal returns.

Foster concluded that estimates of earnings per share have infor-
mation content and that investors rely on the most timely source of

information in setting equilibrium security prices.

The Information Content of Accounting Changes

According to Gonedes and Dopuch (1974), a change in the accounting
techniques used for external reporting purposes can affect capital mar-
ket equilibrium because (1) the change provides accounting numbers that
convey information pertinent to valuing a firm; (2) making the change
has a substantive economic impact independent of the accounting numbers
affected by the change; or (3) the change signals other events that
have economic importance. A number of studies have investigated the
impact of accounting changes on security prices.

Kaplan and Roll (1972) examined the effect that two widely adopted
accounting changes had on stock prices. One change was the switch
from the deferral method to the flow through method of accounting for
investment credits. The other change was from accelerated depreciation
to straight-line depreciation. Neither change had an effect on taxes,
cash, or any other real economic asset or liability of the firm. The
firms simply chose different forms of communicating the same informa-

tion. Kaplan and Roll hypothesized that stock prices would be unaffected

23
by the accounting changes and earnings announcements.

To assess the impact of the accounting changes, weekly abnormal
returns were computed for sixty weeks surrounding the earnings announce-
ments. Kaplan and Roll believed that if a company were able to influ-
ence its stock price by an accounting change, the principal impact
would occur when the announcement was made of the year's earnings.
Firms that increased reported earnings by adopting the flow through
method of accounting for the investment credit experienced positive
average abnormal returns for ten weeks surrounding the earnings
announcements. The increase in security prices was only temporary.
The firms experienced negative average abnormal returns for fifteen
weeks after that.

Kaplan and Roll suggested that investors may have been mislead by
the earnings announcements but revised their expectations about the
firms when they received information about the true reasons for the
reported high earnings in the annual reports. Another explanation
proffered was that quarterly earnings reports issued about thirteen
weeks after the annual earnings announcements may have indicated to
investors that the increased rate of earnings was due to the invest—
ment credit change and could not be maintained.

Kaplan and Roll found that firms that switched from accelerated
to straight-line depreciation appeared to have a small, temporary
increase in abnormal returns around the earnings announcement date.
These results were not statistically significant and the null hypothesis
of no effect could not be rejected.

Gonedes and Dopuch (1974) suggest that the significant results

obtained for the firms changing to the flow through method may have

24
been overstated because a large percentage of the firms switched at
essentially the same time, 1964. This bunching may have caused cross-
sectional correlation among the disturbance terms (estimated using the
market model), understating estimates of their dispersion, and there-
fore overstating the significance of the results. If that were the
case, Kaplan and Roll's estimates would actually be consistent with
the hypothesis that the accounting changes had no effect on stock prices.

Ball (1972) argued that Kaplan and Roll detected a market reaction
to the accounting changes because they used the constant risk version
of the market model. In addition, since a relatively large number of
firms changed to the flow through method at the same time, it may not
have been reasonable for Kaplan and Roll to assume that the accounting
change was independent of the market factor. Ball reasoned that if all
firms changed at the same time, the average return on the firms would
be the market factor, regardless of any effect of the accounting change.
Therefore, use of the market model which removes market effects from
the security returns would not be a valid methodology.

Ball attempted to determine whether the market is mislead by
changes in accounting techniques by analyzing the relationship between
stock prices and a variety of accounting changes over a fourteen year
period. A two-factor cross-sectional model of security returns adjusted
for risk changes was used to test for price adjustments around the dates
of the accounting changes.

Ball believed that firms that change accounting techniques may

not be of constant risk. He argued that the two-factor asset pricing

25
model5 might incorrectly indicate a market adjustment to an accounting
change unless the constant risk assumption were relaxed. He therefore
incorporated risk changes in the two-factor model, estimating risk by
using an ordinary least squares regression on price data for 50 months
prior to and 50 months after the date of the accounting change.

Ball found little abnormal price movement in the nine years before
the accounting changes. There was little unusual price behavior in
the month of the accounting change and in the 19 months after the
accounting change. Ball concluded that changes in accounting techniques
are not associated with market adjustments for the average firm.

Ball's study tested and supported the hypothesis of capital market
efficiency. Gonedes and Dopuch (1974) state that it is possible to
approach Ball's results with the maintained hypothesis of capital market
efficiency. If this is done, his results can be viewed as evidence that
accounting changes have no information content.

Ball expressed surprise that some of the categories of accounting
changes included in his study did not exhibit strong market behavior,
especially changes in inventory valuation from first in, first out
(Fifo) to last in, first out (Lifo). This change would affect taxable
income and, therefore, the after tax cash flow to stockholders. Ball

reasoned that this change did not affect stock prices because at the

 

~ 5Security returns defined by the one-factor or market model,
R1 - “1+8im3m.+ui’ are conditional upon the ex-poste value of the mar-

'ket return. Estimates of a1 and 8m are the intercept and slope of

the linear relationship between R1 and Rm .

The two-factor capital asset pricing model expresses the ex-ante
expected return of a security in terms of the rate of return on a risk-
free asset (R) plus a risk premium which is the risk measure 81m multi-
jplied by the difference between the expected return on the market (Rm )
and the risk free rate; E(R1) - Rf + (E(RIn ) -Rf)81m-

26
same time it reduced tax payments, increases in factor prices may have
reduced cash flow.

Sunder (1973) analyzed the relationship between changes in inven-
tory valuation methods and stock price behavior. His study included
two samples of firms, those firms that changed their inventory valuation
method from Fifo to Lifo and those that changed from Lifo to Fifo.

Sunder assumed that the market is efficient with respect to infor-
mation about accounting changes and tested for information content.

He hypothesized that if the market's expectations about a firm are
formed primarily on the basis of reported earnings, the market price

of firms switching from Fifo to Lifo would decline while the market
price of firms changing from Lifo to Fifo would increase. If the mar-
ket's expectations about a firm are formed on the basis of real economic
value, the market value of firms switching to Lifo should remain un-
changed or increase while the market value of firms switching to Fifo
should remain unchanged or decline.

Sunder used the market model to test the stock price behavior for
twenty-four months surrounding the accounting change. He found that
the average price of the stocks of firms switching to Lifo rose 5 per-
cent higher than that of the market in general during the twelve months
'preceding the accounting change. There was no clear trend in stock
prices during the year following the accounting change.

It appeared that the stock prices of firms switching to Fifo
declined slightly, but this could not be supported statistically
‘because of the small size of the sample (22 firms). Sunder concluded
that the market reacts to information concerning the real economic

value of the firm.

27

In measuring the association between accounting changes and stock
prices, Sunder assumed that the relative risk of the stocks remained
unchanged. He tested this assumption upon completion of his study and
found a significant increase in the relative risk of firms that changed
to Lifo. Sunder concluded that using the market model which assumed
constant risk may have yielded misleading results.

Sunder (1975) replicated his earlier study and attempted to measure
the association between the accounting and price changes by abstracting
the effect of risk changes. The relative risk of each stock was esti-
mated for each of the 24 months surrounding the time of an inventory
accounting change. A maximum of 41 months and a minimum of 25 months
of data were used to estimate the relative risk for each month. The
estimates of relative risk were used to estimate the residuals which
provided a measure of association between the accounting changes and
price changes.

Sunder found that the average risk of firms changing to Lifo was
lower in the pre-change months than in the post-change months. During
the two year study period, the average risk increased by 5.4 percent,
with 78 percent of the total change occurring in the 12 months preceding
the accounting change. For firms that made the accounting change from
Lifo to Fifo, the average relative risk decreased by 5.3 percent.

The average abnormal price change for firms switching to Lifo was
4.7 percent after being adjusted for relative risk and changes in rela-
tive risk. The increase of 4.7 percent during the 12 months preceding
‘the change was only slightly less than the 5.3 percent increase Sunder
found when he used a model assuming constant risk. During the 12 months

following the accounting change, no significant abnormal price changes

28
were observed.

Sunder concluded that changes in the market price of stocks are
associated with the economic value of firms rather than changes in
reported earnings. Investors appear to use information about the
underlying economic realities of firms rather than reported earnings
in forming their expectations about the securities.

None of the studies cited thus far have dealt with nondiscretion-
ary accounting changes. Harrison (1977) investigated the market
effects of discretionary and nondiscretionary accounting changes. He
hypothesized that the signal emitted from the nondiscretionary
accounting change is more likely to be interpreted as information about
the change than about other events in the firm. Discretionary changes
can serve as signals about firms' production-investment decisions,
expectations about cash flows, attempts to smooth income, etc. Deter-
mining the information content of discretionary accounting changes is
more difficult because of the presence of these confounding variables.

Harrison's study examined four separate categories of accounting
changes: discretionary changes with positive effects on net income;
discretionary changes with negative effects on net income; nondiscre-
tionary changes with positive effects on net income; and nondiscretionary
changes with negative effects on net income. To test the hypothesis
that the relative discretion of management and the sign of the net
income effect of an accounting change have information content,
Harrison matched the change firms with nonchange firms of similar
'relative risk and industry membership. The matched firms were dichot-
<mnized into high risk and low risk groups because it had been suggested

by Gonedes (1975) that the market reaction to accounting changes may

29
vary across risk classes.

The two-factor asset pricing model was used to compare the total
returns of the treatment and control groups. Since the expected returns
of the two groups were made equivalent by matching on relative risk, any
differences in realized returns could be attributed to the variables
of interest. The null hypothesis of no effect was tested over a 13
month period surrounding the post-year end month when most firms issued
their preliminary earnings or annual reports.

Harrison found that both discretionary and nondiscretionary
accounting changes that increased net income were associated with unique
stock market behavior. The results also indicated that the discretion
available to management in making the accounting changes possessed
information content.

When the return differences of the two risk classes were plotted,
it appeared that the high risk and low risk classes of firms were
affected differently by the accounting changes. Harrison suggested
that further research needs to be done before any conclusions can be

drawn about the apparent risk class dependency.

Information Content and Changes in Systematic Risk
Collins and Simonds (1977) investigated the information content of
SEC line of business (LOB) disclosures by examining the impact of the
«disclosure requirements on the systematic risk of multirproduct firms.
Using the framework of Rubinstein (1973) which demonstrated the relation-
ship between the operating activities of a multi-product firm and the
systematic risk (beta) parameter specified in the two-factor capital

asset pricing model, they indicated how segmental reporting may be used

30
by investors to assess systematic risk.

Collins and Simonds hypothesized that:

To the extent that the model represents a reason-

able characterization of the critical factors on which

investors base their assessments of the riskiness of

securities and to the extent that the initial disclosure

of segmental data under the SEC LOB program conveys 'new'

(and what is perceived to be reliable) information with

which to assess these parameters, one would expect to

observe changes in the systematic risk parameter of

those firms most affected... (p. 12)

The study examined the effect of LOB requirements on the systematic
risk of three groups of firms: a treatment group that reported little
or no segmental data prior to the enactment of the SEC LOB disclosure
requirements and two control groups. Estimates of the systematic risk
were calculated by using the market model and ordinary least squares
procedures on 40 months of return data, ending with the month of interest.

To test for a shift in systematic risk, Collins and Simonds used
a series of analysis-of-covariance (ANCOVA) procedures. The test pro-
cedures were applied repeatedly to each group during the time period
surrounding the enactment of the SEC segmental data requirements
because the exact location of the possible shift in risk was not known
a priori. The ANCOVA analysis was conducted at both the individual
firm and portfolio level.

At the portfolio level, Collins and Simonds found a significant
shift in the systematic risk of the treatment group during the hypothe-
sized impact period. The control groups experienced no significant risk
changes in the same period. When the ANCOVA beta—change analysis was
‘performed at the individual firm level, the hypothesis that each indi-

‘vidual treatment firm's risk remained constant was rejected. The same

hypothesis was not rejected for the control groups.

31

Collins and Simonds concluded that SEC line of business profit
data do provide useful information to investors for assessing the
riskiness of a firm.

A study by Bar-Yosef and Brown (1977) also tested for information
content by analyzing the variability of systematic risk. Bar-Yosef and
Brown re-examined the information content of stock splits which had
originally been evaluated by Fama, Fisher, Jensen and Roll (FFJR)(1969).
The methodology of FFJR assumed that the systematic risk of securities
is unaltered during the period surrounding a split.

Bar-Yosef and Brown reasoned that although it is unlikely that
stocks undergo a permanent change in risk as a result of stock splits
since neither the firm's financial or operating characteristics are
altered, it is possible that the risk is unstable around this time.
Since stock splits are often preceded by temporally high earnings and
followed by increased total dividends, the splits may coincide with
market expectations of increased cash dividends. Bar-Yosef and Brown
hypothesized that the split announcement may reduce uncertainty regard-
ing the permanence of the higher earnings and signal to investors that
higher total cash dividends are forthcoming.

The information content of stock splits was evaluated by examining
inoving betas for 108 months surrounding the split month. Betas were
estimated using the "symmetric method" and 60 months of return data.6

Firms that increased total cash dividends after the split were

 

6The "symmetric method" measures systematic risk in month t by
Irtilizing data from.months t_ through t_1 and from months t t rough
t . Collins and Simonds (1977) used the "leading method" which
calculates beta in month to by using return data from months t_n to to.
Either method is appropriate.

32
evaluated separately from firms that decreased total cash dividends
after the split.

The analysis showed that the risk of both groups became temporarily
larger during the period surrounding the split. The systematic risk
rapidly returned to its pre-split level for firms increasing dividends
but remained high for firms decreasing dividends.

Bar-Yosef and Brown concluded that firms which split their shares
after experiencing temporally high earnings reduce investor uncertainty
about future earnings and cash dividends. The split announcement in—
creases the uncertainty for investors of firms which have pre-split
earnings that indicate no future increases in dividends.

Bar-Yosef and Brown state:

Our finding that systematic risk increases prior

to a split suggests that moving beta techniques should

be considered when the information content of other

financial events is examined. (p. 1080)

Since past research has shown that moving beta analysis is an
effective method to test for information content, it will be used in

this research to examine the impact of a change in accounting for

research and development expenditures.

CHAPTER THREE
RESEARCH DESIGN AND METHODOLOGY

The effect of FASB Statement No. 2 on the systematic risk of firms
can be evaluated by assessing the relative stability of the beta coef-
ficient around the time of the accounting change. A time period must
be identified within which the market risk adjustment would most likely
have taken place if Statement No. 2 provided investors with new infor-
mation.

In April 1973, the FASB placed on its technical agenda a project
on accounting for research and development and similar costs. The
Board recognized that research and development expenditures constitute
a significant element of the United States economy1 and was concerned
about the number of alternative techniques used to account for and
report those costs. The Discussion Memorandum was issued on December
28, 1973, and a public hearing on the subject was held on March 14, 1974.

In deliberations following the hearing, the Board concluded that
the Statement of Financial Accounting Standards under consideration
should only address accounting for research and development costs and
'not similar costs such as start-up, marketing research and promotion

costs, and costs of training new personnel. An Exposure Draft of the

 

1In 1973, total research and development expenditures were over
.30 billion dollars. Approximately two—thirds of all research and
development was conducted by business enterprises.

33

34
proposed statement was issued on June 5, 1974, and the statement was
adapted in October 1974. Statement No. 2 was effective for fiscal
years beginning on or after January 1, 1975, although earlier applica-
tion was encouraged.

Since the market's reaction to published accounting information
can be anticipatory, the earliest point at which a firm's systematic
risk would be expected to shift would be at the time the Exposure Draft
was issued in June 1974.2 An adjustment in risk would not be expected
at the time the Discussion Memorandum was issued because the FASB did
not advocate any particular accounting alternative in the Discussion
Memorandum. It simply presented the arguments for and against each
alternative and reviewed the current authoritative literature pertinent
to accounting for research and development and similar costs. Since
the FASB's decision to require immediate expensing was not known until
the Exposure Draft was issued, a market reaction to the proposed
accounting change would not be expected before that time.

Data sheets obtained from Accountipngrends and Techniques indi-
cated that many of the firms that deferred research and development
costs during 1973 switched to expensing those costs in 1974. Firms

that complied with Statement No. 2 at the end of 1974 were selected

for this study because they would be among the first firms to experience

 

2Ball and Brown (1968) found that much of the information con-
tnained in annual accounting reports is anticipated through alternative
snources of information. It is not unreasonable to assume that once
time requirements of Statement No. 2 were known by the market, infor-
mation about R and D costs was develOped and leaked to the market
thirough analysts'reports based on conversations with corporate manage—
fluent or corporate management's comments published in the press.

35

risk changes if Statement No. 2 had an information effect.3 Since this
study will only analyze those firms that changed to expensing at or
near the end of 1974, the latest potential beta shift point would be
March 1975. Financial statements for the year-end are normally issued
within three months.

The relevant time period in which a market risk adjustment would
have taken place if Statement No. 2 provided investors with new infor-
mation would be from June 1974 to March 1975. Whether or not the
systematic risk of firms was affected by the expensing and disclosure
requirements will be determined by the analysis. Figure 1 presents a

time continuum of events related to the issuance of Statement No. 2

Sample Selection

To test for the effect of Statement No. 2 on systematic risk, this
study examined firms that provided incomplete information about research
and development costs in their annual reports and lO-K reports prior to
the statement. Firms were classified as providing incomplete informa-
tion if they did not disclose the amount of research and development
costs capitalized and amortized either in footnote or schedule form.
The amount of research and development costs expensed may or may not
have been disclosed. Because these firms did not provide information
about the capitalized research and development costs, investors would
be unable to make adjustments to the reported earnings to determine

what the earnings would have been if the firms had expensed all R and D

 

3Since the potential impact period was defined as from the time of
the Exposure Draft (June 1974) to the time the annual reports were pub-
lished, only firms with November or December 1974 year-ends were in-
cluded in the study. This criterion shortened the potential impact period.

36

A.:.a was

 

as moaned
may no
mwcqumoz
usapsmv
Avozoou Awesome Awesome
N .02 uuouv aboaouosux
unassumumv ousooexuv cosmosouwnv
chad mums chma «sad «sad nnma. mean csad
.009 .mn: .oon .uoo case .009 scan .w:<
i. l a w L. a + a t.
nu on mm an no so He mm H such:

.

canoe amuse m Hmaucouomlllllm

N .02 unassumum cu vaumaum
moco>m noowuuuu no assauuaoo mash

H NMbuHh

37
expenditures. Therefore, one would expect these firms to be most
affected by the accounting and reporting requirements of Statement No. 2.

Data sheets obtained from Accountipngrends and Techniques were
used to identify those firms that capitalized all or part of their
research and development expenditures prior to the issuance of State-
ment No. 2 and changed to expensing those costs at or near the end of
1974 (November 30 to December 31, 1974). An AICPA computer print-out
of footnote disclosures concerning R and D accounting changes was used
to supplement the list.

The 1973 lO-K reports of the firms identified as having changed
their treatment of R and D costs were reviewed to determine the amount
of information disclosed about the capitalized R and D expenditures.
Those firms that provided complete information, as defined above, were
eliminated from.the sample. A shift in systematic risk would not be
expected for those firms because investors had sufficient data to assess
the firms' risk and return under both the expensing and capitalizing
alternatives prior to Statement No. 2.

Firms from the populations of both the New York Stock Exchange
(NYSE) and American Stock Exchange (ASE) were included in the sample.
The NYSE firms were required to have monthly stock price returns avail-
able on the Center for Research in Security Prices (CRSP) tape for the
77 month period from August 1970 to December 1976. The ASE firms were
required to have complete data available on stock prices, dividends,
and capital changes for the same period.

An initial sampling criterion was that firms that had other
accounting changes at the time they complied with Statement No. 2 could

not be included in the sample. However, during the sampling procedure

38
it was determined that 11 of the treatment firms (24 percent) had
changed their inventory cost flow assumptions from Fifo to Lifo at the
end of 1974.4 In order to maintain an adequate sample size, the cri-
terion was changed to eliminate those firms that had accounting changes
other than a change from Fifo to Lifo at the same time they revised
their treatment of R and D costs.

A control group consisting of firms that expensed and disclosed
all research and development costs prior to Statement No. 2 was matched
to the treatment group. A shift in the systematic risk of the control
group firms would not be expected since their accounting and reporting
of R and D costs did not change as a result of Statement No. 2.

The treatment and control firms were matched on the basis of secu-
rity exchange and industry membership. Since market risk is determined
in part by industry membership, a disproportionate industry membership
in either the treatment or control group could cause risk adjustment
differences between the two groups if the inherent riskiness of those
industries changed during the test period. Matching the firms by 3-
digit Standard Industrial Classification (SIC) code controlled for a
factor that could have affected the treatment and control groups in a
systematic manner. The major industry groupings for each of the samples
is presented in Table 3.

As can be seen from Table 3, firms from the metal, machinery, and
electronics industries comprise a large percentage (approximately 55

percent) of the treatment and control firms. Firms in the chemical and

 

4Twelve of the matched control firms (27 percent) changed their
accounting policies from Fifo to Lifo at the end of 1974.

39
TABLE 3

Industry Membership of Sample Firms

 

 

 

3-Digit SIC Number of

_Code Industry Name Matched Firms
100 Metals - Miscellaneous 2
131 Oil - Crude Producers l
240 Forest Products 1
284 Cosmetics 2
300 Tire and Rubber Goods 1
322 Containers - Metal and Glass 2
324 Building Materials - Cement 2
335 Rolling and Drawing Non-Ferrous Metals 1
343 Bldg. Materials - Heat, Air Cond., and Plumb. 1
349 Fabricated Metal Products 2
352 Farm and Garden Machinery 1
353 Machinery - Oil Well Equipment 1
354 Machine Tools 1
356 Machinery - Industrial 1
357 Electronic Computer Equipment 2
363 Electric Household Appliances 1
366 Radio and TV Transmitting Equipment 2
367 Electronics 1
371 Auto Parts and Accessories 3
372 Aircraft and Parts 3
376 Guided Missiles and Space Vehicles 1
381 Engineering, Laboratory and Research

Equipment 2
391 Silverware and Plateware l
394 Toys and Amusement - Sporting Goods 1
481 Telephone Companies 1
492 Natural Gas Transmission 1
506 Electrical Apparatus and Equipment 1
737 Computer Data Processing 1

Number of Matched Firms in the Treatment and Control Groups - 40 per

group.

4O

pharmaceutical industries, which carry out a great deal of basic
research, are not represented in the sample. Firms in these industries
almost exclusively expensed all research and development costs.

It was also determined during the sample selection that very few
firms on either the New York Stock Exchange or American Stock Exchange
capitalized all of their research and development expenditures. The
majority selectively capitalized their R and D costs. The fact that
firms carrying out risky, basic research expensed those costs and that
few firms capitalized all R and D costs might indicate that firms
attempted to accurately represent the results of their R and D activi-
ties prior to Statement No. 2.

The firms in the treatment and control groups were matched as much
as possible on the basis of their 1973 total asset size. Table 4 pre-
sents the asset size distribution of the treatment and control groups.
The effect of the selection and matching criteria on the sample size is
explained in Table 5.

Table 6 presents the research intensity of the firms in the treat-
ment and control groups which was defined as the ratio of total research
and development expenditures to total sales and revenue. Although the
treatment and control firms were not matched on the basis of research
intensity, the research intensities of the firms in the treatment and
control groups are quite similar. Table 6 indicates that 18 of the
treatment firms (45 percent) and 19 of the control firms (47.5 percent)
spent less than 1 percent of total sales and revenue on research and
development in 1974. Only 2 firms in the treatment group and 2 firms
in the control group had R and D expenditures that were greater than 5

percent of total sales and revenue. All of those firms were in the com-

;nater equipment and computer data processing industries.

41

TABLE 4

1973 Asset Size of Treatment and Control Firms

 

 

 

 

Asset Size Treatment Control
(000 deleted) Firms Firms
$25,000 or less 6
$25,001 - 50,000 7
$50,001 — 200,000 9 6
$200,001 - 500,000 6 10
$500,001 - 1,000,000 4 4
Over $1,000,000 ‘_8 ._1
4_o m
TABLE 5
Effect of Selection Criteria on Sample Size
Firms with 11/30 to 12/31 year ends that changed their
R and D accounting policy in 1974* 68
Less: Firms with other accounting changes in 1974 ._g
59
Less: Firms with incomplete data on prices, dividends,
and capital changes 19'
49
Less: Firms that could not be matched according to
3-digit SIC code ___
Number of firms in treatment sample 49_

 

*Firms had incomplete data about R and D costs prior to compliance with
Statement No. 2.

42

TABLE 6

1974 Research Intensity of Sample Firms*

 

 

Research Intensity Treatment Control
(Percent) Firms Firms
< 1.0 18 19
1.0 - 1.9 9 9
2.0 - 2.9 7 8
3.0 - 3.9 2 l
4.0 - 4.9 2 l
5.0 and greater _2_ __2
:9 .29.

 

*Research intensity is defined as the ratio of total research and
development expenditures to total sales and revenue.

Estimates of Systematic Risk

To determine whether there was a significant shift in systematic
risk associated with the accounting and disclosure requirements of
Statement No. 2, betas were estimated for each group for each month
from June 1973 to December 1976. Although the potential impact period
‘was defined as June 1974 to March 1975, betas were estimated for a
longer interval to evaluate their behavior prior to and after any
adjustment as a result of Statement No. 2.

To more precisely estimate systematic risk, the individual securi-
‘ties of each group were combined into equally weighted portfolios. The
precision with which the beta coefficient can be estimated is inversely
related to the magnitude of unsystematic risk. Unsystematic risk can

In: substantially reduced by grouping individual securities in portfolios.

43
Collins and Simonds (1977) determined that a test for adjustments in
systematic risk is best conducted at the portfolio level. Firm level
beta changes are difficult to detect.

A limitation of the portfolio level analysis is that it does not
provide information about the beta behavior of individual firms. If
the firms in the portfolio undergo changes in their relative risk that
are not of the same algebraic sign, the shift in the portfolio beta
could be small or non-existent while the betas of some individual firms
could have changed appreciably. Since the relative risk of the treat-
ment firms would be expected to increase if Statement No. 2 provided
information, it was decided that the beta shift analysis should be con-
ducted at the portfolio level.

The beta coefficients for each month were estimated using ordinary
least squares (OLS) regression procedures and the market model on
return data for a thirty-five month period ending with the month of
interest. For example, the beta coefficient for June 1973 was esti-
mated using return data from August 1970 through June 1973.

The market model is of the form:

it
it

m
oi + Bimfimt + uit eq. (1)

where

it = return of portfolio 1 in period t

W?
ll

mt return on the market portfolio in period t

i’ B = intercept and slope of the linear relationship

N 0.:
between Rit and Rmt

8 portion of portfolio i's return that varies

C?

it w
independently of Rmt'

44
The monthly return of each portfolio is the average of the monthly
returns of all the securities in that portfolio. The return on each

security is defined as:

’\.

R11: ' [(Pit + Dit)/Pi,t-1] '1 eq' (2)
where

Pit = price of security i at end of month t

Dit = dividends per share paid on security i in month t.

The return on the market is computed using the Standard and Poor's
(S&P) Composite Index. This index of common stock performance is mar—
ket value weighted, that is, the price of each stock is weighted by the
number of shares outstanding. The S&P Composite Index includes 500 of
the largest stocks in the United States in terms of stock market value.

The monthly returns for the market are in the form of:

m

Rmt = [kPmt + Dmt)/Pm,t-1] -1 eq. (3)
where

’L

Rmt is the common stock total return during month t

mt is the value of the S&P Composite Index at the

end of month t

Dmt is the estimated dividends received during month t
and reinvested at the end of month t.

Data Sources
Monthly stock price relatives for the NYSE firms were generated
:from the CRSP tape. Monthly closing stock prices for the ASE firms

were collected from the ISL DailLStock Price Record and Barron's.

45
Data on cash dividends, stock dividends, and stock splits were gathered
from the ISL Daily Stock Price Record and Standard and Poor's Annual

Dividend Record.

 

Statistical Testing Procedures

An analysis of covariance procedure developed by Harvey (1976) was
used to test the stability of each portfolio's systematic risk during
the period surrounding the adoption of Statement No. 2. If the exact
location of a possible shift in beta had been known a priori, a single
ANCOVA or Chow test could have been applied to each portfolio at the
hypothesized shift point.5 Since a single potential shift point could
not be identified, a series of Chow tests developed by Harvey was
applied repeatedly from June 1973 through December 1976. This period
encompassed the hypothesized shift interval of June 1974 to March 1975.

The test procedure repeatedly divides the overall time series of
monthly returns into two adjacent sub-periods. Recursive residuals
calculated for each of the sub-periods are then compared to determine
if a significant shift in beta has occurred.

The following regression model is used to derive the recursive

residuals:
yj = xj' B + uj, j = l,...,n eq. (4)

‘where

x. is a k x l vestor of fixed observations on the independent

variables - xjl

sz

5The ANCOVA procedure is frequently referred to as a CHOW test

‘because of the extensions Chow (1960) made to the ANCOVA procedure.

If

46

B is a k x 1 vector of coefficients = 81

B2

yj is the jth observation on the dependent variable (the

return on the portfolio)

u is a disturbance term.

In its expanded form the model is:

y. = le Bl +'sz 82 + uj.

J
le = 1, then
y3. = (1) Bl + sz 82 + uj eq. (5)

= a +-xj2 82 +-uj.

Equation (5) is equivalent to the market model explained earlier.

The n-k recursive residuals are defined as:

Y3 ' xi bJ-l

 

' ' _1 5 , j=k+l,...,n eq. (6)
[1 + xj (Xj_1 Xj_1) xj]

where

‘4
u.
II

0‘
I

the jth return on the portfolio

x 8 a k x 1 vector of fixed observations on the

independent variables = l .1

1:
ml

the estimates of the coefficients obtained

from the first j observations 8 Fa

LB-

 

X - a j x k matrix of full rank consisting of the first j

observations on the independent variables.

47

The denominator of the equation depends only on the observations
on the independent variables and is introduced to standardize the recur-
sive residuals so that under the null hypothesis the variance is 02.

The recursive residuals are normally and independently distributed with
a mean of zero under the null hypothesis.

To analyze the constancy of the regression parameters, the n
observations generated by Equation 4 are ordered according to time and
allocated into p mutually exclusive groups, with m1 (>k) observations
in the first group, m2 observations in the second group, etc. Let

ni = m1+m2+...m for i=1,...,p and set no=k.

i,
The p—l test statistics are:

 

n
(i=n 2 +1 sz) / “1
F‘ = 1'1 , i = 2,...,p. eq. (7)
i ( 21-1 U 2) / n. -k
j=k+l j 1'1

The test statistics have F-distributions with mi and ni_fdc degrees
of freedom and are statistically independent under the null hypothesis.

In this study, 75 recursive residuals were computed from 77 obser-
vations on the portfolio returns and market index and the least squares
estimates of beta. The 77 observations were ordered according to time

in 7 groups with 13 observations in the first group, 11 observations in

each of the next 5 groups and 9 observations in the last group.

 

6The number of groups and observations per group is arbitrary. The

77 observations were ordered into groups of 11 so that the recursive
residuals for the impact period of June 1974 to March 1975 would be in-
cluded in one group. The program used to calculate the recursive
residuals requires the first group to contain m+k observations. There-
famathe first group contains 13 observations and the last group contains
9 observations.

48

Eleven recursive residuals were calculated for each of the first six
groups and 9 recursive residuals were calculated for the last group.

The test procedure first compares the sum of the first 11 recur-
sive residuals to the sum of the second 11 recursive residuals and come
putes a test statistic. The sum of the first 22 recursive residuals
is then compared to the sum of the next 11 recursive residuals and a
test statistic is again computed. This procedure is repeated until the
sum of the first 66 recursive residuals has been compared to the sum
of the last 9 recursive residuals.

To determine whether the regression coefficients were constant
over the entire interval, each of the test statistics is compared to
an F-statistic with m1 and n1_1-k degrees of freedom. A significant
test statistic indicates that there was a shift in the beta coefficient
between periods ni_1 and mi. Figure 2 summarizes the sequence of tests.

An assumption of the recursive F-test is that if there is a sig-
nificant test result, then the shift in beta occurred between the last
observation (month) in the interval n and the first observation in

i-l

the subsequent interval, m The test procedure has the greatest

1.
power when the shift occurs between those two points. However, if a
shift in beta occurs within an interval, a significant test may also
result. Therefore, the recursive F-tests identify the sub-period in
which the beta coefficients changed but do not identify the specific
month of the shift. The beta coefficients must be plotted in order
to determine the specific month of a significant shift.

If Statement No. 2 provided no new information to assess a firm's

systematic risk, the beta coefficient of the treatment portfolio (as

well as of the control portfolio) would be relatively stable throughout

49

ousooooum

 

 

 

 

ummu Him
I A VA
aomooum «0
assumes xIHIw: one we Sues ousomooum ammo new I A VA
TH s A TH c A V
Somoouu mo moouwoo xi : mam E sues ouaooooua umou uma I n: H+ w m m: H+xuh
d HIHG
fl: NEHn—FN: a" a a
I . _ _ _ _ _
as «a ma NE xAHE

Awesouw A com“ mouoouo mcoaum>uomno 55v
monouw a coma vouoouo mcoeum>uomoo :

 

musmoooum mocmwum>oo mo mfimhama< ho>umm
ago as mumoolm mo ooaosoom onu mo soaumuumSHHH

N MMDOHh

50
the hypothesized impact period. If Statement No. 2 had information
content, the beta coefficient of the treatment group would be expected
to increase in a month within the impact period. Figures 3 and 4
illustrate (hypothetically) the behavior of the treatment portfolio

beta in each situation.

Difference Analysis

The risk stability of each portfolio was evaluated separately
using the recursive F-tests. The results of the separate testsindicate
if the systematic risk of each portfolio was stable during the hypothe-
sized impact period but do not determine if there were significant
differences in the risk behaviors of the two groups. If the accounting
and reporting policies of Statement No. 2 provided new information,
significant differences in the risk behaviors of the two portfolios
would be expected during the designated impact period.

To assess whether the risk behaviors of the treatment and control
portfolios were dissimilar, monthly differences in the returns of the

two portfolios were calculated. The differences were defined as:

fiDt ' i"It - RC: eq. (8)
where

in.

fiTt 3 the return on the treatment portfolio in month t

it

Ct

the return difference in month t

the return on the control portfolio in month t.

A constrained model based on the Sharpe-Lintner capital asset
pricing model (CAPM)‘was used to regress the return differences against

the market return to estimate monthly beta coefficients for the

51

FIGURE 3
Hypothetical Beta Behavior if Statement No. 2
Possessed No Information Content*
imt

1.2

1.1

1.0mm

 

 

 

 

0.9
0.8
35 47 57 67 77 ”nth
FIGURE 4
Hypothetical Beta Behavior if Statement No. 2
Possessed Information Content**
Bimt
1.2
LL Arv‘v‘:vA.
1.0V““‘ ‘;‘~;"/”’—7~
0.9
0.8
month

 

 

35 47 57 67 77

*For illustration purposes, the beta of the portfolio is assumed
to be 1.0.

**For illustration purposes, the beta of the portfolio is assumed
to be 1.0 before the shift due to Statement No. 2.

52
"difference portfolio." If the relative risk of the treatment port-
folio did not shift more significantly than the relative risk of the
control portfolio, the beta coefficient of the "difference portfolio"
would be relatively stable.7

The constrained model is of the form:

'U W
Rjt I RFt + 8(Rmt I RFt) I uj

’b
or Rjt - RH - “imt - RFt) + uj eq. (9)
where
jt = the return on portfolio j in period t

the risk free rate of return in period t
(the rate of return on U.S. Treasury bills)

Rn;

mt the return on the market portfolio in period t

Cc’

the dist rbance term which is independent of
RF and
t mt.

If the return differences from equation (8) are substituted for

Rjt’ equation (9) becomes:
W N 'b ’b
(RT: I RFt)I(RCt I RFt) I BT(Rmt I RFt)I BC(lfimt I RFt) + (“T I “0)
'U W
(IT: I RCt) I (ET I BGuam: I RFt) I (“T I "6)

RD I 8D (kmt I RFt) I 3D eq' (10)

 

7The results of the difference analysis will determine if the
risk behaviors of the two groups were different and will indicate the
sub-period in which they differed, but they will not reveal whether
the risk of one or both groups changed. If the risk of only one
portfolio shifted, the analysis cannot identify the "shift" portfolio.
Therefore to determine if Statement No. 2 affected the risk of the
treatment group, separate analyses as well as the difference analysis
‘were conducted.

53
The recursive F-tests described earlier were used to evaluate the
constancy of the "difference portfolio" beta. The constrained model
was used to estimate the betas so that any significance in the recur-
sive F—tests would be due to changes in beta only.
Recall that the recursive residuals are computed using the model

yj = x ' B + u , j-l,...,n, where B is a k x 1 vector of coefficients.

:1 i
If the constrained model is not used, a significant recursive F-test
may be the result of changes in 81 (the intercept or a in the market
model) or 82 (the systematic risk). Since a difference in the risk
behavior of the treatment and control portfolios was defined as a shift
in the beta coefficient (systematic risk) of the "difference portfolio,"
it is important that the recursive residuals be calculated in a way
that only shifts in the systematic risk affect the F-tests. This can
be accomplished by using the constrained model in which the intercept
of the linear relationship between the security and market returns is
defined as the risk-free rate of return. Since the risk-free rate of
return in any period is the same for both the treatment and control
portfolios, this factor drops out of the equation for the return dif-
ferences (eq. 10). Therefore, the recursive residuals are computed
with only one coefficient, the systematic risk of the "difference
portfolio," and the recursive F-tests are affected by changes in the
systematic risk only.

The n-k recursive residuals for the "difference portfolio" are

defined as:

54

y - X 'b -
j = 3' 1 114 A: , j=k+1,...,n eq. (11)
[1 + xj (xj_1 xj_1) xj]

I38

 

where
y = the monthly return difference,
3 m 3 RT R
RDt t Ct
the return on the market minus the risk free rate of return,
'b
(Rmt I RFt)

j the least squares estimate of the "difference portfolio"
beta obtained from the first j observations

X
I

OI
I

X = a vector of the first j observations on the independent
variable, xj.

The null hypothesis of the difference analysis is stated as

follows:

H ' There was no significant change in the beta coefficient

0 of the "difference portfolio" during the potential
impact period.
H1: There was a significant change in the beta coefficient

of the "difference portfolio" during the potential
impact period.

Since the potential impact period was defined as the 10 month
interval from June 1974 to March 1975, the risk stability of the
"difference portfolio" was tested at several points within that inter-
‘val as well as at numerous points outside the impact period. The 77
return observations were ordered into 26 groups with 4 observations in

the first group, 3 observations in each of the next 24 groups, and l

55
observation in the last group.8 A shift in the beta coefficient of the
"difference portfolio" during the impact period indicates that there
were differences in the risk behaviors of the treatment and control

groups as a result of Statement No. 2.

 

81a the separate analyses, the return observations were ordered
into 7 groups so that the systematic risk of each portfolio during
the impact period could be compared to the risk before and after the
impact period. The analyses could also have been conducted with more
groups and fewer observations in each group in which case the risk
stability of each portfolio would also have been evaluated at one or
‘more points within the impact period.

CHAPTER FOUR

RESULTS OF THE ANALYSIS

This chapter presents data on the analysis of the risk behaviors
of the treatment and control groups during the period of June 1973 to
December 1976. The results of the analysis indicate that both the
treatment and control groups experienced significant shifts in risk
during the hypothesized impact period of June 1974 to March 1975.

There were no differences in the risk changes of the treatment and
control groups. Firms that altered their accounting and reporting
policies as a result of FASB Statement No. 2 experienced shifts in risk
that were not statistically different than the risk changes of firms
that did not change their accounting and reporting methods. These

results suggest that Statement No. 2 had no information effect.

Test for Constancy of the Risk Parameters

The moving beta estimates calculated for the treatment and control
groups are presented in Tables 7 and 8 and are plotted in Figure 5.
Inspection of Table 7 indicates that there were dramatic decreases
in the risk of the treatment group in months 47 and 48. The beta
coefficient decreased by 23 percent between months 46 and 47 and by 29
percent between months 47 and 48. The risk of the treatment group
decreased until it reached a minimum at month 53. There was then a

.28 percent increase in risk between months 53 and 54.

56

57
TABLE 7

Treatment Portfolio Moving Beta Coefficients Estimated
Using 35 Months of Return Data and the Market Model

 

 

 

Observation
First Last Month Beta Coefficient
1 35 June 1973 1.8133
2 36 1.8777
3 37 1.7587
4 38 1.7510
5 39 1.8640
6 40 1.8538
7 41 December 1973 1.7161
8 42 1.6563
9 43 1.6371
10 44 1.6592
11 45 1.6480
12 46 1.6583
13 47 June 1974 1.2721
14 48 .8983
15 49 .9179
16 50 .8581
17 51 October 1974 .8281
18 52 .7583
19 53 December 1974 .7309
20 54 .9385
21 55 .9518
22 56 March 1975 .9599
23 57 .9674
24 58 .9959
25 59 1.0013
26 60 .9926
27 61 .9956
28 62 1.0034
29 63 .9898
30 64 .9925
31 65 December 1975 .9896
32 66 1.1009
33 67 1.0952
34 68 1.0853
35 69 1.0775
36 70 1.0756
37 71 1.0556
38 72 1.0529
39 73 1.0451
40 74 1.0437
41 75 .9753
42 76 .9805

43 77 December 1976 1.0015

58

TABLE 8

Control Portfolio Moving Beta Coefficients Estimated
Using 35 Months of Return Data and the Market Model

 

 

 

Observation
First Last Month Beta Coefficient
1 35 June 1973 1.5673
2 36 1.6246
3 37 1.5442
4 38 l 5588
5 39 1.6253
6 40 1.6535
7 41 December 1973 1.5432
8 42 1.5048
9 43 1 5241
10 44 1 5131
11 45 1.4835
12 46 1.4958
13 47 June 1974 1.1969
14 48 .8195
15 49 .8114
16 50 .7848
17 51 October 1974 .7407
18 52 .6955
19 53 December 1974 .6817
20 54 .8454
21 55 .8366
22 56 March 1975 .8442
23 57 .8725
24 58 .8860
25 59 .8818
26 60 .8723
27 61 .8746
28 62 .8772
29 63 .8662
30 64 .8646
31 65 December 1975 .8605
32 66 .9226
33 67 .9177
34 68 .9076
35 69 .9001
36 70 .8973
37 71 .8808
38 72 .8787
39 73 .8675
40 74 .8660
41 75 .7877
42 76 .7887

43 77 December 1976 .8079

 

59

FIGURE 5

Moving Portfolio Betas Estimated Using 35 Months
of Return Data and the Market Model

 

 

 

or?

O
°.
s
u o o
G 'H w-l
0.4 .4.4
E o o o g
UIH H'H -
o u u o ~g
o H o u
H O O O
Eds. can. a
9
"S
O
°.
.N
p
3.
.0
(D
0
°.
,v
D
8.
'8
(D
3.:
'3';
O
I:
0
°.
.01
ID
0
9
Pa
'
3.
..Q
Q
0
‘3
*3
O
‘3
-8
3.
*8
3.
02-2 on“: 00’: as”: or“: oz“: 00’: oe’b os’o ov°d‘
6139 OI'TOJIUOd

60

The control group also experienced substantial decreases in risk
in months 47 and 48. The beta coefficient of the control group de-
creased by 20 percent between months 46 and 47 and by 32 percent
between months 47 and 48. The risk continued to drop until month 53.
Between months 53 and 54, there was a 24 percent increase in the risk
of the control group.

The shifts in the risk of the treatment and control groups occurred
during the designated impact period of June 1974 to March 1975. The
Exposure Draft was issued in month 47 (June 1974) and the year-ends
of the firms in the treatment and control groups were in month 53
(December 1974).1

To analyze the observed beta changes in a formal manner, the
recursive F—tests developed by Harvey and described earlier were
applied to the treatment and control groups. The 77 monthly observa-
tions of the portfolio returns were ordered into 7 groups with 13
observations in the first group, 11 observations in each of the next
5 groups, and 9 observations in the last group. Figure 6 illustrates

the ordering of the monthly observations.

FIGURE 6

Ordering of the 77 Monthly Return Observations
into 7 Groups

1-13 1 14-24 ,1 25-35 I 36-46 L 47-57 LA58-68 . 69-77
Months Aug.'70- Sept.'71- Aug.'72- July '73- June '74- May '75- Apr.'76-
IAug.'7l July '72 June'73 May '74 Apr. '75 Mar.'76 Dec.'76

 

 

1Two NYSE firms in the treatment group have 11/30 year-ends.
All other firms in the treatment group and all firms in the control
group have December year-ends.

61

Seventy-five recursive residuals were calculated from the 77
observations. The recursive F-statistieacomputed by comparing the sums
of squared recursive residuals for the seven groups are presented in
Table 9. The treatment and control groups were analyzed separately.

Analysis of Table 9 indicates that a significant shift in beta
occurred between the first 46 months and the subsequent 11 months for
the treatment group. The computed F-statistic is 3.328 which is sta-
tistically significant at an u-level of .01. The results of the recur-
sive F-tests are consistent with the graph of the treatment group betas
(Figure 5) which exhibited a sizable decrease in beta between months 46
and 47 and months 47 and 48 and an increase between months 53 and 54.
Inspection of Table 9 and Figure 5 reveals that the risk of the treat-
ment group was relatively stable prior to month 47 and after month 53.

The recursive F-tests indicate that the control group also experi-
enced a significant shift in risk between the first 46 months and the
succeeding 11 months. The computed F-statistic is 4.188 which is
satistically significant at an o-level of .01. A shift in the beta
coefficient of the control group between the first 35 months and the
following 11 months is also indicated at an o-level of .05. The come
puted F-statistic is 2.724. This result is not consistent with the
plot of the control group risk (Figure 5) which appears to be rela-
tively stable between months 35 and 46. The beta coefficient of the
control group is 1.5673 at month 35 and varies between 1.6535 and
1.4835 from.months 36 to 46.

The risk behavior of the control group can be evaluated prior to
month 35 if the beta coefficient is estimated using 20 months of

return data. Figure 9 (page 70) plots the beta coefficient of the

62

TABLE 9

Recursive F-Statistics Computed for the
Treatment and Control Portfolios

 

 

 

2 Compared to Degrees of
Zdj for months: 2312 for months: F-value Freedom
Treatment Group:

1-13 14-24 .487 (11,11)
1-24 25-35 .347 (11,22)
1-35 36-46 1.938 (11,33)
1-46 47-57 3.328* (11,44)
1-57 58-68 1.101 (11,55)
1-68 69-77 .054 (9,66)
Control Group:
1-13 14-24 .407 (11,11)
1-24 25-35 .250 (11,22)
1-35 36-46 2.724** (11,33)
1-46 47-57 4.188* (11,44)
1-57 58-68 .688 (11,55)
1-68 69-77 .251 (9,66)

 

* - significant at an a-level of .01

** - significant at an a-level of .05

63
control group for months 20 to 77. Examination of Figure 9 reveals
that the control portfolio beta decreased from approximately 1.57 in
month 20 to 1.37 in month 26. The beta coefficient was relatively
stable between months 26 and 31 but increased to approximately the
month 20 level (1.56) in month 35 and to 1.68 in month 36.

The significant recursive F—test indicated at an a-level of .05
was apparently caused by the increase in beta in month 36. The beta
coefficient was stable between months 36 and 46. Since the shift in
beta occurred prior to the designated impact period (and in the con-
trol portfolio), it is unlikely that it is related to the issuance
of Statement No. 2.

As explained in Chapter 3, the results of the recursive F-tests
may be influenced by changes in alpha (the intercept of the linear
relationship between the portfolio and market returns) as well as by
changes in the beta coefficient because of the model used to calcu-
late the recursive residuals. It is doubtful that changes in alpha
(if any) significantly affected the results of the recursive F-tests
since the beta coefficients of the treatment and control groups
decreased so dramatically between months 46 and 48 (46 percent and 45
percent, respectively). Nonetheless, to be assured that the signifi-
cant test results were not due to changes in alpha, the beta coeffi-
cients of the treatment and control portfolios were re-estimated using
the constrained model and the recursive F-statistics were recomputed.2

The re-estimated beta coefficients are presented in Table 10.

 

2The beta coefficients were estimated using 35 months of return
data. To conduct the recursive F—tests, the 77 observations were
ordered into groups of 11 as in the first series of F-tests.

64

TABLE 10

Moving Beta Coefficients Re-estimated Using the
Constrained Model and 35 MOnths of Return Data

 

Beta Coefficient

 

 

Month Treatment Group Control Group
June 1973 1.7249 1.4854
1.7982 1.5507
1.6938 1.4872
1.6893 1.5048
1.8180 1.5840
1.8393 1.6378
December 1973 1.7122 1.5350
1.6457 1.4926
1.6329 1.5111
1.6614 1.4979
1.6443 1.4624
1.6596 1.4737
June 1974 1.2755 1.1924
.9243 .8241
.9504 .8165
.8881 .7886
October 1974 .8439 .7437
.7885 .7092
December 1974 .7843 .7088
.9538 .8465
.9618 .8383
March 1975 .9671 .8431
.9719 .8668
.9945 .8799
.9986 .8766
.9889 .8639
.9923 .8661
.9987 .8667
.9871 .8576
.9886 .8567
December 1975 .9848 .8514
1.0939 .9162
1.0873 .9104
1.0831 .9056
1.0756 .8984
1.0734 .8954
1.0547 .8803
1.0542 .8803
1.0439 .8670
1.0443 .8672
.9851 .7972
.9894 .7979
December 1976 1.0154 .8227

 

65

The results of the second series of recursive F-tests were identi-
cal to the results of the first series for both the treatment and con-
trol groups. The recursive F-statistics are presented in Table 11.

Analysis of Table 11 indicates that both the treatment and control
groups experienced significant shifts in beta between the first 46
months and the following 11 months. The computed F-statistics of 3.263
(treatment group) and 3.980 (control group) are statistically signifi-
cant at an a-level of .01. A shift in the control group beta between
the first 35 months and the succeeding 11 months was also indicated at
an a-level of .05. The F-statistic is 2.368. Therefore, the treatment
and control portfolios have experienced significant changes in systematic

risk during the designated impact period.

Analysis of Observed Beta Decline

It appears that since both the treatment and control groups expe-
rienced significant shifts in relative risk between June 1974 and
January 1975, the changes in the beta were not related to the issuance
of Statement No. 2. The relative risk of the control group would not
‘have changed as a result of the statement because the firms in the
control group expensed and disclosed their research and development
expenditures prior to Statement No. 2. If the expensing and disclosure
requirements had provided investors with new information to evaluate
thezrisk of firms, an increase (not a decrease) in the beta coefficient
(1f the treatment group would be expected because of increased leverage
and, in some cases, decreased net income.

The beta estimates used in this risk analysis were determined

tflirough ordinary least squares procedures. According to Johnston

66

TABLE 11

Recursive F-Statistics for the Treatment and Control Portfolios
Beta Coefficients Estimated Using the Constrained Model

 

 

 

2 Cpmpared to Degrees of
ZUj for months: Zdj for months: F-value Freedom
Treatment Group:

1-13 14-24 .501 (11,11)
1-24 25-35 .486 (11,22)
1-35 36-46 1.765 (11,33)
1-46 47-57 3.263* (11,44)
1-57 58-68 1.126 (11,55)
1-68 69-77 .064 (10,66)
Control Group:
1-13 14-24 .427 (11,11)
1-24 25-35 .396 (11,22)
1-35 36-46 2.368** (11,33)
1-46 47-57 3.980* (11,44)
1-57 58-68 .825 (11,55)
1-68 69-77 250 (10,66)

 

* - significant at an a-level of .01

** - significant at an a-level of .05

67
(1972), these estimates can be dramatically affected by extreme out-
lier observations. If a small number of abnormal returns occurred
near the time of the shift in beta, the observed decline would not
reflect a true change in relative risk but would reflect the fact that
several abnormal observations did not adhere to the otherwise stable
relationship between the portfolio and market returns.

To determine if extreme outlier observations caused the very sig-
nificant shifts in beta, the returns of the treatment and control
portfolios were plotted against the returns of the market portfolio
for months 1 through 77 (Figures 7 and 8). It appears from inspection
of Figure 7 that the outlier observations at months 47 and 48 and
‘months 50 and 51 were responsible for causing the changes in beta for
the treatment group. The observations at months 47 and 48 seem to
have caused the beta coefficient to drop sharply at those points
‘while the outlier observations at months 50 and 51 led to the substan-
tial increase in beta that followed. Figure 8 reveals that the control
group exhibited the same pattern of outlier observations.

If the beta coefficients are estimated using a moving regression
Llength of 20 months, it becomes apparent that the significant decline
111 beta is not the result of an altered relationship between the
portfolio and market returns. The betas estimated using 20 months of
portfolio and market data are plotted in Figure 9. Figure 9 reveals
that the betas of the treatment and control groups return to approxi-
mately their pre-month 47 levels when the intervals used for estima-
tion do not include the observations at months 47, 48, 50, and 51.

The betas estimated for month 71 are the first betas calculated

over 20 month intervals that did not include the outlier observations.

68
FIGURE 7
Monthly Returns of the Treatment Portfolio

Plotted Against the Returns on the Market
for August 1970 to December 1976

 

 

AA
A
O
‘s
A
a
‘ n
a a
a
5..
A
.5
5..
... ...
. h
52 so
. q .
4“‘. (,7
A 6 g A
45..
48 figs a
‘ g 46
a0 “3
. . a
a
49 A .
‘ A
ID
“
D
s
v
Q
a II
I f r '1
- .o.13 .08 0.18 0.24

-b.oa 0100 o
HRRKET RETURNS

0.40

0.40

0.32 I

RETURNS
0:18 0:24

89LIO

0.00 P0311

1

j

69
FIGURE 8
Monthly Returns of the Control Portfolio

Plotted Against the Returns on the Market
for August 1970 to December 1976

 

 

O
a A
a ’9 “
23
i“
A
'
9
‘ -O-18 -b.aa 0100 nice 0113 0124

HRRKET RETURNS

Moving Portfolio Betas Estimated Using 20 Months
of Return Data and the Market Model

Treatment
Portfolio

70

FIGURE 9

Control

Portfolio

I

1E.00

r2

2k.00 3b.00 3E.00 43.00 43.00 ab.00 sk.oo 60.00 6E.00
MONTHS

20.00

 

ov’z

oo’z

09’:

OT

9°: or“: oz“:
8138 01103180d

on“

0030

V
08°C

09'

 

71
The increase in beta to the pre-month 47 level was not evident when
the moving regression length was 35 months because the outlier observa-
tions had not yet dropped out of the beta estimation interval and
therefore caused the beta coefficient to remain at a lower level.

Examination of Figures 7 and 8 reveals that the returns on the
treatment and control portfolios were quite stable from month 44 to
month 50. The returns on the treatment portfolio ranged from minus 3
percent to minus 10 percent. The control group returns ranged from 1
percent to minus 8 percent during the same period. The returns on the
market portfolio were very volatile in this 7 month period. The mar-
ket return was minus 2.7 percent in month 46, 10 percent in month 47,
*minus 17 percent in month 48, and minus 8 percent in month 49. This
temporary divergence between the study groups' and the market's be-
havior caused the outlier observations and led to the significant
decline in beta.

In month 51, the returns of the treatment and control portfolios
and the market portfolio all increased significantly. The market
return increased from minus 12 percent in month 50 to 17 percent in
month 51. The returns on the treatment and control portfolios in-
creased from minus 8 percent to 11 percent and minus 8 percent to 10
percent, respectively. The substantial increase in the returns of the
market portfolio and the treatment and control portfolios appear to
have led to the increase in beta in month 54.

According to reports in Barron's, the market experienced serious
declines throughout 1974. Factors contributing to the decline included
rising interest rates, a high rate of inflation, real declines in the

GNP for two quarters, and concern over the growing power of the Arab

72
oil block to drive the economy into a depression.

The prime rate rose steadily during the first half of 1974, reach-
ing nearly 12 percent at the end of May. The market experienced a
downward trend throughout this period. Investment confidence was re-
stored temporarily in June when the prime rate was reduced slightly
and predictions were published that the rate would be reduced to a range
of 7 to 8 percent by the end of 1974.

The increase in market activity was short-lived. The market fell
steeply in July when the prime rate was increased further and the
Labor Department disclosed that the cost of living had risen at an
annual rate of 12 percent in June. At the end of September the stock
market plummeted to a 12 year low. In October 1974, the stock market
finally recovered. This was due in part to overall good third quarter
earnings reports.

The reason that the treatment and control portfolios did not
exhibit price behaviors that were as volatile as the market's may be
that the treatment and control portfolios are not very diversified.
As indicated in Chapter 3, approximately 55 percent of the firms in the
matched treatment and control groups are in the metal, machinery and
electronics industries. It is possible that the price behaviors of the
predominant industries in the treatment and control portfolios were not
as volatile as the price behavior of the market in general during the
hypothesized impact period. The returns on the study portfolios
would not, therefore, be as volatile as the returns on the market
since 55 percent of the study firms are in those industries.

According to Standard and Poor's Trade and Security Statistics,

Securitngrice Index Record, the composite market index decreased by

 

73
approximately 8 percent between June and July 1974 (months 47 and 48).
The price indices of the metal fabricating and miscellaneous metal
industries decreased by approximately 6 percent and 3 percent, respec-
tively, between months 47 and 48 while the price index of the machinery
industry dropped by 7 percent during the same time. It is very likely
that differences in the industry composition of the market portfolio
and the treatment and control portfolios led to the disparity in the
groups'behaviors and therefore the observed decline in the risks of
the treatment and control groups.
Comparison of the Treatment and Control
Group Risk Changes

The recursive F-tests verified that the relative risks of the
treatment and control groups changed significantly between June 1974
and January 1975. The F-tests were applied to the treatment and con-
trol groups separately. To assess whether the relative risks of the
treatment and control portfolios behaved differently between months 35
and 77 (the overall study period), monthly changes in the beta of each
group were computed. The monthly percentage changes are presented in
Table 12 and graphed in Figure 10. It appears from inspection of the
table and graph that the risk changes of the two groups were very
similar over the entire interval of June 1973 to December 1976.

To empirically evaluate whether the risk changes of the treatment
and control portfolios were significantly different, recursive F-tests
'were applied to the "difference portfolio" betas as explained in
Chapter 3. The observations were ordered into groups of 3 so that the
risk changes of the treatment and control portfolios could be compared

at several points within the impact period of June 1974 to March 1975.

74
TABLE 12

Percentage Changes in the Beta Coefficients of
the Treatment and Control Portfolios

 

Treatment Group Control Group Differences in

Between

 

h.

Percentage Percentage Percentage
Months Change Change Change
35-36 .0355 .0365 -.0010
36-37 -.0634 -.0494 -.0140
37-38 -.0044 .0095 -.0139
38-39 .0645 .0427 .0218
39-40 -.0055 .0174 .0229
40-41 -.0743 -.0667 -.0086
41-42 -.0348 -.0249 -.0099
42-43 -.0116 .0128 -.0244
43-44 .0135 -.0072 .0207
44-45 -.0068 -.0196 .0128
45-46 .0062 .0083 -.0021
46-47 -.2329 -.1998 -.0431
47-48 -.2938 -.3153 .0215
48-49 .0218 -.0099 .0317
49-50 -.0650 -.0328 .0327
50-51 -.0351 -.0562 .0211
51-52 -.0843 -.0610 .0233
52-53 -.0361 -.0198 -.0163
53-54 .2840 .2401 .0439
54-55 .0140 -.0104 .0246
55-56 .0085 .0091 -.0006
56-57 .0078 .0335 -.0257
57-58 .0295 .0156 .0139
58-59 .0054 -.0047 .0101
59-60 -.0087 -.0109 .0022
60-61 .0030 .0026 .0004
61-62 .0078 .0030 .0048
62-63 -.0136 -.0125 -.0011
63-64 .0027 -.0018 .0045
64-65 -.0029 -.0047 .0018
65-66 .1125 .0722 .0403
66-67 -.0052 -.0053 .0001
67-68 -.0090 -.0110 .0020
68-69 -.0072 -.0083 .0011
69-70 -.0018 -.0031 .0013
70-71 -.0186 -.0184 -.0002
71-72 -.0026 -.0024 -.0002
72-73 -.0074 -.0127 .0053
73-74 -.0013 -.0017 .0004
74-75 -.0655 -.0904 .0249
75-76 .0053 .0013 .0040
76-77 .0214 .0243 -.0029

75
FIGURE 10
Percentage Changes in the Beta Coefficients of the Treatment

and Control Portfolios (Beta Coefficients Estimated
Using 35 Months of Return Data and the Market Model)

i
E

Treatment Portfolio
Control Portfolio

l

 

7

15.00 11.00 013.00

05.00

93.00

31.00

 

 

m-‘o u-‘o or‘o III-'0 ”30‘3”: “pg-'3 ”Unfifsaudwo- oz-o- 33°0-

arm"

76
The recursive F-tests were applied at months 47, 50, 53, and 56 within
the impact period as well as at numerous points outside the impact
period.

Table 13 presents the test statistics for the recursive F-tests.
Analysis of Table 13 discloses that there were 2 significant F-tests
at an a-level of .05. A significant test statistic (4.512) resulted
when the sum of the squared recursive residuals from the first 40
observations was compared to the sum of the squared recursive residuals
for months 41 through 43. A second significant test statistic (2.947)
resulted from the comparison of the sum of the squared recursive
residuals for the first 64 observations to the sum of the squared
recursive residuals for months 65 through 67. Both of these test
statistics were outside the impact period of June 1974 to March 1975.
Therefore, the risk changes of the treatment and control groups in
months 47, 48, and 54 were not significantly different from one another.

Figure 11 plots the series of test statistics from the recursive
F-tests and the F-distribution at an a-level of .05 for 3 and ni_1-1
degrees of freedom. Figure 11 illustrates that the test statistics
were well below the significance level except at months 41 and 65
[(3,39) and (3,63) degrees of freedom].

The beta coefficients of the "difference portfolio" are plotted
in Figure 12.3 Inspection of Figure 12 reveals that there were no
substantial increases or decreases in beta between months 40 and 43

or between months 64 and 67. The "difference portfolio" beta was

 

3The beta coefficients were estimated using 35 months of "return
difference" data.

77
TABLE 13

Recursive F-Statistics Computed for the
"Difference Portfolio"

 

Compared to Degrees of

NZ

 

Xuj for months Zuj2 for months: F-value ‘Freedom
1-4 5-7 .277 (3,3)
1-7 8-10 1.190 (3,6)
1-10 11-13 .597 (3,9)
1-13 14-16 .201 (3,12)
1-16 17-19 2.138 (3,15)
1-19 20-22 .451 (3,18)
1-22 23-25 .568 (3,21)
1-25 26-28 .523 (3,24)
1-28 29-31 1.793 (3,27)
1-31 32-34 .226 (3,30)
1-34 35-37 .033 (3,33)
1-37 38-40 .635 (3,36)
1-40 41-43 4.512* (3,39)
1-43 44-46 .758 (3,42)
1-46 47-49 1.343 (3,45)
1-49 50-52 1.745 (3,48)
1-52 53-55 1.371 (3,51)
1-55 56-58 2.646 (3,54)
1-58 59-61 .362 (3,57)
1-61 62-64 .159 (3,60)
1-64 65-67 2.947* (3,63)
1-67 68-70 .136 (3,66)
1-70 71-73 .547 (3,69)
1-73 74-76 .387 (3.72)
1—76 77 .233 (1,75)

 

* - significant at an a-level of .05.

78

FIGURE 11

Comparison of the Critical Values of the
F-Distribution and the Test Statistics

for the "Difference Portfolio"

r. a..
mofiumaumu
umoe

:OfiHOwuhom
oudouowwan:

mOofldu
aosuansuumsoua 0;”
so mmaaaslemosusnu

1....

guesses

....c muuzsowa

«in

:P

 

a £51151

aft

Y
”'0

00"

 

79
FIGURE 12

"Difference Portfolio" Beta Coefficients Estimated Using
35 Months of Return Differences and the Constrained Model

14nIIWfiifiI

EEEFIIFEmI

0'4 .00

55.00

 

ch.

“.00

o— s—o—o—o—HWM
W \AV’. V ....-H-O—O-
00 C.

 

if};

b--—
D.”

 

007 I 00-0 00-0 ’0 02-1: 00-0 02-0- 09-01“
use 011041304 305335510

80

relatively stable between months 35 and 77.

It should be noted that when a large number of tests are conducted
(25 tests in this case) and a few tests are significant, the signifi-
cant tests could be the result of chance. Therefore, the significant
test statistics at month 41 and 65 may not indicate true differences
in the risk behaviorsof the treatment and control groups in those
months. Examination of the percentage risk changes in Table 12 reveals
that the beta coefficienusof the treatment and control groups decreased
by 7.4 percent and 6.7 percent, respectively, between months 40 and 41.
Between months 64 and 65, the beta of the treatment group decreased by
.29 percent while the beta coefficient of the control group decreased
by .47 percent. The risk changes of the treatment and control groups
appear to be very similar in those two months.

To determine whether the two significant tests occurred by chance,
a binomial test was conducted. The probability of obtaining x signifi-

cant tests by chance when N tests are conducted is given by:

N— e . 11
p(x)= qu" q()
x
where
P = alpha level
Q = l-P
N N
___.— '
x3(N-x)! , . denotes factorial.

81
The results of the binomial test indicate that the probability
that exactly 2 significant tests occurred by chance is 23 percent.4
Therefore, it is very likely that there are no real differences in the

risk behaviors of the treatment and control groups in months 41 and 65.

 

IThe probability that 2 or less significant tests occurred by
chance is 59.5 percent, [p(2) + p(1)].

CHAPTER FIVE

SUMMARY AND CONCLUSIONS

The objective of this study was to investigate whether Statement
of Financial Accounting Standards No. 2 provided investors with infor-
mation to assess the riskiness of firms. Statement No. 2 requires all
research and development expenditures as defined by the statement to
be charged to expense and to be disclosed in the financial statements
in each period for which an income statement is presented. One of the
Financial Accounting Standards Board's goals in issuing Statement No. 2
was to provide useful financial information about research and develop-
ment costs. The Board believed that the capitalization of any research
and development expenditures was not useful in assessing the earnings
potential of an enterprise and therefore would not enhance the ability
of investors to predict the risk and return on an investment. Many
have criticized the FASB's expensing decision and have questioned
whether this accounting and reporting policy provides investors with
relevant information to assess the risk and return of firms undertaking
research and development projects.

The information content of Statement No. 2 was assessed by testing
the stability of the systematic risk of firms that changed their
accounting and reporting policies to comply with Statement No. 2. It
was hypothesized that if Statement No. 2 provided new information to
evaluate the riskiness of a security, a shift in the security's beta

82

83
would occur. A sample of firms that changed their R and D accounting
policies at the end of 1974 was matched to a sample of firms that
expensed and disclosed all research and development expenditures prior
to the issuance of Statement No. 2. No adjustment in systematic risk
would be expected for firms that had expensed all R and D costs prior
to Statement No. 2 because the accounting and reporting policies of
those firms did not change.

Monthly beta coefficients were estimated for each of the samples
which were treated as equally weighted portfolios of securities.
Recursive F-tests were performed separately on each portfolio to
determine whether there was a shift in the portfolio's systematic
risk. It was hypothesized that if Statement No 2 possessed information
content, a shift in the relative risk of the treatment group (change
firms) would occur at some point between June 1974, the time the Expo-
sure Draft was issued, and March 1975.

The results of the recursive F-tests indicated that the beta
coefficients of the treatment and control groups shifted significantly
between June 1974 and March 1975. Analysis of the plots of the beta
coefficients revealed that both the treatment and control groups
experienced significant declines in risk in June 1974 and again in
July 1974. The betas of both groups increased significantly in
January 1975.

To determine whether there were significant differences in the
beta shifts of the treatment and control groups, a "difference port-
folio" was created which consisted of monthly differences in the
returns of the treatment and control portfolios. Monthly beta coef-

ficients were estimated for the "difference portfolio." It was

84

hypothesized that if the shifts in the relative risks of the treatment
and control groups were significantly different, the beta coefficient
of the "difference portfolio" would not be stable during the impact
period of June 1974 to March 1975.

The recursive F-tests conducted on the "difference portfolio" indi-
cated that there were no significant differences in the beta shifts of
the two groups. Examination of the plots of the monthly returns on
the treatment and control portfolios and the market suggested that the
significant shifts in the relative risks of both groups resulted from

extreme outlier observations around the time of the shifts.

Conclusions

The results of this analysis suggest that Statement No. 2 possessed
no information content. Although the beta coefficient of the treatment
group was not constant, the instability of the beta coefficient does
not reflect a true change in relative risk as a result of Statement
No. 2. This conclusion is supported by the fact that the control group
experienced beta changes in the same months which were not statistically
different than those of the treatment group.

It.appears that the market viewed the changes necessitated by State-
ment No. 2 as mere bookkeeping variations having no real economic sub-
stance. Although Statement No. 2 resulted in all research and develop-
ment expenditures being treated as having no future value, the market's
expectations of the future payoffs of firms' research and development
expenditures apparently did not change. The market's perception of the
riskiness of firms formerly capitalizing some or all of their research

and development costs was not altered by Statement No. 2.

85

The FASB's contention that the capitalization of any research and
development expenditures would not be useful in assessing the risk and
return on an investment was not supported by the results of this study.
It appears that either capitalization of R and D costs conveyed suffi-
cient information to evaluate the risk of a security or that the market
obtained information about firms' R and D expenditures from alternative
accounting and non-accounting sources. The accounting and reporting
policies of Statement No. 2 may have had no effect on investors' per-
ceptions of systematic risk because they provided redundant information.

One of the concerns of opponents of Statement No. 2 was that the
regulation would have an adverse impact on the ability of firms to
raise capital for future research and development projects. The fact
that the relative risk of the treatment group did not increase as a
result of Statement No. 2 may provide preliminary evidence that firms

were not adversely affected by the regulation.

Implications of the Results

The Financial Accounting Standards Board has established a formal
review procedure for statements that have been in effect for at least
two years. The initial reviews will include Statement No. 2. One of
the Board's primary justifications for expensing research and develop-
ment costs was that this accounting alternative would provide relevant
information for investment and credit decisions. A tentative conclu-
sion of this study is that Statement No. 2 provided no new information
to assess the riskiness of firms.

Many have argued that the decision to expense all research and

development expenditures was not based on sound accounting theory.

86

Some have cautioned that Statement No. 2 creates questionable new
accounting theory with potentially wide application. If the degree of
uncertainty of future benefits is a valid criterion for expensing
research and development costs (it was argued in Chapter 1 that it is
not valid for development expenditures), then the criterion could be
extended to many investments, especially investments in long-lived
plant and equipment in fields where there are rapid technological
advances.
The FASB's treatment of research and development expenditures
appears to be in direct contrast to its definition of an asset in the
proposed conceptual framework for accounting. In the Exposure Draft,
"Objectives of Financial Reporting and Elements of Financial Statements
of Business Enterprises," the FASB states:
A resource other than cash must have three character-
istics to be represented as an asset in the financial
statements of an enterprise: (a) the resource must,
singly or in combination with other resources, con-
tribute directly or indirectly to future cash inflows
(or to obviating future cash outflows), (b) the enter-
prise must be able to obtain the benefit from it, and
(c) the transaction or event giving rise to the enter-
prise's right or interest in the benefit must already
have occurred. (page 21)

Many development expenditures which Statement No. 2 now requires to be

expensed would be considered assets under this definition.

The FASB contended that selective capitalization of research and
development expenditures was not a viable accounting alternative because
no set of conditions could be established for capitalization that could
be objectively and comparably applied by all enterprises. The Inter-

national Accounting Standards Committee (IASC) has recently issued

International Accounting Standard 9, "Accounting for Research and

87

Development Activities." The IASC states that the allocation of
research and development costs to accounting periods is determined by
their relationship to the expected future benefits to be derived from
the R and D activities. In cases where there is little direct rela-
tionship between the amount of research and development costs and
future benefits because the amount of such benefits is too uncertain,
the IASC requires immediate expensing. When the product or process
is technically or commercially feasible and the enterprise has adequate
resources to enable the product or process to be marketed, the costs
of development activities may be deferred to future periods.

The IASC established the following criteria which must be satis-

fied for development costs to be deferred:

(a) the product or process is clearly defined and the
costs attributable to the product or process can
be separately identified;

(b) the technical feasibility of the product or process
has been demonstrated;

(c) the management of the enterprise has indicated its
intention to produce and market, or use, the product
or process;

(d) there is a clear indication of a future market for the
product or process or, if it is to be used internally
rather than sold, its usefulness to the enterprise can
be demonstrated; and

(e) adequate resources exist, or are reasonably expected
to be available, to complete the project and market

the product or process.

88

International Accounting Standard 9 requires the disclosure of
the total research and development costs expensed in each period,
including the amortization of deferred development costs, the balance
of unamortized development costs, the additions to the deferred develop-
ment cost account, and the basis for amortization. The comprehensive
disclosure requirement of IAS 9 will enable investors to make adjusting
calculations to the reported earnings numbers to compare firms that
have expensed all research and development costs with firms that have
selectively capitalized development expenditures.

It appears that in light of the results of this study, the criti-
cism of Statement No. 2, and the proposed conceptual framework for
accounting, the FASB should reconsider its position on accounting for
research and development costs. International Accounting Standard 9

may be a useful model for the Board to consider.

Limitations of the Study

The effect of Statement No. 2 on investors' perceptions of risk
was evaluated at the portfolio level because past research indicated
that firm level beta analysis lacks statistical power. A limitation
of the portfolio level analysis is that it is not capable of providing
information which may be disaggregated to the individual firm level.
It is possible, therefore, that the relative risksof some treatment
firms increased as a result of Statement No. 2 although the beta
coefficientsof the portfolios decreased significantly.

The firms in this study were drawn from the New York Stock Exchange
and the American Stock Exchange and were not randomly sampled. The

results of this analysis, therefore, cannot be generalized beyond the

89
firms in the sample. Logical but not statistical inferences can be
made about the effect of Statement No. 2 on other firms to the extent
that those firms are similar to the firms in the study.

One of the sampling criteria of this study was that a firm must
have changed its method of accounting for research and development
costs at the end of 1974. (Although Statement No. 2 was not effective
until fiscal years beginning on or after January 1, 1975, the FASB
encouraged earlier application.) This sampling criterion may have
biased the results of the study. Firms that complied with Statement
No. 2 before the effective date may have done so because their earnings
or financial positions were less affected by the statement. These
firms, for example, may have capitalized fewer of their research and
development expenditures and therefore reduced their assets by less
as a result of Statement No. 2. Firms that delayed making the
accounting change until 1975 or later may have capitalized a greater
percentage of their R and D expenditures. The financial positions and
financial ratios of these firms may have been more adversely affected
by the statement due to larger asset write-offs. Although the results
of this study indicate that Statement No. 2 had no information content,
it is possible that Statement No. 2 provided information about the
risk of firms that postponed making the accounting change until 1975

or later.

Suggestions for Future Research
Further research may overcome some of the limitations of this cur-
rent study. An extension of this study would be to evaluate the risk

behavior of firms that delayed changing their R and D accounting

9O
policies until fiscal years beginning on or after January 1, 1975.

Statement No. 2 may have provided useful information to assess the
riskiness of those firms.

No firms whose equity securities are traded over the counter (OTC)
were included in this study. Past research has shown that the OTC mar-
ket relies heavily on the information in annual reports to make invest-
ment decisions due to the lack of more timely sources of information.1
An analysis of the risk stability of OTC firms that changed their R and D
accounting policies may indicate whether Statement No. 2 provided useful
financial information to OTC investors.

The information content of Statement No. 2 could also be investi-
gated with a different research design. The information content could
be measured by determining if abnormal stock returns, adjusted for
relative risk and changes in relative risk, were associated with the
accounting change. Either the market model or the two-factor model

could be used to estimate the price residuals.

 

1See Grant (1977).

APPENDIX A

Treatment Sample

Aeroflow Dynamics

Aeronca Incorporated

Alaska Interstate Corporation
American Hoist and Derrick
Applied Data Research Inc.
Arvin Industries

Austral Oil Company

Avco Corporation

Avon Products

Belden Corporation

Big Three Industries

Braun Engineering Company
Cetec Corporation

Cyprus Mines Inc.

Del Laboratories Inc.

The Goodyear Tire and Rubber Company
Hanna Mining Company

Hasbro Industries Inc.
Hazeltine Corporation

Hoffman Electronics Corporation
Honeywell Inc.

Insilco Corporation

Lockheed Aircraft Corporation
Lone-Star Industries

Martin Marietta Corporation
McDonnell Douglas Corporation
Napco Industries Inc.
National Can Company

P & F Industries

Puerto Rican Cement Company
Raymond Precision Industries
Roblin Industries

The Signal Companies
Sundstrand Corporation
Superscope Inc.

United Industrial Corporation
UOP

western Union Corporation
weyerhaeuser Company

White Consolidated Industries Inc.

91

Stock Exchangg

 

ASE

ASE

NYSE
NYSE
ASE

NYSE
ASE

NYSE
NYSE
NYSE
NYSE
ASE

ASE

NYSE
ASE

NYSE
NYSE
ASE

NYSE
NYSE
NYSE
NYSE
NYSE
NYSE
NYSE
NYSE
ASE

NYSE
ASE

NYSE
ASE

ASE

NYSE
NYSE
NYSE
NYSE
NYSE
NYSE
NYSE
NYSE

APPENDIX B

Control Sample

Allis Chalmers Corporation
American Petrofina Inc.
Anchor Hocking Corp.

Athlone Industries

Barry Wright Corporation
Bertea Corporation

Boeing Company

Boise Cascade Corporation
Borg-Werner Corporation
Burroughs Corporation
Cincinnati Milacron Inc.
Citation Companies

Cleveland Cliffs Iron Company
Conrac Corporation
Continental Telephone Company
Dynalectron Inc.

Eastern Company

Federal Signal Corporation
Florida Gas

Gabriel Industries Inc.
Gillette Company

Hastings Manufacturing Company
Ideal Basic Industries Inc.
International General Industries
Kaiser Cement and Gypsum Co.
LaMaur Inc.

Northrop Corporation

Questor Corporation

Revere Copper and Brass Inc.
Sargent-Welch Scientific
Scovill Manufacturing Company
Smith International Inc.

TRW Inc.

Tech-Syn Corporation
Texasgulf Inc.

Thiokol Inc.

Torin Corporation

Triangle Industries

Uniroyal

watkins-Johnson

92

Stock Exchangg

NYSE
ASE

NYSE
NYSE
ASE

ASE

NYSE
NYSE
NYSE
NYSE
NYSE
ASE

NYSE
NYSE
NYSE
ASE

ASE

NYSE
NYSE
ASE

NYSE
ASE

NYSE
ASE

NYSE
ASE

NYSE
NYSE
NYSE
NYSE
NYSE
NYSE
NYSE
NYSE
NYSE
NYSE
NYSE
NYSE
NYSE
NYSE

LIST OF REFERENCES

Ball, Ray. "Changes in Accounting Techniques and Stock Prices,"
Empirical Research in Accounting: Selected Studies; 1972.
Supplement to Journal of Accounting Research, 1-38.

 

 

Ball, Ray and Brown, Philip. "An Empirical Evaluation of Accounting
Income Numbers," Journal of Accounting Research (Autumn 1968),
159-177.

Bar-Yosef, Sasson and Brown, Lawrence. "A Reexamination of Stock
Splits Using Moving Betas," Journal of Finance (September 1977),
1069-1080.

 

Bierman, Harold, Jr., and Dukes, Roland E. "Accounting for Research
and Development Costs," Journal of Accountancy (April 1975),
48-55.

 

Booz-Allen and Hamilton, Inc. Management of New Products. Booz-Allen
and Hamilton, Inc., 1968.

 

Chow, Gregory. "A Test of Equality Between Sets of Observations in
Two Linear Regressions," American Economic Review (July 1960),
591-605.

 

Collins, Daniel and Simonds, Richard. "SEC Line of Business Disclosure
and Market Risk Adjustments" (unpublished manuscript, Michigan
State University), 1977.

Corbin, Donald A. "Accounting Standards for Research and Development,"
Management Accountipg_(0ctober 1975), 47-48.

 

Dukes, Roland E. Market Evaluation of Accountinngnformation: A Cross
Sectional Test of Investor Response to ExpensipgTResearch and
Development Expenditures (unpublished Ph.D. dissertation, Stanford
University), 1974.

Fama, Eugene. "Efficient Capital Markets: A Review of Theory and
Empirical Werk," Journal of Finance (May 1970), 383-417.

 

Fama, E., Fisher, L., Jensen, M., and R011, R. "The Adjustment of
Stock Prices to New Information," International Economic Review
(February 1969), 1-21.

Financial Accounting Standards Board. Statement of Financial Accounting
Standards No. 2 - Accountinggfor Research and Development Costs,
1974.

93

94

Foster, George. "Stock Market Reaction to Estimates of Earnings per
Share by Company Officials," Journal of Accounting Research
(Spring 1973), 25-37.

Gee, Robert E. "A Survey of Current Project Selection Practices,"
Research Management (September 1971), 38-45.

 

Gerlach, John and wainwright, Charles Anthony. Successful Management
of New Products. Hasting House Publishers, Inc., 1968.

Gonedes, Nicholas. "Risk, Information, and the Effects of Special
Items on Capital Market Equilibrium," Journal of Accounting
Research (Autumn 1975), 220-256.

Gonedes, Nicholas J., and Dopuch, Nicholas. "Capital Market Equilibrium,
Information Production, and Selecting Accounting Techniques:
Theoretical Framework and Review of Empirical Work," Studies of
Financial Accountinngbjectives: 1974. Supplement to Journal
of Accounting Research, 48-129. ‘

 

 

 

Grant, Edward B. Interim Information and the Information Content of
Annual Earningg Announcements (unpublished Ph.D. dissertation,
Michigan State University), 1977.

 

 

Harrison, Tom. "Different Market Reactions to Discretionary and Non-
discretionary Accounting Changes," Journal of AccountingLResearch
(Spring 1977), 84-107.

Harvey, A. C. "An Alternative Proof and Generalization of a Test for
Structural Change," The American Statistician (August 1976),

 

International Accounting Standards Committee. International Accounting
Standard 9: Accountingpfor Research and Develgpment Activities,
1978.

Jensen, Michael. "Capital Markets: Theory and Evidence," Bell Journal
of Economics and Management Science (Autumn 1972), 357-398.

Johnson, Orace. "A Consequential Approach to Accounting for R & D,"
Journal of Accounting Research (Autumn 1967), 164-173.

 

Johnson, Orace. "Contra-Equity Accounting for R & D," The Accounting
Review (October 1976), 808-822.

Johnston, J. Econometric Methods. McGraw-Hill Book Company, 1972.

 

Kaplan, Robert and Roll, Richard. "Investor Evaluation of Accounting
Information: Some Empirical Evidence," Journal of Business
(April 1972), 225-257.

Kiger, J. "Empirical Investigation of NYSE Volume and Price Reactions
to the Announcement of Quarterly Earnings," Journal of Accounting
Research (Spring 1972), 113-128.

95

Lev, Baruch. "On the Association Between Operating Leverage and Risk,"
Journal of Financial and Quantitative Analysis (September 1974),
627-639.

Mansfield, Edwin. "Industrial Research and Development: Character-
istics, Costs, and Diffusion of Results," American Economic Review,
Paper and Proceedings (May 1969), 65-71.

May, Robert. "The Influence of Quarterly Earnings Announcements on
Investor Decisions as Reflected in Common Stock Price Changes,"
Empirical Research in Accounting: Selected Studies, 1971.
Supplement to Journal of Accounting Research, 119-163.

Milburn, Alex J. "An Empirical Study of the Relationship of Research
and Development Expenditures to Subsequent Benefits," (Unpublished
Research Study, University of Illinois), 1971.

Newman, Maurice S. "Equating Return from R & D Expenditures,"
Financial Executive (April 1968), 26-33.

 

Rubinstein, Mark. "A Mean-Variance Synthesis of Corporate Financial
Theory," Journal of Finance (March 1973), 167-181.

SchiffENMichael and Fabricant, Solomon. "Comments on Proposed State-
ment of Financial Accounting Standards-m-Accounting for Research
and Development Costs," Faculty Working Paper, College of Business
and Public Administration, New York University, 1974.

Siegel, Sidney. Nonparametric Statistics. McGraw-Hill Book Company,
Inc., 1956. '

 

Sunder, Shyam. "Relationship Between Accounting Changes and Stock
Prices: Problems of Measurement and Some Empirical Evidence,"
Empirical Research in Accounting: Selected Studies 1973.
Supplement to Journal of Accountinngesearch, 1-45.

 

 

Sunder, Shyam. "Stock Price and Risk Related to Accounting Changes in
Inventory Valuation," The Accounting Review (April 1975),
305-315.