#4

.73?! .hupxﬁuu
. .
u

V 4 ‘ v

.r

 

 

K....V...kw..§.5u. WW.
I A lit. 44,; ‘

AI. 3.x...”

«an.» :.,.m.«r:.m 1
:x. . v. ‘

. 11):! ..

IHESIS
\

UNIV|ERS SITY LIBRAR

llllllillllillllllll ’Wllllllll! WIN"

3 1293 01413 8956

LIBRARY

Michigan State
University

 

 

 

           

 

 

 

 

 

 

 

 

 

 

 

 

 

This is to certify that the

dissertation entitled

PENSION ACCOUNTING CHOICE: MANAGERIAL RESPONSE
TO SEC SCRUTINY OF DISCOUNT RATES

presented by

Norman Harrison Godwin

has been accepted towards fulﬁllment
of the requirements for

Ph . D . degree in Account ing

 

46(4th

6 Major professor

Date 6-91‘9‘

MS U is an Afﬁrmative Action/Equal Opportunity Institution 0-1277 1

 

 

PLACE IN RETURN 80X to remove We checkout from your record.
TO AVOID FINES return on or before dete due.

DATE DUE DATE DUE DATE DUE

      

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

MSU IeAn Afﬁrmative ActIoNEquel OpportunIIy InetItqun

 

PENSION ACCOUNTING CHOICE: MANAGERIAL RESPONSE
TO SEC SCRUTINY OF DISCOUNT RATES

By

Norman Harrison Godwin

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

DOCTOR OF PHILOSOPHY

Department of Accounting

1996

ABSTRACT

PENSION ACCOUNTING CHOICE: MANAGERIAL RESPONSE
TO SEC SCRUTINY OF DISCOUNT RATES

By

Norman Harrison Godwin

This study examines ﬁrm response to a September, 1993 letter written by the
Chief Accountant of the Securities and Exchange Commission (SEC) recommending
that ﬁrms reduce their assumed pension discount rates for 1993. Three hypotheses are
developed concerning the degree to which a ﬁrm decreases its discount rate in light of
the SEC recommendation. Based on positive accounting theory and the results of prior
pension research, it is posited that differences in compliance across ﬁrms will reﬂect
ﬁrm differences in the potential for being a target of SEC investigation, leverage, and
the adjusted funding ratio. A fourth hypothesis is developed concerning the degree to
which ﬁrms that comply alter another accounting choice in response to compliance. It
is posited that of those ﬁrms that comply, those with incentives to report smoothed
earnings will smooth the effect of compliance by changing another pension assumption,
the salary progression rate.

All four hypotheses are tested via multivariate regression. A sample of 600
ﬁrms selected from Compustat is used for tests of hypotheses one through three. As
expected, ﬁrms with a high potential for being a target of SEC investigation complied

to a greater degree to avoid that scrutiny while ﬁrms with high leverage complied to a

lesser degree to avoid the costs of tightening debt covenants. Contrary to expectations,
ﬁrms with low adjusted funding ratios complied to a greater degree. This result is
interpretted as consistent with the theory of political costs. A smaller sample of 277
ﬁrms is used for tests of hypothesis four. As expected, ﬁrms that complied with the
recommendation decreased their salary progression rates to mitigate the negative effect

from compliance.

ACKNOWLEDGMENTS

I would like to thank the members of my dissertation committee - Dr. Joseph
Anthony (chairman), Dr. Kathy Petroni, Dr. Richard Simonds, and Dr. Franklin
Boster - for their time, effort, and encouragement throughout this process. I would
especially like to thank Joe for his guidance and attitude and Kathy for the excellent
experience I gained as her research assistant.

1 would like to thank Dr. C. Wayne Alderman for being a role model in
academia, a mentor in life, and most of all, a friend. I would like to thank Dr. Kim
Key for listening to all the gory details of this process. A special thanks goes to the
members of my doctoral class - Brian Ballou, Barbara Esteves, and Dave Qi - for four
years of friendship and support.

I would like to thank my wife, Mary Ann, for enjoying these four years with
me. There is no better wife in the world. I would like to thank my children -
Harrison, Sara Caitlin, and Clay - for their excitement when I come home and for
helping me leave the pressures of school at school. I would also like to thank my
parents for instilling in me the qualities necessary to complete a doctoral program.

Finally, and most importantly, I would like to thank my God and Savior Jesus

Christ. Through His death, I have eternal life. Soli Deo gloria.

iv

TABLE OF CONTENTS

LIST OF TABLES .............................................................................. vii

LIST OF FIGURES ............................................................................ viii
CHAPTER

1. INTRODUCTION AND OVERVIEW .............................................. l

1.1 SEC recommendation .......................................................... 1

1.2 Overview of hypotheses ....................................................... 4

1.3 Overview of research design .................................................. 6

1.4 Summary of results ............................................................. 6

1.5 Organization of remaining chapters ......................................... 7

2. ACCOUNTING FOR PENSIONS ................................................... 9

2.1 PAS 87 ............................................................................ 9

2.1.1 Liabilities .............................................................. 11

2.1. 2 Exmnse ................................................................ 12

2.1.3 Assumptions ........................................................... 12

2.2 ERISA ........................................................................... 16

2.3 PAS 87 vs. ERISA ............................................................ 17

2.4 Summary ........................................................................ 18

3. HYPOTHESIS DEVELOPMENT .................................................. 19

3 .1 Compliance with the SEC’ 5 recommendation ............................ 20

3.1.1 RetentiaLIQLSEQinvestigatinn .................................... 21

3.1.2 Everage ............................................................... 23

3.1.3 Eundingratio .......................................................... 26

3.2 Smoothing the effect of compliance ........................................ 27

3.3 Summary ........................................................................ 3O

4. RESEARCH DESIGN ................................................................. 31

4.1 Sample selection ............................................................... 31

4.2 Tests of compliance ........................................................... 32

4.2.1 Dependentxatiable ................................................... 32

4.2.2 Independenuariables ................................................ 33

4.2.3 Wattles ..................................................... 35

V

4.3 Tests of smoothing ............................................................ 37

4.3.1 Dependentlai’iable ................................................... 37

4. 3. 2 lndemndenuariables ................................................ 39

4. 3. 3 Cbmmlxatiables ..................................................... 41

4. 3. 4 AdditienaLsmoethingtcsts .......................................... 43

4.4 Summary ........................................................................ 45

5. RESULTS ................................................................................ 46
5.1 Descriptive statistics .......................................................... 46

5.2 Results for tests of compliance .............................................. 50

5 2 1 Altematixemmnrefommplianee .............................. 56

5 2 2 Wm ............................... 58

5 2 3 Eunhetanalxsimthxpcthesism ................................ 58

5 2 4 Alternatixuxplanatiomfonhxmthesisthree ..................... 59

5.2.5 Risleeffeet ............................................................. 62

5.2.6 Analysisoiﬁnnslestmsampleseleetm ....................... 65

5.3 Results for tests of smoothing ............................................... 67

5.3.1 ResumfmmaddmonaLsmmthimsts .......................... 74

5.4 Summary ........................................................................ 79

6. CONCLUSIONS ........................................................................ 80
6.1 Summary of research ﬁndings ............................................... 80

6.2 Contributions ................................................................... 82

6.3 Suggestions for future research ............................................. 83

LIST OF REFERENCES ...................................................................... 84

vi

Table

10

11

12

13

14

LIST OF TABLES

Sample selection ........................................................................ 32
Descriptive statistics for 600 sample firms ......................................... 48
Changes in disclosed actuarial assumptions, by year ............................ 49
Correlation coefﬁcients ................................................................ 51
Regression results for hypotheses one, two, and three .......................... 55
Regression results with alternative measure of compliance ..................... 57
Correlations between discount rates and debt ratings ............................ 64
Descriptive statistics for ﬁrm groups based on 1992 discount rate ............ 66
Correlation matrix ...................................................................... 69
Regression results for hypothesis four .............................................. 71
Tobit regression results ............................................................... 73
Comparison of subsample and non-subsample ﬁrms ............................. 75
Yearly correlations between changes in discount rates and salary rates ...... 76
Regression results for alternative hypothesis for hypothesis four ............. 78

vii

LIST OF FIGURES

Figure

1 Effect of differential actuarial assumptions ........................................ 15
2 Comparison of discount rates and SEC benchmark .............................. 61
3 Predicted signs for hypothesis four regression coefﬁcients ..................... 7O

viii

Chapter One
INTRODUCTION AND OVERVIEW

This study examines ﬁrm response to a September, 1993 letter written by the
Chief Accountant of the Securities and Exchange Commission (SEC) recommending
that ﬁrms reduce their assumed pension discount rates for 1993. The general questions
of interest are as follows: what impact did such a recommendation have on accounting
choice; did certain ﬁrm characteristics systematically affect response to the
recommendation in a manner consistent both with current accounting theory such as
positive theory and with the results of previous pension research; and did some ﬁrms
alter other accounting choices to mitigate the effect of compliance. Answers to these
questions will contribute to the growing understanding of how accounting choices are
made.

This introductory chapter continues in section 1.1 with a discussion of the
SEC’s recommendation for lower discount rates. Sections 1.2 and 1.3 contain
overviews of the hypotheses tested and the research design, respectively. The results
are summarized in section 1.4. Section 1.5 concludes this chapter with the
organization of the remaining chapters.
1 .1 SEC recommendation

Although several years have passed since the highly publicized issuance of
Statement of Financial Accounting Standards No. 87, "Employers' Accounting for
Pensions" (PAS 87), deﬁned beneﬁt pension plans have continued to draw attention.
One aspect of the accounting for pensions that received recent scrutiny from the
ﬁnancial press, regulators, and political bodies is the discount rate assumption, a
critical component in the calculation of both net periodic pension expense and pension
liabilities. This scrutiny arose in the early 1990’s due to concerns that ﬁrms were

intentionally assuming discount rates that were too high given the declining interest rate
1

2

environment of the early 1990’s. Assuming high discount rates results in less
conservative estimates of pension expense and pension liabilities.

During 1993, the SEC took action. On September 22, SEC Chief Accountant
Walter Schuetze sent a letter to the Emerging Issues Task Force (EITF) of the
Financial Accounting Standards Board (FASB) making public the SEC's expectation
that registrants comply at the next measurement date with the FASB's original
recommendation in PAS 87 concerning discount rates. The FASB had held that to
choose an appropriate discount rate, ﬁrms could consider the rates of return on ”high-
quality ﬁxed-income investments currently available " (PAS 87, par. 44)‘. In its letter,
the SEC deﬁned "high-quality ﬁxed-income investments" as those receiving a Moody's
rating of Aa or higher. This deﬁnition effectively established the accepted set for 1993
discount rates as a range from 7 % to 7.5 %. The SEC threatened that any discount rate
unreasonably high would trigger an ofﬁcial inquiry.

Although some type of admonishment from the SEC to consider lowering rates
could have been anticipated by fu'ms due to falling interest rates, the extreme degree of
the SEC's position apparently was not expectedz. Timothy Lucas, chairman of the
EITF and recipient of Schuetze's letter, commented that he could recall no previous
announcement from the SEC concerning this matter. Financial press articles referred
to the SEC's position as "lobbing what amounts to a grenade into the books,” an
"unprecedented warning, " and a " great stir. " Firms appear to have been caught off
guard. In effect, those ﬁrms with high discount rates for 1992 were now faced with

the unexpected impact of a lower discount rate for 1993.

 

1 Even with this position from the FASB, in 1986, the year in which many ﬁrms early-
adopted PAS 87, there was wide variation in the choice of discount rates [Stone and
Ingram (1988)].

2 Jereski, L. ”SEC is Challenging Funding for Plans,” 77w Wall Street Journal.

3

It is reasonable to assume that ﬁrms that chose a high discount rate in 1992
generally would be dissatisﬁed with the SEC's position. A portion of this
dissatisfaction can be linked to ﬁrm responses to discussion memorandums and
exposure drafts issued during the development stages of PAS 87. Concerned about
potential volatility in pension expense and liabilities that would result from choosing
discount rates based on current settlement rates, some of the 408 respondents suggested
that an "averaging technique should be used to smooth potential year-to-year changes
(in discount rates)" (PAS 87 , par. 201). However, the FASB argued that "(i)f a
ﬁnancial measure purports to represent a phenomenon that is volatile, the measure must
show that volatility or it will not be representationally faithful" (PAS 87, par. 174).
Therefore, the FASB required that the discount rate be chosen as a current settlement
rate deﬁned as the rate of return on high—quality ﬁxed-income investments currently
available.

A potentially more pervasive source of ﬁrm dissatisfaction with the SEC’s
position is the effect that compliance has on the ﬁrm’s ﬁnancial statements. Decreasing
the discount rate results in increased pension liabilities disclosed in the footnotes (e. g. ,
the Accumulated Beneﬁt Obligation) for the current year and increased pension expense
for the following year. In a worst case scenario, the decrease in the discount rate could
necessitate a separate reduction to stockholder's equity due to funding deﬁciencies as
measured by the increased liability. In a best case scenario, the effect on disclosed
liabilities and pension expense could be immaterial.

The latter scenario is improbable for the majority of pension sponsors. As
indicated by the median sample ﬁrm discount rate for 1992 of 8.25 % , compliance with
the recommendation required a decrease of about one percent for at least half of all
pension sponsors. Such a decrease can have a substantial negative impact on pension
accounts. For example, in terms of the effect on disclosed liabilities, a one percent

decrease in the discount rate can potentially result in an increase in the pension liability

4

of 20% or more [Winklevoss (1993)]. In terms of pension expense, the effect of
reducing the discount rate one (two) percent has been estimated to be an average of
4.5% (8.8%) of after-tax income3. Thus, compliance with the SEC's recommendation
can be very costly.
1.2 Overview of hypotheses

Several empirical questions arise as a result of the SEC's recommendation.
First, what are the implications of an SEC recommendation? In other words, does a
recommendation from the SEC have enough inﬂuence to alter an accounting choice that
negatively impacts reported expenses and liabilities. Second, why is there variation, if
any, in the degree of compliance across ﬁrms? Did factors suggested by existing
theory and prior results from pension research inﬂuence the degree of compliance?
Third, given that ﬁrms complied with the recommendation to some degree, did some
ﬁrms alter other accounting choices to mitigate the effect of compliance? Answers to
these questions will add to the understanding of how accounting choices are made.

This study examines each of these questions. Three hypotheses are developed
concerning the degree to which a ﬁrm decreases its discount rate in light of the SEC
recommendation. It is posited that differences in compliance across ﬁrms will reﬂect
ﬁrm differences in certain economic and ﬁnancial characteristics. A fourth hypothesis
is developed concerning the degree to which ﬁrms that comply alter another accounting
choice in response to compliance. It is posited that of those ﬁrms that comply, those
with incentives to report smoothed earnings will smooth the effect of compliance by
changing another pension assumption, the salary progression rate. An overview of
each of the four hypotheses follows below.

The ﬁrst of the three compliance hypotheses is based on the political cost
hypothesis. The political cost hypothesis predicts that those ﬁrms under political

 

3 This estimation, based on approximately 2000 ﬁrms from Compustat, was provided
by Sue West of SEI Capital Markets Research.

5

scrutiny will make accounting choices that help to avoid increased scrutiny and the
costs associated with that scrutiny [Watts and Zimmerman (1986)]. Because the SEC’s
recommendation represents increased political scrutiny of accounting practice, ﬁrm
compliance should be increasing in the perceived and actual costs of that scrutiny.
However, because the perceived and actual costs of scrutiny are difﬁcult if not
impossible to measure, this study focuses on the potential for a ﬁrm to be an actual
target of SEC investigation. Firms with a high potential for investigation should be
willing to comply to avoid the costs associated with an SEC investigation. Thus,
hypothesis one predicts that compliance with the recommendation is positively
associated with the potential for being a target of SEC investigation.

The second hypothesis is based on positive accounting theory’s debt-equity
hypothesis, which predicts that the larger a ﬁrm’s leverage, the more likely that ﬁrm is
to make income increasing accounting choices [Watts and Zimmerman (1986)]. The
debt-equity hypothesis is based on the assumption that leverage proxies for the presence
of and closeness to debt covenants, assumptions validated by Duke and Hunt (1990)
and Press and Weintrop (1990). Highly leveraged ﬁrms are therefore predicted to
make income increasing accounting choices to avoid the costs associated with the
violation of debt covenants. Because compliance with the SEC’s recommendation
reduces income as well as increasing disclosed liabilities, ﬁrms with high leverage
should be hesitant to comply. Thus, hypothesis two predicts that compliance is
negatively associated with leverage.

The development of hypothesis three draws from the results of previous pension
research. Feldstein and Morck (1983) and Francis and Reiter (1987) ﬁnd evidence
consistent with ﬁrms choosing pension discount rates to create the appearance of a
better funded plan. Proposed reasons for this behavior include avoiding the necessity
of making contributions to the pension fund [Feldstein and Morck (1983)], avoiding
potential political costs of having an unfunded plan [Francis and Reiter (1987)], and

6

avoiding negative perceptions of ﬁnancial condition [Harper and Strawser (1993)].
Compliance with the recommendation reduces the reported funded status of the pension
plan. Firms with less funded plans on an adjusted basis should therefore be hesitant to
comply so that they may avoid the costs associated with reporting a less funded plan.
Thus, hypothesis three predicts that compliance is positively associated with the funded
status of the plan when adjusted to a common discount rate.

Hypothesis four is based on the smoothing hypothesis. The smoothing
hypothesis predicts that ﬁrms choose accounting procedures to create stable earnings
streams. Given that a ﬁrm complies with the recommendation and has an incentive to
report smooth earnings, then that ﬁrm can offset the negative effect of compliance by
reducing another pension assumption, the salary progression rate. The expectation of
this type of smoothing behavior is supported by Ronen and Sadan (1981) who observe
that smoothing can be accomplished through changes in pension actuarial assumptions.
Thus, hypothesis four predicts that ﬁrms with incentives to report smooth earnings who
lower their discount rate in compliance with the SEC are more likely to lower their
salary progression rate.

1.3 Overview of research design

Tests of the above hypotheses are conducted on a sample of up to 751 ﬁrms
using archival data collected from publicly available sources. Hypotheses one, two and
three are tested by regressing a measure of compliance onto variables representing the
three hypotheses as well as several control variables. Hypothesis four is tested by
regressing the smoothing measure onto both the measure of compliance and several
variables representing the incentive to report smoothed earnings. Various tests are
conducted to discern the robustness of the results to alternate speciﬁcations.

1.4 Summary of results
The data in the study are consistent with the SEC’s recommendation having a

substantial impact on accounting choice. The decrease in the discount rate for 1993 for

7

the average ﬁrm was approximately one percent. This decrease translated into a
signiﬁcant decrease in the reported funded status for the average ﬁrm. In addition, the
decrease was much more pronounced than in prior years, both in the magnitude of the
change and in the percentage of ﬁrms making a change. This result illustrates that the
change in the discount rate for 1993 was well outside the historical trend, suggesting
the introduction of an outside inﬂuence in 1993, namely the SEC recommendation,
which signiﬁcantly altered accounting choice.

As expected, ﬁrms that faced a greater potential for being a target of SEC
investigation complied with the recommendation to a greater degree (i.e. , had larger
decreases in their discount rates) to avoid being the target of SEC scrutiny. In
addition, ﬁrms with higher leverage complied to a lesser degree to avoid the costs of
tightening debt covenants. Contrary to expectations, ﬁrms with lower adjusted funding
ratios complied to a greater degree than ﬁrms with higher adjusted funding ratios.
Assuming that the adjusted funding ratio proxies for political sensitivity to the SEC in
the pension area, such ﬁrm behavior can be interpreted as consistent with the theory of
political costs. Firms with greater sensitivity, as measured by a low adjusted funding
ratio, complied to a greater degree to avoid the costs associated with political scrutiny.

Finally, the data are consistent with ﬁrms smoothing the effect of compliance by
decreasing the salary progression rate. The average decrease in the salary rate for
1993 was approximately one-half percent, a decrease well beyond the historical change
both in magnitude and in the percentage of ﬁrms making a change. Regression results
conﬁrm that ﬁrms that complied with the SEC’s recommendation decreased their salary
rates to mitigate the negative effect from compliance.

1.5 Organization of remaining chapters

The remaining chapters are organized as follows. Chapter two discusses the

accounting requirements for pension plans as prescribed by PAS 87 and the statutory

requirements as prescribed by the Employee Retirement Income Security Act of 1974.

8

Reviews of relevant literatures and development of the hypotheses to be tested are
contained in chapter three. Chapters four and ﬁve detail the research design and
present empirical results, respectively. Concluding remarks are contained in chapter

six.

Chapter Two
ACCOUNTING FOR PENSIONS

Employers sponsoring deﬁned beneﬁt pension plans are currently subject to two
sets of regulations: the Employee Retirement Income Security Act of 1974 (ERISA)
and Statement of Financial Accounting Standards No. 87 , "Employers' Accounting for
Pensions” (FAS 87). ERISA established statutory law for pension plans while FAS 87 ,
issued by the Financial Accounting Standards Board (FASB) in 1985, established
accounting and ﬁnancial reporting rules. The two regulations are related but not equal.

This chapter continues with a review of FAS 87 in section 2.1. The two
objectives of this review are to explain the pension accounting process prescribed by
FAS 87 and to highlight the importance of actuarial assumptions in that process.
Following this review is a discussion of ERISA in section 2.2. Section 2.3 follows
with a comparison of the two regulations, speciﬁcally focusing on the actuarial
assumptions under each regulation. The objective of this comparison is to demonstrate
that funding considerations under ERISA can be separated from accounting
considerations under FAS 87. Section 2.4 summarizes the chapter.

2.1 FAS 87

Although it could be adopted as early as 1985, FAS 87 became effective in
1987 for all ﬁrms sponsoring deﬁned beneﬁt pension plans‘. The standard was the
culmination of an eleven year effort by the FASB to establish better accounting and

reporting rules for pension plan sponsors. This effort began in 1974 in response to

 

4 A deﬁned beneﬁt pension plan is a “pension plan that deﬁnes an amount of pension
beneﬁt to be provided (to the employee at retirement or other termination of
employment), usually as a function of one or more factors such as age, years of
service, or compensation.” (FAS 87, glossary). In contrast, a deﬁned contribution
pension plan is a plan that “speciﬁes how contributions to the (employee’s pension)
account are to be determined instead of specifying the amount of beneﬁts the individual
is to receive” at retirement or other termination of employment (FAS 87, glossary).

9

10

both passage of ERISA and criticism of Accounting Principles Board Opinion 8,
"Accounting for the Cost of Pension Plans" (APB 8), the only accounting
pronouncement governing pension accounting at that time. In 1980, the FASB issued
two standards. The ﬁrst, Statement of Financial Accounting Standards No. 35,
"Accounting and Reporting by Deﬁned Beneﬁt Pension Plans,” (FAS 35) addressed
reporting by actual plans5. The second, Statement of Financial Accounting Standards
No. 36, "Disclosure of Pension Information," (FAS 36) amended APB 8 by requiring
new disclosures by plan sponsors such as net assets, vested beneﬁts, and assumed rates
of return. However, FAS 36 made no changes in the accounting provisions for the
measurement of pension expenses and liabilities. The standard was an admittedly
interim step towards the more comprehensive PAS 87 which would supersede both
APB 8 and FAS 36. The passage of PAS 87 in 1985 continued "the evolutionary
search for more meaningful and more useful pension accounting" (FAS 87, par. 5), but
the FASB noted that even this standard was unlikely to be the ﬁnal step in that
evolution. Since 1985, however, pension accounting has remained relatively stable“.
The purpose of FAS 87 was twofold. First, it established a uniform calculation
of pension expense in order to provide a more representationally faithful,
understandable, and comparable measure of pension costs. Second, it provided for

more detailed and understandable disclosures of pension liabilities in order to further

 

5 Note that FAS 35 established the accounting and reporting standards for the annual
ﬁnancial statements of deﬁned beneﬁt pension plans. The focus of this study is not on
the accounting prescribed for the pension plan itself, but on the employer’s accounting
for pensions. FAS 36 and ultimately FAS 87 established standards for such employer
accounting for pensions.

6 Along with FAS 87, the FASB issued Statement of Financial Accounting Standards
88, "Employers' Accounting for Settlements and Curtailments of Deﬁned Beneﬁt
Pension Plans and for Termination Beneﬁts. " Aside from FAS 88 and three
amendments to FAS 87, no new pension standards have been released by the FASB
since 1985.

11

improve ﬁnancial reporting. In each of these objectives, actuarial assumptions play a
key role. The following paragraphs discuss the liability and expense measures under
the standard, deﬁne the most inﬂuential assumptions, and explain how these
assumptions affect the liabilities and expense.

2.1.1 Liabilities

FAS 87 requires the estimation and disclosure of three liability measures: the
vested beneﬁt obligation (VBO), the accumulated beneﬁt obligation (ABO), and the
projected beneﬁt obligation (PBO). The ABC and PBO are both actuarial present
values of future beneﬁts payable to employees for service to date. Given that pension
beneﬁts are deﬁned as a function of future salary levels, the difference between the
ABC and PBO is that beneﬁts under the ABO are based on current and past salary
levels while beneﬁts under the PBO are based on projected future salary levels. The
VBO is that portion of the ABO that has vested. The VBO and ABC provide
information about the sponsor's obligation if the plan is discontinued today while the
PBO provides information on the obligation assuming that the plan continues in effect
and that expected future events such as salary increases occur. As a result, the PBO
will generally be the largest measure followed by the ABC and then the VBO. All
three are reported as of the end of the ﬁscal period.

FAS 87 also requires the recognition of a liability on the balance sheet if certain
conditions are met. If cumulative pension expense to-date exceeds cumulative
contributions to the pension plan, then an accrued pension liability must be included in
the balance sheet. The value of this liability is the difference between the two
cumulative amounts. Also, when the ABO exceeds plan assets, an additional minimum
liability must be recorded for the difference. This required additional minimum
liability is reduced by any accrued pension liability already recognized on the balance

sheet.

12

2.1.2 Expense

FAS 87 also requires a uniform calculation of pension expense comprised of the
following components: service cost, interest cost, return on plan assets, and
amortization of prior service cost, gains or losses, and the transition asset/obligation.
Of main concern to this study are service and interest costs. Service cost is the
actuarial present value of beneﬁts earned by employees for service in the current
period. In other words, it is that portion of the PBO that is attributable to service
rendered in the current period. Like the PBO, service cost reﬂects future salary levels.
Interest cost is simply the increase in the PBO due to the passage of time and is
calculated as the product of the beginning PBO balance and an appropriate interest rate.

2.1.3 Assumptions

Actuarial assumptions are extremely important to both the measurement of
pension liabilities and the calculation of pension expense. Recognizing their
importance, the FASB requires the disclosure of three main assumptions used in the
accounting for pensions prescribed by FAS 87: the discount rate, the salary
progression rate, and the expected return on plan assets. Each rate is to be chosen
explicitly, meaning that each should reﬂect the best estimate of the plan's future
experience with respect to that assumption alone (FAS 87, par. 191). Below is a
discussion of the most inﬂuential assumptions, the discount and salary progression rates
(FAS 87, par. 195), and how each impacts the accounting under PAS 87.

Under APB 8, the discount rate was deﬁned as an average expected return on
funds invested or to be invested (APB 8, Appendix A). FAS 87 separates this
expectation into two components, the discount rate and the expected return on plan
assets. This separation is done due to the FASB's belief that pension information is
more understandable if the asset-related aspects of the pension plan are separated from

the liability-related aspects (FAS 87, par. 180). Thus, the discount rate is deﬁned in

13

FAS 87 as the "interest rate used to adjust for the time value of money" (FAS 87,
glossary) while the expected return is deﬁned as its name indicates.

FAS 87 states that "discount rates shall reﬂect the rates at which pension
beneﬁts could be effectively settled" (FAS 87, par. 44). The FASB's intention is that
the discount rate be chosen by the sponsor as the best estimate of implicit rates in
annuity contracts that could appropriately settle the pension obligation. In keeping with
this intention, the FASB expects each sponsor to annually reevaluate its discount rate to
determine whether it remains the best estimate of an effective current settlement rate.
As a result, the FASB expects discount rates to change as interest rates rise or fall.

Of all assumptions made in current pension accounting, the discount rate has the
most signiﬁcant impact. First, it serves as the discount factor for the present value
calculations of both the PBO and service cost. Second, it serves as the interest rate at
which interest cost is calculated on the PBO. Note that the end-of-year discount rate is
used for the PBO calculation while the beginning-of-year rate is used for the service
and interest costs calculations. In general, as the discount rate decreases, the PBO,
service cost, and interest cost increase7.

The salary progression rate has the second most signiﬁcant impact on pension
beneﬁts and expenses. It is related to the discount rate in that each incorporates
expectations of the same economic environment (FAS 87, par. 195); however, the
salary progression rate is chosen explicitly and is therefore independent from the
discount rate. The salary rate reﬂects the sponsor's best estimate of future salary
increases to be granted to its employees. Because the PBO and service cost are both
based on future salary levels, the salary rate impacts the magnitude of each. Due to its

effect on the PBO, the salary rate also indirectly impacts the magnitude of interest cost.

 

7 Calculation of interest cost is affected in two ways: the rate used to calculate interest
is decreased but the PBO is increased by the decrease in the rate. The latter increase is
proportionally larger, thus causing interest cost to increase [Winklevoss (1993)].

14

Like the discount rate, the end-of-year salary rate is used for the PBO calculation while
the beginning-of-year rate is used for the service cost calculation. In general, as the
salary progression rate decreases, the PBO, service cost, and interest cost decrease3.

To illustrate how the discount and salary rates are generally used in the
calculation of pension expense, the following example is taken from Fogarty and Grant
(1995). Starting with a 45 year old employee with 15 years of service who will retire
at age 65, the authors compute the annual pension beneﬁt for that employee by
projecting, via the salary progression rate, the current salary ($50,000) to the
employee's last year of expected employment and by assuming a pension beneﬁt
formula of 70 percent of that ﬁnal year's salary. This annual beneﬁt is then multiplied
by the present value factor for an ordinary annuity over the assumed expected post
retirement period (10 years) to obtain the total beneﬁt due the employee during
retirement. This total beneﬁt is then discounted back to the end of the year for which
pension expense is being calculated to obtain the present value of the total future
beneﬁt. Finally, this present value is divided by the employee's remaining years of
service to obtain the current period's annual service cost component of pension
expense.

One objective of Fogarty and Grant (1995) is to illustrate the impact of differing
actuarial assumptions on the service cost component of pension expense. In Table 1 of
their paper, the authors report values for annual service cost, computed as described
above, under differing actuarial assumptions. A subset of that table is reported in

Figure 1.

 

8 The third assumption under FAS 87 is the expected return on plan assets. While this
study does not examine the expected return, it represents the long-term (e.g. , ﬁve-year
average) expected return on plan assets invested or to be invested (FAS 87, par. 45)
and is used in the calculation of pension expense but not in the calculation of pension
liabilities. For a detailed examination of the expected return, see Blankley and
Swanson (1995).

15

 

 

 

 

 

 

Discermtrate
9% 10%
Salami: 2% $ 3.246 $ 2.613
3% $ 3,945 $ 3,176
Figure 1

Effect of differential actuarial assumptions

Figure 1 illustrates well the different aspects of changing the discount rate
and/or the salary rate. First, it illustrates that changing the discount rate has a larger
effect than changing the salary rate. Starting with a 10% discount rate and a 3% salary
rate, the annual pension expense in the example is $3,176. Reducing the discount rate
one percent increases pension expense to $3,945, an increase of 24.2%. In contrast,
reducing the salary rate one percent decreases pension expense to $2,613, a decrease of
17.7%. Second, it illustrates the offsetting nature of the two rates. Reducing both
rates one percent increases pension expense to $3,246 from $3,176, and increase of
only 2.2%. Overall, the example demonstrates the importance of assumptions in

pension accounting9.

 

9 It should be noted that the ﬁgures in the example are based on a simpliﬁed calculation
of service cost and are sensitive to the age of the employee. Fogarty and Grant’s
(1995) example also includes a 60 year old employee. For that employee, the effects of
the two rate changes are still offsetting, but the effect of a change in the salary rate is
now greater than the effect of a change in the discount rate. However, as the mean age
of the civilian workforce as of 1990 was 37.7 years and rising [US Department of
Labor Bulletin 2426 (1993)], the ﬁgures of the 45 year old employee appear to be more
relevant to the average pension plan.

16

2.2 ERISA

ERISA became law on September 2, 1974. Its main goal was to protect the
interests of plan participants by establishing new rules for participation, adding
mandatory vesting schedules, requiring disclosures of plan information, creating a
system to insure pension payments, and setting minimum funding standards. Of
primary concern to this study are the funding standards and the actuarial assumptions
underlying these standards.

Under ERISA, the minimum required contribution for a given year includes
several components: normal costs, supplemental costs, the balance in the funding
standard account, and an additional funding charge, if necessary. Normal costs are
those costs attributable to the current year of service by plan participants while
supplemental costs are the current year's amortized portion of costs attributable to
previous periods. The funding standard account balance is the amount of excess or
deﬁcit in funding as of the previous year, and the additional funding charge is a
requirement designed to improve the funding of highly under-funded plans [see
Winklevoss (1993) for a more technical discussion].

In calculating normal and supplementary costs, actuarial assumptions
concerning future interest returns on plan assets, salary increases, and various types of
decrements (e.g. , mortality) must be made. Like FAS 87, the most inﬂuential of these
assumptions are the funding interest rate and the equivalent salary increase rate.
Unlike PAS 87, the funding interest rate is an expectation of investment returns from
the plan's funding policy, not a rate at which pension obligations could be currently
settled. The funding interest rate and equivalent salary increase rate are chosen as the
actuary's best estimate of the long-term outlook for the plan considering past plan
performance and current market trends. This long-term perspective is encouraged by
ERISA so that contributions to the pension fund, and therefore participant security, will

be relatively stable (i.e. , unaffected by temporary market swings).

17

2.3 FAS 87 vs. ERISA

The main difference between ERISA and FAS 87 that is of interest to this study
lies in the choice of assumptions, especially the choice of discount factors. While
ERISA discounts future beneﬁts payable to participants for service in the current year
(i.e., normal costs) at the expected rate of investment return, FAS 87 discounts those
beneﬁts (i.e., service cost) at the current settlement rate. This difference stems from a
divergence of perspectives. Actuaries are directed towards a more long-term
perspective in choosing funding interest rates so that temporary ﬂuctuations in the
market or in ﬁrm performance do not result in changes in the rate for statutory
funding“). On the other hand, employers are directed towards a more short-term
perspective in choosing discount rates. As a result, rates used for statutory funding
under ERISA can be different from the rates used for accounting under FAS 87
[Blankley and Swanson (1995)].

This difference in rates was not always the case. Under APB 8, the discount
rate was equivalent to the ERISA funding interest rate. Thus, changes in the discount
rate mirrored changes in the funding interest rate. However, under PAS 87, the
discount rate can be changed while the funding interest rate remains the same. This
type of disparity in assumption choice is illustrated in two 1993 reports prepared by
Buck Consultants, a worldwide pension consulting ﬁrm. In The Health of America 's
Pension Plans and Economic Pension Actuarial Assumptions Survey, Buck Consultants
reports that average discount rates fell from 8.2% in 1992 to 7.3% in 1993 while
average funding interest rates remained steady at 8.2%.

The most important aspect of this ability to change the discount rate without
changing the funding interest rate is the effect on cash ﬂows or funding. Research on

pensions prior to FAS 87 documented an association between changes in discount rates

 

‘0 Per discussions with pension actuaries, funding interest rate and equivalent salary
increase rate assumptions for statutory funding purposes change infrequently.

18

and funding [e.g., see Healy and Palepu (1990)]. However, because statutory funding
requirements are unaffected by changes in accounting discount rates under FAS 87,
Blankley and Swanson (1995) ﬁnd no evidence that actuarial assumptions under FAS
87 were manipulated to inﬂuence cash outﬂows to the pension plan. As a result, the
choice of discount and salary rates under FAS 87 can be separated from funding
considerations [Ali and Kumar (1993)].
2.4 Summary

In summary, sponsors of deﬁned beneﬁt pension plans are subject to ERISA,
which is concerned with the funding of pensions, and FAS 87, which is concerned with
the overall accrual accounting for pensions. Under FAS 87, ﬁrms are required to
disclose three measures of the pension liability and recognize pension expense.
Actuarial assumptions play a major role in measuring these amounts. In addition, the
assumptions chosen under FAS 87 are different than those under ERISA. As a result,
changing such assumptions does not affect minimum cash outﬂows required for
statutory funding. Therefore, sponsors of pension plans can choose assumptions under
FAS 87 without direct cash ﬂow implications, thus allowing them to choose
assumptions in reaction to other contracts of the ﬁrm. Chapter three will develop

hypotheses concerning this type of behavior.

Chapter Three
HYPOTHESIS DEVELOPMENT

In their seminal article, Jensen and Meckling (1976) characterize the ﬁrm as a
nexus of contractual relationships in which principals engage agents to perform some
task on the principals' behalf. Emphasizing transaction costs between these principals
and agents and opportunistic behavior resulting from incomplete contracts, Watts and
Zimmerman (197 8) began positive accounting theory. Under positive accounting
theory, the ﬁrm's contracting parties hold private information and seek to maximize
individual utility. Contracting and information costs are assumed to be non-zero and
accounting is assumed to be an integral part of resulting contracts, both formal and
informal. As a result, accounting choice affects these contracts and therefore ﬁrm
value. Based on this role of accounting, positive accounting theory attempts to explain
and predict managerial choices of ﬁnancial accounting/ reporting practices [Watts and
Zimmerman (1986)].

Watts and Zimmerman (1990) explain that managers select accounting practices
from an "accepted set" that is determined by the contracting parties. The accepted set
is deﬁned by restrictions which produce "best" or ”accepted” accounting practices even
in the absence of mandated accounting standards. The discretion embodied in the
accepted set gives rise to accounting choice studies where accounting choice is
explained by examining variables representing a manager's incentive to select one
accounting practice over another or to select a certain combination of accounting
practices. According to Watts and Zimmerman (1990), these studies are usually
performed under one of three major contractual settings: management bonus plans, debt
contracts, and the political process.

The SEC’s recommendation concerning discount rates provides a setting in

which to examine the determinants of accounting choice. As discussed in chapter one,
19

20

in 1993 the SEC recommended that ﬁrms reduce their discount rates to a range of 7 %
to 7.5 %. Because compliance with this recommendation reduces income and increases
liabilities, ﬁrm compliance could be contingent on the potential economic impact of
these ﬁnancial statement effects as well as the non-ﬁnancial statement ramiﬁcations of
such effects (e.g. , tightening of debt covenants). In addition, because the SEC is a
political body with the ability to impose costs on ﬁrms, compliance could also be
contingent on the costs associated with SEC scrutiny due to noncompliance.

Therefore, it is posited that differences in compliance across ﬁrms will reﬂect ﬁrm
differences in certain economic and ﬁnancial characteristics.

This chapter continues with the development of four hypotheses concerning ﬁrm
compliance with the SEC’s recommendation. Speciﬁcally, compliance is hypothesized
to be increasing in the potential for being a target of SEC scrutiny, decreasing in
leverage, and increasing in pension funding ratios. These three hypotheses are
developed in section 3.1. Furthermore, of those ﬁrms that comply, those with
incentives to report smoothed earnings are hypothesized to smooth the effect of
compliance by changing another pension assumption, the salary progression rate. This
hypothesis is developed in section 3.2. Section 3.3 summarizes the chapter.

3. 1 Compliance with the SEC's recommendation

The decision of whether to comply or not with the SEC's recommendation can
be considered as a tradeoff between the costs and beneﬁts associated with compliance.
The costs of compliance are those costs (e.g. , tightening of debt covenants) that may
arise as a consequence of the increases in expenses and disclosed liabilities generated
by compliance. The beneﬁts of compliance are the avoidance of the costs associated
with increased political scrutiny from the SEC. Hypotheses associated with these costs
and beneﬁts of compliance are developed below.

21

3.1.1 RetentiaLfQLSEtLimLestigatiQn
Accounting choice has been hypothesized to be affected by the potential and

actual costs associated with scrutiny from political bodies [Watts and Zimmerman
(1978)]. This reasoning has been formalized into the political cost hypothesis which
predicts that those ﬁrms under political scrutiny will make accounting choices that help
to avoid increased scrutiny and the costs associated with that scrutiny [Watts and
Zimmerman (1986)]. The SEC's recommendation, originating from a powerful
political body with the ability to impose signiﬁcant costs on ﬁrms, represents a speciﬁc
case of increased political scrutiny of accounting practice. Avoidance of the costs
associated with this increased political scrutiny is therefore a beneﬁt of compliance.

SEC scrutiny begins in the Division of Corporate Finance with an examination
of a ﬁrm's annual ﬁling [Skousen ( 1991)]. If the examination uncovers a violation,
then the SEC either will send a letter of comments to the registrant detailing the
problems with the ﬁling or in cases of substantial violations, will refer the problem to
the Division of Enforcement [Kile (1993)]. When a letter of comments is sent, the
registrant has the opportunity to respond to the issues raised. If the SEC is satisﬁed
with the response, then the investigation is dropped. If the SEC is not satisﬁed, then it
can instruct the registrant to better comply in future ﬁlings or require the registrant to
submit an amended ﬁling. When the violation is referred to the Division of
Enforcement, ofﬁcial enforcement actions, summarized in the SEC's Accounting and
Auditing Enforcement Releases (AAER's), are conducted. Ofﬁcial enforcement actions
can result in the suspension of trading of a registrant's security, civil injunctions, or
referrals to criminal authorities.

Accounting research has shown that action taken by the SEC imposes tangible
costs on the targeted registrant. Negative cumulative abnormal returns have been
documented for ﬁrms ﬁling amended ﬁnancial statements in response to letters of

cement [Kile (1993)], for ﬁrms receiving AAER's [Feroz, Park, and Pastena (1991)],

22

for ﬁrms whose trading of stock is suspended by the SEC [Howe and Schlarbaum
(1986)], and for ﬁrms that are targets of general SEC litigation [Nourayi ( 1994)]. In
addition, Feroz et al. (1991) report that of all ﬁrms receiving AAER's, top managers
are forced to resign or are ﬁred in seven out of ten cases, stockholders sue management
in eight out of ten cases, and the auditor is censured in four out of ten cases.

While violations concerning discount rates would most likely be handled
through the letter of comments process, the costs documented for all forms of SEC
action suggest that the perceived costs of SEC scrutiny would include loss of ﬁrm
value, deterioration of managerial reputation, and potential litigation. Other costs
could include the actual costs of responding to SEC investigations, the potential for the
SEC to scrutinize other areas of ﬁnancial reporting, lack of discretion in future years in
the area prompting the SEC's scrutiny, and degradation of ﬁnancial reporting
credibility“. Compliance with the SEC's recommendation concerning discount rates
should be increasing in these perceived and actual costs of SEC scrutiny.

While the relation between the costs of SEC scrutiny and compliance is
predicted to be positive, the inability to measure these costs precludes a direct test of
the prediction. However, a related prediction is that compliance is positively
associated with the potential for being an actual target of SEC investigation. Due to the
large number of reports ﬁled annually with the SEC and the disproportionately small
staff of reviewers, the SEC reviews only selected annual ﬁlings in a given year
[Skousen (1991)]. Thus, those ﬁrms that are more likely to be selected for such a
review should have a stronger incentive to comply. As a result, it is expected that
compliance with the SEC's recommendation is increasing in the potential for being a

target of SEC investigation. As discussed in chapter four, the interaction of total

 

‘1 These examples of potential costs were obtained in discussions with several audit
partners and managers of Big Six accounting ﬁrms who are involved with pension
accounting and/ or SEC communications.

23

assets and the discount rate before the recommendation will proxy for the potential for

being a target of SEC scrutiny. This leads to the following hypothesis:

[1,: Compliance with the SEC's recommendation is positively associated
with the potential for being a target of SEC investigation.

3.1.2 Leverage
During the late 1970's and early 1980's, researchers began to investigate the ability of

leverage to explain accounting policy/procedure choice [see Holthausen and Leftwich
(1983) and Watts and Zimmerman (1986) for a review of these studies]. These
investigations were based largely on the testable implications from agency theory as
described by Jensen and Meckling (1976) and were extensions of the introductory step
toward a positive theory of accounting taken by Watts and Zimmerman (1978). The
actual testing of leverage arose from the assumption that contracting costs are positive
in the negotiating, writing, evaluating, and renegotiating processes associated with debt
contracts. Researchers reasoned that as a ﬁrm moved closer to violation of its debt
covenants, the ﬁrm would choose or support accounting procedures to mitigate the
costs of such a violation. However, in lieu of the costly activity of obtaining data on
debt covenants and in an attempt to ascertain whether incurring these costs might prove
worthwhile in future tests, researchers included various measures of leverage as proxies
for closeness to violation of debt covenants. As a result, researchers actually tested the
following debt-equity hypothesis as Opposed to the more theoretical covenant-based
hypothesis:

24

Ceteris paribus, the larger a ﬁrm's debt/equity ratio, the more likely the ﬁrm's
manager is to select accounting procedures that shift reported earnings from
future periods to the current period [Watts and Zimmerman (1986, p. 216)].

The results of the early tests of the debt-equity hypothesis were consistent with
the predictions of the hypothesis. In their review of the growing positive accounting
literature, Holthausen and Leftwich (1983) concluded that leverage was one of two
ﬁrm-speciﬁc factors that were systematically related to accounting policy/procedure
choice. Watts and Zirmnerman (1986) concluded similarly, noting that the empirical
regularity of the explanatory power of leverage was a previously unrecognized
phenomenon. While both of these reviews provided informal evidence of the
descriptive validity of the debt-equity hypothesis, Christie (1990) provided a much
stronger conclusion by aggregating test statistics across the studies reviewed by
Holthausen and Leftwich (1983) and Watts and Zimmerman (1986). He found that
leverage was one of six variables with signiﬁcant explanatory power across studies.
Thus, the debt-equity hypothesis was found to be descriptive and generalizable.

Tests of the debt—equity hypothesis were based on the assumption that leverage
is an appropriate proxy for closeness to violation of debt covenants. Duke and Hunt
(1990) and Press and Weintrop (1990) both provide evidence that this assumption is
reasonable. Employing correlation analysis, Duke and Hunt (1990) ﬁnd a signiﬁcantly
positive relation between the existence of debt covenants and each of their seven
leverage proxies as well as a signiﬁcantly positive relation between their leverage
proxies and the closeness to debt covenants that are based on retained earnings. Press
and Weintrop (1990) provide similar results, concluding that the ”overall pattern of
results on tests of (debt covenants based on leverage, net worth, and working capital)

suggests that leverage level is associated with closeness to constraints" (p.81). Thus,

25

the assumption that leverage is an appropriate proxy for closeness to debt covenant
violation was validated.

Leverage has been included in investigations of pension accounting choice
because these choices affect both earnings and liabilities. Some researchers have found
leverage to be explanatory. Francis and Reiter (1987) ﬁnd a signiﬁcant relation
between adjusted funding ratios and leverage. Ali and Kumar (1993) ﬁnd that the
income effect from adoption is closely associated with closeness to debt covenants.
Other researchers, especially those investigating pension assumptions, have had no
success with leverage. Feldstein and Morck (1983) ﬁnd that leverage cannot explain
the assumed discount rate under APB 8. Similarly, Thomas (1988) ﬁnds that leverage
cannot explain the funding interest rate assumed by ﬁrms for statutory purposes.

There are two potential reasons for the failure of these latter two studies to
detect a relation between leverage and assumption choice. First, the studies focus only
on the level of the discount rate. In contrast, this study will focus on the change in the
discount rate brought on by the SEC's recommendation. Investigating the change is
advantageous because it places ﬁrms with varying discount rates on a comparable basis
by differencing away the previous discount rate. More importantly, it also allows for a
focus on the earnings and liability effects of a discount rate change instead of on the
discount rate itself. Second, both studies were conducted when APB 8 was in effect.
Ali and Kumar (1993) ﬁnd evidence consistent with the hypothesis that accounting
incentives have more inﬂuence under FAS 87 than under APB 8. Therefore, this study
is conducted in a time period under the accounting of FAS 87.

Because compliance will reduce earnings due to the increase in pension
expense, ﬁrms that are close to violation of debt covenants, as indicated by higher
leverage, should make more aggressive (i.e. , higher) pension assumptions to avoid the
decrease to earnings and subsequent tightening of debt covenants that would result from

full compliance. Stated differently, compliance with the SEC's recommendation is

26

expected to be negatively associated with the degree of leverage. As discussed in
chapter four, long-term debt divided by total assets will proxy for leverage. This leads

to the following hypothesis:

H,: Compliance with the SEC 's recommendation is negatively associated
with leverage.

3.1.3 Eundingratio

The pension funding ratio is deﬁned as the ratio of pension assets to pension
liabilities. This ratio has been the subject of several previous empirical studies [e.g.,
Feldstein and Morck (1983), Francis and Reiter (1987), and Harper and Strawser
(1993)]. Of most importance to the current study is Feldstein and Morck (1983) who
examine the relation between a ﬁrm's discount rate and its funding ratio. Standardizing
each sample ﬁrm's unfunded vested liability to the average discount rate chosen by all
their sample ﬁrms, the authors ﬁnd a signiﬁcantly positive relation between a ﬁrm's
discount rate and the size of its unfunded vested liability. This relation implies that
while ﬁrms with assets exceeding liabilities choose lower discount rates, ﬁrms with
liabilities exceeding assets choose higher discount rates to appear less unfunded.
Feldstein and Morck (1983) conclude that "(i)t is clear...that ﬁrms do engage in
strategic attempts to reduce their reported unfunded vested pension liabilities. . . . " (p.
202). In their investigation of pension funding strategies, Francis and Reiter (1987)
conclude similarly, noting that "ﬁrms may be using higher disclosure rates to create the
'appearance' of well funded pension plans" (p. 56).

Based on the above ﬁndings, it appears that ﬁrms manipulate their funding
ratios through pension assumption choices to give the impression of a better funded
plan. Proposed reasons for this behavior include avoiding the necessity of making
contributions to the pension fund [Feldstein and Morck (1983)], avoiding potential
political costs of having an unfunded plan [Francis and Reiter (1987)], and avoiding

27

negative perceptions of ﬁnancial condition [Harper and Strawser (1993)]. These
propositions from prior research suggest that the funding ratio signals the overall
quality of the pension plan, where lower quality plans are faced with additional costs in
the form of increased contributions, political scrutiny, and negative ﬁnancial statement
user perceptions.

If the true quality of a ﬁrm's pension plan is measured as the funding ratio
adjusted to some standard or common discount rate, then those ﬁrms with lower
adjusted funding ratios (i.e. , lower overall pension plan quality) should be hesitant to
comply with the SEC's recommendation, for compliance will reveal their plan's
inferior quality. In contrast, those ﬁrms with higher adjusted funding ratios (i.e.,
higher overall pension plan quality) should be willing to comply to reveal their plan's
superior quality. Thus, compliance should be increasing in the adjusted funding ratio,

which leads to the following hypothesis:

H,: Compliance with the SEC's recommendation is positively associated
with the adjusted funding ratio.

3.2 Smoothing the effect of compliance

Patterns associated with reported earnings appear to be a focal point of many
ﬁrm managers. As Barth, Elliott, and Finn (1992) note, many ﬁrms devote
considerable space in their annual reports to emphasize their earnings patterns. This
emphasis on earnings patterns is puzzling, for there is no current theory of ﬁrm value
that explicitly incorporates earnings patterns. In an attempt to explain why ﬁrms would
be concerned with earnings patterns, Barth et al. (1992) suggest three plausible
scenarios that would make earnings patterns value relevant. Firms could be rewarded
in the market through the correlation of their earnings patterns with growth,
persistence, and risk; through the market's increased ability to predict future earnings

due to the consistent earnings patterns; or through the private information revealed by

28

the manager through the earnings pattern. While it is unclear which of these three
scenarios, if any, is valid, Barth et al. (1992) ﬁnd that the market assigns a premium to
the value of ﬁrms that exhibit a non decreasing earnings pattern for a period of at least
ﬁve years duration. Thus, not only do ﬁrms appear to be concerned with earnings
patterns, but the market may eventually reward such behavior as well.

Research on earnings patterns is not new. The smoothing hypothesis, which
can be traced back to as early as Hepworth (1953), has been the subject of much
accounting research in the last few decades. The hypothesis predicts that managers
choose accounting procedures to create stable earnings streams. Ronen and Sadan
(1981) explain that income smoothing behavior, broadly speaking, is an attempt to
signal information to users of ﬁnancial statements. This broad deﬁnition is consistent
with Barth et al. '8 (1992) third scenario. However, Ronen and Sadan (1981) note that
the more operational deﬁnition used in accounting research is that smoothing is a
means to dampen ﬂuctuations of earnings around some normal or expected level.

Much of the research on the smoothing hypothesis has been an attempt to
demonstrate the descriptive validity of the hypothesis. Several researchers have been
successful in doing so [see Ronen and Sadan (1981) for a detailed review of early
smoothing studies]. For example, Hand (1989) ﬁnds that ﬁrms smooth transitory
decreases in earnings through the recognition of "paper gains" associated with debt-
equity swaps. Other researchers have analytically explored the motivations for
smoothing, suggesting that smoothed earnings increase ﬁrm value through decreased
perceived bankruptcy risk [Trueman and Titman ( 1988)], enhanced perceived value
from potential shareholders [Dye (1988)], and clearer perceptions of managers'
expectations of future cash ﬂows [Sub (1990)]. Still other researchers have attempted
to use a costly contracting framework to explain why ﬁrms smooth earnings [e.g. ,
Moses (1987) and Beattie et al. (1994)]. While these studies ﬁnd associations between

smoothing behavior and variables such as ﬁrm size, existence of bonus plans, earnings

29

variability, and outside ownership concentration, a consistent set of explanatory
variables for smoothing has yet to be developed.

In the area of pensions, smoothing issues have surfaced in both practice and
research. In practice, the most notable area of smoothing concerns is found in the
development stages of PAS 87 . As stated earlier, practicing professionals who
responded to the discussion memorandums and exposure drafts for FAS 87 were
concerned about potential volatility in pension expense and liabilities. Some of these
respondents suggested that an averaging method with respect to actuarial assumptions
could be used to smooth potential effects from year-to-year changes in such
assumptions. It is clear that the issue of smoothed expenses (or lack thereof) associated
with pensions has been a point of interest since the beginning of FAS 87.

In pension research, smoothing has been investigated directly. For example,
Ali and Kumar (1993) ﬁnd that the income effect from adoption of FAS 87 was used to
smooth income. Although not addressing smoothing issues, other pension research has
produced results that are consistent with smoothing. For example, Bodie, Light,
Morck and Taggart (1985) ﬁnd that more proﬁtable ﬁrms choose lower discount rates.
While the authors characterize this ﬁnding as ﬁrms attempting to build ﬁnancial slack
in the form of over funded pension plans during more proﬁtable years, another
plausible explanation is that ﬁrms who are more proﬁtable choose lower rates to
decrease earnings while ﬁrms who are less proﬁtable choose higher rates in order to
boost earnings.

The SEC's recommendation is another setting in which to investigate smoothing
behavior. Because some managers are more reluctant to bear the perceived costs of
reporting reduced earnings, some ﬁrms that reduce their discount rate in compliance
with the recommendation should be expected to choose other accounting practices to
offset the reduction in earnings caused by the change in the discount rate. One option

available to these ﬁrms is the choice of another pension assumption, the salary

30

progression rate. The expectation of this type of smoothing behavior is supported by
Ronen and Sadan (1981) who observe that smoothing can be accomplished through
changes in pension actuarial assumptions. As illustrated in section 2.1.3, while an
identical decrease in the salary rate will not completely offset a decrease in the discount
rate, as the discount rate is reduced, the salary rate can also be reduced to partially
offset the negative effect on earnings from the decrease in the discount rate. Therefore,
of those ﬁrms that comply, those with incentives to report smoothed earnings should be
more likely to smooth the effect of compliance by decreasing the salary progression

rate. This leads to the following hypothesis:

H,: Firms with incentives to report smooth earnings who lower their
discount rate in compliance with the SEC are more likely to lower
their salary progression rate.

3.3 Summary

This chapter contains the development of the four hypotheses that arise as a
result of the SEC’s recommendation. Compliance with the recommendation is posited
to be increasing in the potential for being a target of SEC investigation, decreasing in
leverage, and increasing in the adjusted funding ratio. In addition, ﬁrms that both
comply and have incentives to report smoothed earnings are posited to decrease the
salary progression rate to offset the effect of compliance. The research design

employed to test these four hypotheses is explained in chapter four.

Chapter Four
RESEARCH DESIGN

This chapter details the research design employed to test hypotheses one
through four. Section 4.1 summarizes the process of selecting sample ﬁrms. Sections
4.2 and 4.3 describe the tests for compliance and smoothing, respectively. Section 4.4
summarizes the chapter.

4.1 Sample selection

The sample selection process begins with all ﬁrms on Standard and Poor’s
Compustat Annual Tapes (1993 version). Firms from this population are included in
the ﬁnal sample if they meet the following requirements mcessary to test hypotheses
one through four. First, ﬁrms must sponsor a pension plan subject to the reporting
requirements of FAS 87. Second, because discount rates are sensitive to interest rates,
all sample ﬁrms must face the same economic conditions; thus, sample ﬁrms must have
a December year-end. Imposing these two restrictions yields 1,530 ﬁrms. This subset
is further reduced by 289 ﬁrms by eliminating all ﬁrms in the ﬁnancial industry (SIC
codes 6000-6999). An additional 490 ﬁrms are eliminated from the sample because
they do not have non-missing discount rate and salary rate information for both 1992
and 1993. Finally, so that the sample is comprised of ﬁrms that are faced with a
decision to comply or not with the SEC’s recommendation, all ﬁrms with 1992
discount rates below eight percent are removed. Eight percent is chosen as the cutoff
because full compliance to the seven percent lower bound of the SEC’s range would
require a decrease of one percent, an accounting change that can have a signiﬁcant

impact on the ﬁnancial statements [Winklevoss (1993)]. This ﬁnal restriction results in

31

32

a loss of 151 ﬁrms and leaves a ﬁnal sample of 600 ﬁrms”. Table 1 summarizes the

sample selection process.

Table 1
Sample selection

 

 

 

Firms Firms

Requirements for inclusion in ﬁnal sample Lost Remaining
Firm must sponsor a pension plan subject to FAS 87 and

have a December year-end 1,530
Firm must have a SIC code other than 6000-6999 289 1,241
Firm must have non-missing discount and salary rates for

1992 and 1993 490 751
Firm must have a 1992 discount rate greater than or equal

to eight percent 151 600

 

 

 

 

4.2 Tests of compliance

Hypotheses one, two, and three posit that compliance with the SEC's
recommendation is contingent on the potential for being a target of SEC scrutiny,
leverage, and the adjusted funding ratio, respectively. These hypotheses are tested via
multivariate regression. Dependent, independent, and control variables used in the
regression are discussed below.

4.2.1 Dependenuariable

The change in a ﬁrm’s discount rate from 1992 to 1993 proxies for compliance

with the SEC’s recommendation and serves as the dependent variable. Given that the

 

12 Descriptive information on the 151 ﬁrms lost because their 1992 discount rates are
below eight percent is reported in chapter ﬁve.

33

SEC established a target range to which ﬁrms could move their discount rates in
compliance, measuring compliance as the raw change in the discount rate requires a
control for the level of the discount rate before the change. Such a control is needed
because a ﬁrm moving from eight to seven percent complies the same as a ﬁrm moving
from nine to seven percent although the measure of compliance is different. Therefore,
the discount rate chosen in 1992 is included in the regression to account for a ﬁrm’s
discount rate before the SEC’s recommendation.

Measuring compliance as the change in the discount rate also raises the concern
that compliance is mis-measured by some amount of change that would have been made
without the SEC’s recommendation. Ideally, the expected change could be estimated
by projecting the 1993 discount rate via longitudinal analysis of past discount rates as
compared to factors such as the historical interest rates on Aa corporate bonds.
However, an analysis of this type is precluded by the lack of a sufﬁciently long data
series. Only six observations (1987-1992) of the discount rate are available for most
ﬁrms, as FAS 87 was not required to be adopted until 1987. An alternative to
projecting a ﬁrm speciﬁc discount rate is to estimate an expected change for all ﬁrms
given macro economic conditions. However, subtracting such an expected change
from the actual change only alters the compliance measure by a scalar, the effect of
which goes to the intercept term of the regression. Thus, compliance is measured as
the entire change in the discount rate.

4.2.2 Independenuariables

In order to test hypothesis one, a proxy for the potential for being a target of
SEC scrutiny must be employed. As mentioned in chapter three, the SEC reviews
selected annual ﬁlings in a given year due to the large number of reports ﬁled annually
with the SEC and the disproportionately small staff of reviewers. However, it is not
clear which ﬁrms are more likely to be selected for such a review. Feroz et al. (1991)

state that the SEC chooses candidates for formal investigation partly on the value of the

34

message that it can send and partly on the probability that it can be successful in the
investigation. If the SEC investigates certain registrants in order to send a message to
all registrants, the value of that message is assumed to be most likely increasing in ﬁrm
size. If the probability of success is also an important determinant of which ﬁrms are
investigated, then the SEC should be expected to target those ﬁrms that are abusing the
discount rate choice the most (i.e. , those ﬁrms choosing the highest discount rates).
Firms that meet both criteria should therefore have the highest potential for being a
target of SEC scrutiny. Thus, the interaction between a ﬁrm’s total assets and it’s
assumed discount rate for 1992 is included in the regression as the proxy for the
potential for being a target of SEC investigation if a ﬁrm chooses not to comply. Its
coefﬁcient is expected to be negative.

As mentioned in chapter three, many measures of leverage have been employed
in prior research. The measure used for hypothesis two is long term debt divided by
total assets. This measure is consistent with Francis and Reiter (1987), who ﬁnd a
signiﬁcant relation between leverage and funding strategies. The coefﬁcient on this
variable is expected to be positive.

For purposes of hypothesis three, the adjusted funding ratio is deﬁned as the
ratio of pension plan assets to the Accumulated Beneﬁt Obligation (ABO), where the
ABC is "adjusted" to seven percent. The ABC is chosen because if the ABO exceeds
plan assets, a minimum liability must be created that often increases the liability that a
ﬁrm reports on its balance sheet. Also, Barth (1991) demonstrates that market prices
are most closely associated with the combination of the ABC and plan assets.
Therefore, ﬁrms are expected to put most of their emphasis on the ABO.

To calculate the adjusted funding ratio, the ﬁrst step is to calculate what the
ABO would be if a ﬁrm used a seven percent discount rate. Previous research has
employed a linear transformation to obtain an adjusted liability measure [Feldstein and
Morck (1983), Bodie et al. (1985), Francis and Reiter (1987)]. This transformation is

35

accomplished by multiplying the unadjusted ABO by the ratio of the common discount
rate to the disclosed discount rate. The same transformation is employed in this study
with the 1992 ABC and the 1992 discount rate serving as the unadjusted ABC and the
disclosed discount rate, respectively. Once the adjusted ABC is calculated, then the
adjusted funding ratio is determined by dividing 1993 plan assets by the adjusted ABO.
Using 1993 plan assets allows the adjusted funding ratio to measure how well funded
the plan would be on an adjusted basis given contributions to the pension plan during
1993. A ratio less than (greater than) one implies that the pension plan is under (over)
funded. The coefﬁcient on this variable is expected to be negative.

4.2.3 Connolxariables

Previous research on pensions under the guidelines of APB 8 suggests that
capital availability, a proxy for the ability to fund the plan, impacts pension accounting
choice to some degree [e.g., Bodie et al. (1985), Francis and Reiter (1987), and Healy
and Palepu (1990)]. Based on arguments here and elsewhere [Ali and Kumar (1993),
Blankley and Swanson (1995)], funding and accounting decisions can be separated from
each other under the guidelines of FAS 87. Thus, the impact of funding considerations
on accounting choice is not as strong under FAS 87 as under APB 8. However, there
is no doubt that accounting decisions under FAS 87 have at least an indirect effect on
funding decisions. For example, the accrued pension liability that is reported on the
balance sheet is the difference between the cumulative amount of pension expense
recognized to date and the cumulative amount contributed to the plan to date. Thus, if
a ﬁrm does not wish to increase this liability, then as pension expense increases due to
compliance, funding must be increased as well. However, if the ﬁrm lacks the
available capital to increase funding, then compliance may be reduced. This implies
that capital availability may impact the degree of compliance with the SEC's
recommendation. Consistent with Francis and Reiter (1987), capital availability is

measured as cash ﬂows from operations less dividends and capital expenditures.

36

Another factor that may impact compliance is the materiality of the pension
plan. If a ﬁrm’s pension plan is immaterial to its overall ﬁnancial position, then the
ﬁrm could be more likely to comply with the SEC’s recommendation, for the ﬁrm
incurs little cost from choosing a lower discount rate. However, if a ﬁrm’s pension
plan is material to its overall ﬁnancial position, then the ﬁrm could be less likely to
comply.

Because pension plans affect both income and liabilities, the materiality of the
pension plan to overall ﬁnancial position is measured in two ways. The ﬁrst measure
focuses on the size of the plan relative to the size of the ﬁrm and is measured as the
adjusted ABO divided by total assets. The second measure focuses on the impact of
pension expense on net income and is measured as the absolute value of pension
expense divided by the absolute value of net income before pension expense. Absolute
values are taken so that the measure will be non-negative for all ﬁrms; otherwise, ﬁrms
with pension income instead of pension expense or a net loss instead of net income
would have a negative value that would misrepresent the impact of pension expense on
net income. Both measures are included in the regression, but a directional prediction
is made only for the former. Firms with smaller plans relative to ﬁrm size are
expected to be more likely to comply with the SEC’s recommendation. In other words,
the coefﬁcient is expected to be positive.

Other factors that may impact compliance are plan characteristics such as the
structure of beneﬁts (ﬁxed versus percentage of wages) and the group of employees
covered (hourly versus salary). These characteristics may be uniform across industries
[Francis and Reiter (1987)]. Therefore, an industry dummy variable is included to
control for the impact on compliance of systematic differences between plan
characteristics. If this variable also proxies for the degree of unionization in ﬁrms, it

will control for the effects unions may have on compliance. Two digit SIC codes will

37

be used for the dummy variable. This leads to the following regression to be estimated

for tests of hypotheses one, two, and three:

ADR, = a + mun, + BZSCRUTINY, + B3LEV, + [3,,AFR, + BSCAPITAL,
+ BGPSIZE, + [37%, + 13st..., + a,

where
ADR Change in the discount rate from 1992 to 1993;
DR 1992 discount rate;
SCRUTINY Interaction between DR and 1992 total assets;
LEV 1992 total long-term debt divided by 1992 total assets;
AFR 1993 total pension assets divided by 1992 adjusted ABO;
CAPITAL 1992 cash from operations less capital expenditures and dividends;
PSIZE 1992 AFR divided by 1992 total assets;
PE Absolute value of 1992 pension expense divided by the absolute
value of 1992 net income before pension expense;
DIndustry Dummy variable based on two-digit SIC code;
i Firm index.

4.3 Tests of smoothing

Hypothesis four posits that ﬁrms with incentives to report smoothed earnings
will decrease their salary rate in order to partially offset the decrease in earnings
resulting from the decrease in the discount rate. Like hypotheses one through three,
hypothesis four is tested with a multivariate regression. Discussions of the dependent,
independent, and control variables follow below.

4.3.1 Dependeanai'iable

Testing of hypothesis four requires an estimation of the expected change in the
salary rate in the absence of the SEC's recommendation. The difference between this
expected change and the actual change is the amount of smoothing conducted by ﬁrms
and serves as the dependent variable for tests of hypothesis four. This smoothing
measure differs slightly from usual tests of smoothing in that it is not measured in

terms of earnings. However, due to the difﬁculty in estimating the earnings impact of

38

a change in the salary rate and assuming that the earnings impact is positively
associated with the magnitude of the change in the salary rate, the change itself is used
in this study as the smoothing measure.

According to FAS 87, the salary progression rate is comprised of three potential

sources of salary increases: merit, productivity, and inﬂation. As the standard states,

(m)erit increases are those that an individual employee will receive as
that employee progresses through a career.... The second component is
labor's share of productivity gains. The third component attempts to
anticipate general compensation increases that result from inﬂation [par.
203, footnote 16].

Before discussing each component, emphasis should be placed on the salary rate being
an estimate of the future growth in salaries. By nature the rate represents a long term
outlook incorporating all years of service for all employees until retirement. Thus, the
expected salary rate for any given year can be thought of as the previous year's rate
plus or minus any revisions in long term expectations. As introduced above, these
revisions can be associated with merit, productivity, and/or inﬂation.

With this long term outlook, there is little reason to expect a change in the merit
and productivity components for 1993. Winklevoss (1993) states that while the rate of
increase due to merit generally decreases with an employee's age, merit increases on
the whole have little effect on aggregate payroll over time if the distribution of ages
and services among employees remains relatively stable. This statement implies that in
the absence of some ﬁrm speciﬁc layoff or hiring of a certain age group of employees,
the merit component of the 1993 salary rate should be roughly the same as that
component of the 1992 rate. Winklevoss (1993) also states that estimating potential
salary increases due to productivity is very difﬁcult and that the importance of this
component may have decreased over the years. Thus, absent the enactment or

elimination of a proﬁt-sharing plan or the development of a new product that

39

guarantees increased future proﬁtability and thus implies the ability to share more
proﬁts with labor via increased salaries, the productivity component of the 1993 salary
rate should also be roughly the same as that component of the 1992 rate.

Of the three components, the inﬂation factor has the most signiﬁcant impact on
future salaries and is most likely represented by a constant compound rate that is
consistent with long-term average inﬂation rates [Winklevoss ( 1993)]. Thus, short
term ﬂuctuations in inﬂation rates should not affect the salary rate as long as those
ﬂuctuations do not revise long term perspectives of inﬂation. Even if ﬁrms adjust
salary rates based on yearly ﬂuctuations in inﬂation rates, the rate of inﬂation during
1992 and 1993, as indicated by the average annual percentage change in the Consumer
Price Index for all urban consumers (CPI), was stable at 3.0%. Thus, like the merit
and productivity components, there is little if any reason to expect a change in the
inﬂation component of the salary rate. As a result, the entire change from 1992 to
1993 is considered unexpected and is characterized as an attempt to smooth the
earnings effect from compliance.

4.3.2 Indenendentxariables

The ﬁrst independent variable included in the regression is the change in the
discount rate. As discussed earlier, the change in the discount rate proxies for the
degree that a ﬁrm complies with the SEC ’5 recommendation. If a ﬁrms wants to
smooth the effect of compliance, then a large decrease in the discount rate should cause
a large decrease in the salary rate. Thus, a positive coefﬁcient is expected and is
consistent with smoothing behavior.

Additional independent variables that potentially distinguish between ﬁrms that
are more or less likely to smooth earnings are also included in the regression. One of
these variables is the ownership structure of the ﬁrm. Smoothing can be viewed as an
earnings management technique attempting to signal information to reduce the

information asymmetry that persists between managers and owners [Schipper (1989)].

40

As a result, assuming that information asymmetry decreases with the degree of
concentration in ownership, those ﬁrms with more concentrated ownership will be less
likely to smooth earnings. Results from Beattie et al. (1994) and Bange and De Bondt
( 1995) are consistent with this reasoning. Thus, to control for the impact ownership
concentration has on the incentive to smooth the effect from compliance, the ratio of
common stock held by institutional investors to total common stock outstanding, a
measure consistent with Bange and De Bondt (1995), is included in the regression.
Institutional investor data is obtained from Moody ’s Handbook of Common Stocks and
Moody ’s Handbook of Nasdaq Stocks. The coefﬁcient on this variable is expected to
be positive.

A third variable included in the regression is the surprise in earnings. If ﬁrms
want smooth earnings streams, then those ﬁrms with negative unexpected earnings
should be more inclined to smooth the effect of compliance while those ﬁrms with
positive unexpected earnings should be less inclined to offset the reduction to earnings
caused by compliance. As discussed in chapter two, changes in actuarial assumptions
in 1993 affect earnings in 1994. As a result, ﬁrms do not know the amount of
unexpected earnings at the time the assumptions are made. However, assuming that
ﬁrms have perfect foresight with respect to at least the direction of the 1994 unexpected
earnings, then a dummy variable can be included in the regression to represent the sign
of the 1994 unexpected earnings. Unexpected earnings is deﬁned as the change in
earnings before pension expense. The dummy variable takes the value of one if the
1994 unexpected earnings is negative and zero otherwise. The coefﬁcient on this
variable is expected to be positive.

Two other general variables found to be associated with smoothing that are
included in the regression are ﬁrm size as a proxy for potential political costs [Moses

( 1987)] and past earnings variability as a proxy for past propensity to smooth earnings

41

[Beattie et al. (1994)]. Firm size is measured as total assets. Past earnings variability
is measured as the standard deviation of the past ﬁve years of earnings.

4.3.3 CentmLiiatiables

Four additional control variables are included in the regression analysis. First,
the expense associated with a ﬁrm’s pension plan may be immaterial with respect to the
ﬁrm’s total earnings. As a result, a change in the discount rate for such a ﬁrm would
have a small relative effect on the ﬁrm's aggregate condition. Thus, to control for the
impact on earnings of a change in the discount rate, the absolute value of pension
expense divided by the absolute value of net income before pension expense is included
in the regression. As in section 4.2.3, absolute values are taken so that pension income
or a net loss will not cause the measure to misrepresent the true materiality of the
pension plan”. The coefﬁcient on this variable is expected to be negative, as ﬁrms
with plans that impact their earnings to a greater extent would be expected to smooth
the impact of compliance to a greater degree.

Second, in the 1990’s ﬁrms have continued to face increased national and global
competition. As a result, ﬁrms have often been forced to “down-size” or “right-size”
in order to stay competitive. Along with these restructuring efforts, ﬁrms may also
reduce future salary expectations, which would be reﬂected in the salary progression
rate. While it is not clear that any revisions in future salary expectations that might
coincide with these restructurings should occur in 1993 to any greater extent than in
any prior or future years, inclusion of a control variable for decreased salary rates due
to downsizing seems appropriate. Thus, a crude proxy for the extent that a ﬁrm is

down-sizing and reducing future salary expectations due to competition, the change in

 

13 Negative pension expense (i.e., pension income) can result if current service and
interest costs are small compared to large expected returns on investments and/or large
amortizations of actual gains from investments and prior service cost. The regression
testing hypothesis four is estimated both with and without ﬁrms with negative pension
expense. Chapter ﬁve reports those results.

42

the number of employees, is included in the regression. The coefﬁcient on this
variable is expected to be positive.

Third, ﬁrms with a higher 1992 salary rate relative to other ﬁrms may have
more leeway to decrease their salary rate. Therefore, the salary rate chosen in 1992 is
included to account for the level of the salary rate before any change.

Finally, plan characteristics such as the structure of beneﬁts (ﬁxed versus
percentage of wages) and the group of employees covered (hourly versus salary) may
impact the degree in which ﬁrms change the salary rate. As stated in section 4.3.2,
these characteristics may be uniform across industries [Francis and Reiter (1987)].
Therefore, an industry dummy variable is included to control for the impact of
systematic differences between plan characteristics on the change in the salary rate. If
this variable also proxies for the degree of unionization in ﬁrms, it will control for the
effects unions may have on compliance. Unions recently have increased scrutiny of
pension plans. They most likely would be displeased with a reduction in assumed
future salary increases. Two digit SIC codes will be used for the dummy variable.

This yields the following regression to be estimated for tests of hypothesis four:

43
ASR, = a, + mun + BZINSOWN, + B3UEARN, + [3431215. + BSVAR, +
36%, + B7DOWNSIZE, + 888R, + mommy + e,-

where
ASR Change in the salary rate;
ADR Change in the discount rate;
INSOWN Common shares held by institutional investors divided by total

shares;
UEARN Dummy variable taking the value of one if the 1994 earnings surprise
is negative, zero otherwise;

SIZE Total assets;
VAR Standard deviation of earnings over the past ﬁve years;
PE Absolute value of pension expense divided by the absolute value of

net income before pension expense;
DOWNSIZE Change in employees;

SR 1992 salary rate;
Dlndustry Dummy variable based on two-digit SIC code;
i Firm index.

4.3.4 Additionalsmoethinuests

As explained in chapter two, even though pension actuarial assumptions are
chosen explicitly and are therefore independent from one another, the salary rate is
nonetheless related to the discount rate in that each incorporates expectations of the
same economic environment (FAS 87, par. 195). Therefore, an alternative hypothesis
to the claim that the above coefﬁcient on the change in the discount rate is evidence of
smoothing behavior is that the coefﬁcient simply represents the positive correlation
inherent in the two rates due to their common component. One way to attempt to rule
out this alternative hypothesis is to examine the relation between the salary rate and the
discount rate in prior years and compare the current year’s relation to the historical
trend. If the relation between the salary rate and the discount rate is no different in
1993 than in prior years, then any claim of smoothing activity is questionable.

However, if the salary rate is reduced in conjunction with the decrease in the discount

44

rate in a more pronounced way in 1993 than in past years, then such a claim for
smoothing activity is still warranted.

To ascertain which of the above two scenarios is true, a regression equation
with the following variables is estimated. The dependent variable is the change in the
salary rate. The independent variables are the change in the discount rate, a dummy
variable taking the value of one if the year is 1993 and the value of zero otherwise, and
the interaction of the two. If the relation between the two rates is more pronounced in
1993, then the coefﬁcient on the interaction term should be positive and signiﬁcant. A
fourth independent variable, the change in the CPI, is also included in the regression.
It is included for two reasons. First, it controls for any impact that changes in short-
terrn inﬂation may have on yearly changes in the salary rate. Second, and perhaps
more importantly, it helps to determine the appropriateness of the expectation
developed in section 4.3.1 for the change in the salary rate. A signiﬁcant coefﬁcient
on the change in the CPI indicates that salary rates are changed in accord with short-
term inﬂationary changes and, since the change in the CPI was zero in 1993, implies
that an expectation of no change in the salary rate for 1993 is appropriate.

Data for the regression analysis is limited to the years 1987 through 1993, for
adoption of FAS 87 was not required until 1987. The examination is also limited to a
subset of ﬁrms identiﬁed from the ﬁnal sample. This latter limitation of the data set is
necessary because discount and salary rate assumptions must be collected by hand from
the pension footnote in annual reports or 10-K ﬁlings from 1987 to 199014. The subset
of ﬁrms for the regression is comprised ﬁrst of 200 ﬁrms chosen at random and is

supplemented with an additional 77 ﬁrms collected in an earlier pilot study, yielding a

 

14 Although other pension data items disclosed in the footnotes such as plan assets,
components of pension expense, and liability measures are reported on the Compustat
tapes prior to 1991 , pension actuarial assumptions prior to 1991 are not reported.

45

total subset of 277 ﬁrms with pension assumption data from 1987 to 1993. These 277

ﬁrms are used to estimate the following regression:

ASRa = 0a.. + BiADRu + 13291993 + BaADR..* D1993 + B4ACPI. + 8...

where
ASR Change in the salary rate;
ADR Change in the discount rate;
D1993 Dummy variable taking the value of one if 1993; zero otherwise;
ACPI Change in the Consumers’ Price Index;
i Firm index;
t Year index, 1988-1993.

4.4 Summary

This chapter contains a description of both the sample selection process and the
research design employed to test hypotheses one through four. The ﬁnal sample
consists of 600 ﬁrms. All four hypotheses are tested via multivariate regressions.
Descriptive information about the sample and results of the regressions are reported in

chapter ﬁve.

Chapter Five
RESULTS

This chapter presents the results of the tests speciﬁed in chapter four. Section
5.1 presents descriptive statistics for the sample. Sections 5.2 and 5.3 describe the
results for the compliance and smoothing hypotheses, respectively. Section 5 .4
summarizes the chapter.
5.1 Descriptive Statistics

As stated in the introduction, one of the ﬁrst empirical questions arising from
the SEC's recommendation is whether such a recommendation contained enough
inﬂuence to alter accounting choice. The descriptive information reported in Tables 2
and 3 suggests that the recommendation did indeed affect accounting choice. Table 2
contains 1992 and 1993 information for the full sample of 600 ﬁrms. Most noticeable
in the table is the large decrease in the mean 1993 discount rate, down from 8.39% in
1992 to 7.45% in 1993. This decrease is statistically signiﬁcant at p< .01 and is
consistent with the conclusion that the SEC's recommendation had a signiﬁcant impact
on discount rate choice. The information in Table 3 , which reports changes in the
discount rate, salary rate, and rate of return for the years 1988 through 1993 for the
subset of 277 ﬁrms, yields additional support for this conclusion. As seen in panel A,
the change in the discount rate in 1993 is much more pronounced than in all prior
years, both in the magnitude of the change and in the percentage of ﬁrms making a
change. This result illustrates that the change in the discount rate for 1993 was well
outside the historical trend, suggesting the introduction of an outside inﬂuence in 1993,

namely the SEC recommendation, which signiﬁcantly altered accounting choice”.

 

‘5 Data from a random sample of 306 ﬁrms in Blankley and Swanson (1995) illustrate
the same trend in discount rates. The authors focus on the overall relation between
discount rates and interest rates for the time period 1987 through 1993 and treat the
SEC’s recommendation as one factor affecting that relation. The authors conclude that

46

47

Table 2 also shows a decrease in the mean salary rate, down from 5.55% in
1992 to 4.93% in 1993. This decrease is statistically signiﬁcant at p< .01 and is
consistent with the existence of smoothing behavior within the salary rate choice.
Information in Table 3 also is consistent with this conclusion. As seen in panel B and
similar to the change in the discount rate, the change in the salary rate in 1993 is much
more pronounced than in all prior years, both in the magnitude of the change and in the
percentage of ﬁrms making a change. This break from the historical trend is consistent
with the hypothesis that the salary rate was changed to compensate for the change in
the discount rate.

The adverse impact on the ﬁnancial statements of such a large decrease in the
discount rate can be seen in the ABC and the funding ratio reported in Table 2. The
median ABO increased approximately 25% in 1993, up from $53 million to $66
million. While this increase is not statistically signiﬁcant, when compared to the
average percentage increase for the years 1988 through 1992 of 4% (not reported in
Table 2), the increase in 1993 is consistent with the reduction in the discount rate
having a substantial impact on pension liabilities. Moreover, the difference between
the two percentages (21%) is consistent with Winklevoss' (1993) rule-of-thumb of a
20% increase in the pension liability for every 1% decrease in the discount rate. When
combined with a lower percentage increase in contributions, the increase in the ABO
produced a signiﬁcant decrease (p < .01) in the mean funding ratio, down from 1.30 to
1.18. Thus, the decrease in the discount rate clearly had an adverse impact on the
ﬁnancial statements.

Also evident in Table 2 is the fact that pension plans have a material effect on

the overall ﬁnancial position of the ﬁrm. Pension expense for the median ﬁrm was

 

discount rates were slow to reﬂect the decline in interest rates in the early 19905;
however, they do not attribute the decline to the SEC’s recommendation.

48

Table 2
Descriptive statistics for 600 sample ﬁrms

 

 

 

 

 

 

 

 

 

 

 

 

 

Std
Variable Yr N Mean Dev 25% Med 75%
Discount rate 92 600 8.39 .367 8.00 8.50 8.50
93 600 7.45“ .400 7.25 7.50 7.50
Salary rate 92 600 5.55 .802 5.00 5.50 6.00
93 600 4.93“ .803 4.50 5.00 5.50
Rate of return 92 582 9.19 .891 8.50 9.00 9.75
93 581 8.99“ .894 8.50 9.00 9.50
Accumulated Beneﬁt 92 593 499 1856 15 53 270
Obligation (ABC) 93 592 572 2083 17 66 319
Plan assets 92 597 643 2538 18 69 368
93 596 692 2722 20 75 371
Reported funded 92 593 1.30 .439 1.08 1.25 1.49
status 93 592 1.18“ .360 .97 1.14 1.33
Pension expense divided by 92 594 .24 .812 .03 .08 .16
net income before expense 93 592 .34 1.09 03 .07 .17
Plan assets divided 92 597 .16 .171 .06 .11 .21
by total assets 93 596 .17 .187 06 .11 .22
Leverage 92 600 .28 .207 14 .28 .37
93 600 .27 .200 .12 .26 .36
Capital 92 591 -41 403 -34 -1 13
93 590 -16 367 -27 -1 20
Total assets 92 600 3596 1 1270 195 723 2935
93 600 3795 13074 205 725 3189

 

 

"PM 1993 value is signiﬁcantly different than 1992 value at less than the .10(.05)[.01] level

 

49

Table 3

Changes in disclosed actuarial assumptions, by year

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

n=277
Panel A: Change in discount rate
1988 1989 1990 1991 1992 1993
mean 0.17c 017‘ 0.04‘ -0.16° 014C 0%“
median 0.00 0.00 0.00 0.00 0.00 -1.00
st dev 0.46 0.45 0.36 0.41 0.33 0.48
n<0 19 77 29 88 78 254
n>0 58 14 57 21 10 0
n 221 226 228 228 252 277
Panel B: Change in salary rate
1988 1989 1990 1991 1992 1993
mean 0.01 -0.05" -0.06° -0.16c -0.17c —0.63c
median 0.00 0.00 0.00 0.00 0.00 050
st dev 0.47 0.38 0.35 0.53 0.47 0.69
n<0 17 24 25 56 51 173
n>0 23 12 12 8 5 2
n 216 222 225 227 253 277
Panel C: Change in rate of return
1988 1989 1990 1991 1992 1993
mean 0.18“ 0.09c 0.06" 0.05" -0.08" -0.22“
median 0.00 0.00 0.00 0.00 0.00 0.00
st dev 0.53 0.39 0.36 0.36 0.37 0.48
n<0 11 11 16 14 36 76
n>0 50 31 33 29 12 10
n 222 226 227 229 248 266

 

 

 

 

 

 

 

“b"“l signiﬁcantly different from zero at less than the .10(.05)[.01] level

 

50

about 7% (8%) of net income before pension expense for 1992 (1993). In addition,
plan assets for the median ﬁrm equaled about 11% of total assets for both years.

In summary, descriptive statistics on sample ﬁrms are consistent with the SEC’s
recommendation having a signiﬁcant effect on accounting choice. Discount rates were
reduced almost a full percentage point for the average ﬁrm, which resulted in a
signiﬁcant decrease in the reported funded status of the average pension plan. Salary
rates were also reduced more than a half of a percentage point, an event consistent with
smoothing behavior in accounting choice. Further examination of the changes in the
discount and salary rates via regression analysis follows below.

5.2 Results for tests of compliance

The second empirical question arising from the SEC's recommendation is
whether there were any factors that systematically affected ﬁrm compliance. In chapter
three, hypotheses one through three were developed from this underlying question.
Tables 4 and 5 contain evidence helpful in answering this question.

Table 4 reports the Pearson correlation coefﬁcients for the independent
variables used in the multivariate regression estimated to test hypotheses one through
three. High correlations between independent variables indicate the presence of
multicollinearity, which reduces the power of t-tests [Kennedy (1989)]. None of the
independent variable coefﬁcients in Table 4 approach 0.8 to 0.9, the benchmark
Kennedy (1989) gives for a high degree of multicollinearity. Thus, it appears as if
multicollinearity is not a major problem in the data set.

In addition to providing assurance that multicollinearity is not a major problem,
Table 4 reveals that several correlations between variables are signiﬁcantly different
from zero. Of main interest are those correlations involving DR, the 1992 discount

rate. The 1992 discount rate is signiﬁcantly correlated with three independent

Correlation coefﬁcients

51

Table 4

 

 

n=600

DR SCRUTINY LEV AFR CAPITAL PSIZE PE
D“ 1.00
SCRUTINY .08" 1.00
”‘3" .02 -.06 1.00
A“ -.21° -.01 -.01 1.00
“mm -.05 25° -.08’ .03 1.00
”'25 .11° .02 -.14° -.19° -.00 1.00
”3 .14 .01 .02 -07; -.08‘ .22; 1.00

 

“b"“l signiﬁcant at less than the .10(.05)[.01] level

The variables in table 4 are deﬁned as follows.-

DR
SCRUTINY
LEV

AFR
CAPITAL
PSIZE

PE

1992 discount rate;
Interaction between DR and 1992 total assets;

1992 total long-term debt divided by 1992 total assets;
1992 ABC (adjusted to seven percent) divided by 1993 total pension assets;

1992 cash from operations less capital expenditures and dividends;

AFR divided by 1992 total assets;
Absolute value of 1992 pension expense divided by absolute value of 1992 net
income before pension expense.

 

52

variables: SCRUTINY”, PSIZE, and AFR. The positive correlation with SCRUTINY
(p < .05) is consistent with the suggestion of Francis and Reiter (1987) that larger ﬁrms
may choose higher discount rates to avoid potential political costs. The positive
correlation with PSIZE (p < .01) is consistent with the materiality of the pension plan
affecting discount rate choice. Firms whose pension plans are more material to their
overall ﬁnancial positions tend to choose higher discount rates to obtain the beneﬁts of
decreased liabilities and expenses associated with higher discount rates. The negative
correlation with AFR (p < .01) is consistent with the claims of prior pension research
such as Feldstein and Morck (1983) that ﬁrms whose true funded status is low tend to
choose higher discount rates to give the appearance of a better funded plan.

Given that the dependent variable for testing hypotheses one, two, and three is
the change in the discount rate, the signiﬁcant correlations between the 1992 discount
rate and both SCRUTINY and AFR can be problematic. Without controlling for the
impact that these two independent variables have on the level of the discount rate
before compliance with the SEC’s recommendation, the coefﬁcients on these variables
in a multivariate regression are biased. To eliminate this bias, the 1992 discount rate is
included in the regression to control for the level of the discount rate before the
change. Such an inclusion then allows for the regression coefﬁcients on the
independent variables testing hypotheses one, two, and three to represent the effect on
compliance beyond that amount of change necessary because of the variables’ effects on

the prior level of the discount rate.

 

‘6 Before discussing the individual correlations in Table 4, it should be noted that
SCRUTINY, which is the interaction of total assets and the level of the 1992 discount
rate, correlates with total assets at .99. Thus, although the SCRUTINY variable
attempts to capture more than just a size effect, the practical measurement is no
different than the standard measure of ﬁrm size. An alternative measurement for
SCRUTINY is a dummy variable taking the value of one (zero) if a ﬁrm is (not) in the
top quartile of both assets and discount rates. Regression results using this alternative
measure are similar to those reported in Table 5.

53

Table 5 reports the results of the multivariate regression estimated for
hypotheses one, two, and three. Although the ﬁnal sample is comprised of 600 ﬁrms,
21 ﬁrms are lost due to missing data for the independent variables, bringing the total
number of ﬁrms for the regression estimation to 579. Because the results of estimating
the regression equation speciﬁed in chapter four are qualitatively similar with and
without industry dummy variables, Table 5 reports results without those variables. The
adjusted R2 for the regression is .3717.

Most notable in the table is the large t-statistic on the coefﬁcient for DR, the
discount rate chosen in 1992. While DR is included in the regression mame to control
for the level of the discount rate prior to the SEC's recommendation, its coefﬁcient
provides further evidence on the impact of the SEC's recommendation on accounting
choice. The negative coefﬁcient indicates that ﬁrms well above the acceptable range of
discount rates had large decreases in their rates while ﬁrms closer to the range had
smaller decreases. It appears that, in general, ﬁrms did not decrease their discount
rates by some constant amount, say one percent, to satisfy the SEC. Rather, they
decreased their discount rates by the amount necessary to get within the SEC's desired
range. In other words, the SEC appears to have achieved its desired result.

Compliance hypotheses one, two, and three are tested by the coefficients on
SCRUTINY, LEV, and AFR, respectively. Hypothesis one predicts that compliance
will be positively associated with the potential for being a target of SEC investigation
because those ﬁrms with a high potential can avoid investigation by complying. Since
compliance is measured as the difference between the discount rate in 1992 and 1993, a

negative coefﬁcient on SCRUTINY is expected. The coefﬁcient has the predicted sign

 

‘7 White’s (1980) test determines whether any heteroskedasticity present in the
regressors results in an abnormal variance-covariance matrix of the OLS estimator
[Kennedy (1993)]. The null of no heteroskedasticity is not rejected at the .10 level.
Thus, heteroskedasticity does not appear to affect the regression results. In addition,
regression results are robust to inﬂuencial data diagnostics.

 

54

and is signiﬁcant at p= .023 for a one-sided test. This supports hypothesis one. Firms
that faced a higher potential for being a target of SEC scrutiny complied with the
SEC's recommendation to a greater extent than other ﬁrms in order to avoid the costs
of SEC scrutiny.

Hypothesis two predicts that compliance will be negatively associated with a
ﬁrm’s leverage since compliance increases disclosed liabilities and decreases net
income through an increase in pension expense. Thus, a positive coefﬁcient on LEV is
expected. The coefﬁcient has the predicted sign and is marginally signiﬁcant at
p= .072 for a one-sided test. Thus, hypothesis two is weakly supported. Firms with
higher leverage tended to comply less than ﬁrms with lower leverage in order to avoid
reduced earnings and the costs associated with increased leverage and the tightening of
debt covenants.

Hypothesis three predicts that compliance will be positively associated with a
ﬁrm’s adjusted funding ratio because a ﬁrm with a low adjusted funding ratio will want
to hide the true quality of its pension plan while a ﬁrm with a high ratio will want to
comply to reveal the true quality of its plan. Thus, a negative coefﬁcient on AFR is
expected. The coefﬁcient has the wrong sign and is signiﬁcant at p= .050 for a two-
sided test. The positive coefﬁcient implies that those ﬁrms with lower adjusted funding
ratios complied to a greater degree than those ﬁrms with higher adjusted funding ratios.
A potential explanation for this result is developed in section 5.2.4 below.

Although no hypothesis is stated concerning the impact of the materiality of the
pension plan relative to overall ﬁrm ﬁnancial position, a positive coefﬁcient is expected
on PSIZE. The coefﬁcient is as expected and is signiﬁcant at p= .024 for a one-sided
test, indicating that those ﬁrms whose pension plans were more (less) material to their
overall ﬁnancial position complied to a lesser (greater) extent with the SEC’s
recommendation. The coefﬁcients on the other control variables, CAPITAL and PE,

are insigniﬁcant.

55

Table 5
Regression results for hypotheses one, two, and three

ADR, = a + BIDR, + BZSCRUTINY, + B3LEV, + B4AFR, +
BSCAPITAL, + 06PSIZE, + [17PE,. + 2,.

Adjusted R2 = .37

 

 

 

 

 

 

 

 

 

 

n = 579
Independent Coefﬁcients Standard T-statistics P-values‘
Variables errors

Intercept 5.23 0.38 13.87 .001
DR -0.75 0.04 -17.25 .001
SCRUTINY —0.00 0.00 -2.00 .023
LEV 0.10 0.08 1.47 .072
AFR 0.07 0.04 1.97 .050
CAPITAL 0.00 0.00 0.72 .474
PSIZE 0.19 0.10 1.99 .024
PE 002 0.02 -0.82 .413

 

 

 

 

 

" P-values are one-sided for all variables except AFR, CAPITAL and PB

The variables in table 5 are deﬁned as follows:

ADR Change in the discount rate from 1992 to 1993;

DR 1992 discount rate;

SCRUTINY Interaction between DR and 1992 total assets;

LEV 1992 total long-term debt divided by 1992 total assets;

AFR 1993 total pension assets divided by 1992 adjusted ABO;

CAPITAL 1992 cash from operations less capital expenditures and dividends;
PSIZE AFR divided by 1992 total assets;

PE Absolute value of 1992 pension expense divided by absolute value of

1992 net income before pension expense.

 

56

In summary, regression results support hypotheses one and two but not
hypothesis three. In addition, the coefﬁcient on DR is consistent with the decrease in
discount rates being driven by the SEC’s recommendation. Other tests and an
alternative explanation for the results for hypothesis three follow below.

5.2.1 Altematilemeasutefeunmnlianse

As discussed in chapter four, the measure of compliance for the above
regression is the change in the discount rate from 1992 to 1993. Measuring compliance
in this manner requires a control for the level of the discount rate before the change,
which is accomplished by including the level of the discount rate as an independent
control variable in the regression. Instead of measuring compliance in a manner that
requires such an independent control variable, compliance can be measured in a
percentage format. If movement of the discount rate to 7.0 percent is considered full
compliance, the actual change in the discount rate divided by the change necessary for
full compliance yields a percentage of compliance that ranges from zero to one. For
example, a ﬁrm moving its discount rate from 9.0 to 8.0 percent would comply at a
rate of 50 percent [(9.0-8.0)/(9.0-7.0)]. Using this alternative measure for compliance
as the dependent variable in the regression testing hypotheses one through three then
eliminates the need for the inclusion of the 1992 discount rate as a control variable.

Table 6 reports the results of the regression when this alternative measure of
compliance is employed as the dependent variable and the 1992 discount rate is
removed from the independent variables. Most notable is the large decrease in the
adjusted R2, down from .37 in the original regression to .01 in the current regression.
This decrease appears to result solely from the alternative speciﬁcation of compliance
and the removal of the 1992 discount rate from the independent variables, for the
coefﬁcients on the independent variables are very similar to those in the original
regression. Thus, conclusions with respect to hypotheses one, two, and three are

similar if compliance is measured as the change in the discount rate with the 1992

57

Table 6

Regression results with alternative measure of compliance

COMPLIANCE, = a + BISCRUTINY, + BZLEV, + mm +

B4CAPITAL, + B5PSIZE, + BGPE, + 8,.

Adjusted R2 = .01

 

 

 

 

 

 

 

 

 

n = 579
Independent Coefﬁcients Standard T-statistics P-values‘
Variables errors

Intercept —0.79 0.04 —17.97 .001
SCRUTINY -0.00 0.00 -1 .82 .034
LEV 0.09 0.06 1.49 .069
AFR 0.06 0.03 2.24 .026
CAPITAL 0.00 0.00 0.60 .551
PSIZE 0.16 0.07 2.15 .016
PE 001 0.01 -0.76 .447

 

 

 

 

 

 

' P-values are one-sided for all variables except AFR, CAPITAL and PB

The variables in table 6 are deﬁned as follows:
COMPLIANCE Change in the discount rate from 1992 to 1993 divided by the change

SCRUTINY
LEV

AFR
CAPITAL
PSIZE

PE

necessary to move to a discount rate of seven percent;

Interaction between 1992 discount rate and 1992 total assets;

1992 total long-term debt divided by 1992 total assets;

1993 total pension assets divided by 1992 adjusted ABC;

1992 cash from operations less capital expenditures and dividends;
AFR divided by 1992 total assets;

Absolute value of 1992 pension expense divided by absolute value of
1992 net income before pension expense.

58

discount rate as an independent control variable or as a percentage of compliance
ranging from zero (no compliance) to one (full compliance).

5.2.2 Robustnesssesmtonhxncthesimne

Regression results are consistent with hypothesis one. Firms facing a higher
potential for SEC scrutiny, as measured by ﬁrm size, complied to a greater degree than
ﬁrms facing a lower potential for scrutiny. However, that support may be contingent
on the measurement of the dependent variable. To test the robustness of the results to
an alternative measure of ﬁrm size, total sales is substituted for total assets as a
measure of ﬁrm size, and the regression is estimated again. The results are very
similar for both measures of ﬁrm size. Therefore, support for hypothesis one does not
appear to be contingent on variable measurement.

5.2.3 EunheLanalxsisofhxpothesist

Although Duke and Hunt (1990) conclude that the deﬁnition of the leverage is
not a critical issue in a research design, another potential reason for the weak result is
the speciﬁcation of the variable for leverage. As discussed in chapter four, leverage is
measured as long-term debt divided by total assets to be consistent with Francis and
Reiter (1987). Measuring leverage with long-term debt in the numerator seems
appropriate as pension liabilities are long-term in nature; however, the use of assets in
the denominator is somewhat arbitrary. Other measures of leverage based on long-
term debt used in previous studies include, but are not limited to, long-term debt
divided by the market value of equity and long-term debt divided by the book value of
equity. To ascertain the sensitivity of the regression results to the speciﬁcation of
leverage, the regression is estimated with each of these two additional speciﬁcations.
The coefﬁcient on leverage for each of the alternative speciﬁcations is weaker than the
original speciﬁcation of long-term debt divided by total assets. Thus, the speciﬁcation

of leverage is not a reason for the weak leverage result in Table 5.

59

5.2.4 Alternatmnlanatmfcnhmthesiuhtee
Hypothesis three predicts that compliance with the SEC’s recommendation will

be positively associated with a ﬁrm’s adjusted funding ratio because a ﬁrm with a low
adjusted funding ratio will want to hide the true quality of its pension plan through non-
compliance while a ﬁrm with a high ratio will want to reveal the true quality of its plan
through compliance. Regression results are opposite this prediction, indicating that
ﬁrms with lower adjusted funding ratios complied to a greater degree than ﬁrms with
higher adjusted funding ratios. Assuming that the adjusted funding ratio proxies for
political sensitivity, these results are consistent with the theory of political costs.

Firms with greater sensitivity, as measured by a low adjusted funding ratio, complied
with the SEC’s recommendation to a greater degree to reduce the political costs
associated with the SEC scrutiny.

Traditionally, researchers testing the theory of political costs have used ﬁrm
size to proxy for a ﬁrm’s political sensitivity under the assumption that larger ﬁrms are
more politically sensitive and have larger wealth transfers imposed on them [Watts and
Zimmerman (1986)]. However, this practice has been met with criticism. Researchers
have argued that it is difﬁcult to make reliable links between the theory of political
costs and tests employing ﬁrm size due to the lack of explicit links to events in the
political arena [Wong (1988)] and to the poor construct validity of ﬁrm size as a proxy
for political sensitivity [Ball and Foster (1982), Bernard (1989)], among other
reasons“. The current study addresses and abates these two criticisms. First, the
SEC’s recommendation is clearly an event in the political arena from which inferences
about political costs can be drawn. Second, the assumption that sensitivity to the
SEC’s recommendation varies with the adjusted funding ratio (i.e. , that the adjusted

funding ratio has high construct validity) is supported in two ways.

 

13 See Key (1995) for a review of the issues relating to ﬁrm size as a proxy for political
costs.

 

60

First, parties that impose political pressure on ﬁrms have generally focused on
the funded status of pension plans or how pension accounting choices affect a plan’s
funded status. For example, since 1990 the Pension Beneﬁt Guaranty Corporation
(PBGC) has published annually a list of the ﬁfty most unfunded plans in the United
States. A ﬁrm makes the list based solely on its funded status as computed by the
PBGC. In addition, the United States Congress attached pension legislation to the 1994
General Agreement on Tariffs and Trade legislation to increase the degree and speed of
pension plan funding. This attached legislation was in part a response to concerns from
Labor Secretary Robert Reich and PBGC Executive Director Martin Slate that pension
under-funding was a chronic problem (IbeﬂalLStteeleimal, March 15, 1995).
Finally, for the last decade Wind and other ﬁnancial news outlets
have periodically printed articles addressing various aspects of pension funding.

Second, and perhaps more important, the SEC’s action came at a time when
mean ﬁrm discount rates were higher than the rates of return on high-quality ﬁxed
income investments. Figure 2 of Blankley and Swanson (1995) graphically illustrates
the movement of mean discount rates and rates of return on high—quality ﬁxed income
investments from 1987 to 1993. Figure 2 in this study partially replicates Blankley and
Swanson’s (1995) ﬁgure but replaces their mean discount rate data with data from this
study. As seen in the Figure 2, rates of return on high-quality ﬁxed investments were
higher than mean discount rates from 1987 to 1990. However, not until rates of return
on high-quality ﬁxed investments dropped below mean discount rates did the SEC take
any action. This suggests that the SEC’s main concern was not that discount rates
should mirror rates of return on high-quality investments, per se, but that funding
ratios should not be inﬂated by choosing discount rates that are higher than rates of
return on high-quality ﬁxed investments. In other words, discount rates should have a

ceiling so that funding ratios are not grossly inﬂated.

61

 

 

10.5

9.5

Interest rates

7.5

7.0

 

 

10.0 . _

9.0 __

8.5 _

8.0 _

 

 

 

+ Discmmt rates
+ SEC benchmark

 

 

 

 

 

1987

1990 1991 1992 1993

1988 1989

Ye er

 

 

 

 

 

Discount rates
SEC benchmark‘

1281 1288 1282 12%) 1221 1222 1223
8.64 8.82 8.64 8.69 8.54 8.41 7.45
10.13 9.99 9.21 9.39 8.34 8.05 7.22

'SEC benchmark is deﬁned as the rate of return on high-quality ﬁxed income

investments.

Figure 2
Comparison of discount rates and SEC benchmark

62

Given that the SEC’s recommendation appears to have originated from a
concern over inﬂated funding ratios, the assumption that the adjusted funding ratio is a
better proxy for sensitivity to the SEC scrutiny is supported. Thus, the testing of
hypothesis three constitutes a good test of the theory of political costs. Results are
consistent with the theory. Firms with lower adjusted funding ratios complied with the
recommendation to a greater degree to avoid any costs that the SEC might impose on
them.

5.2.5 Riskeffect

Firm risk has been suggested in prior research as a factor that affects pension
accounting choices such as actuarial assumptions and funding strategies [e.g. , Feldstein
and Morck (1983), Bodie et al. (1985), and Francis and Reiter (1987)]. The traditional
argument, especially with respect to funding strategies, has centered around the
function of the Pension Beneﬁt Guaranty Corporation (PBGC). The PBGC is a federal
agency that guarantees payment of basic retirement beneﬁts earned by American
workers participating in private-sector deﬁned beneﬁt pension plans. If a ﬁrm is
unable to fulﬁll its promised pension obligation, the ﬁrm can legally satisfy that
obligation by transferring all pension assets and 30% of the market value of the ﬁrm’s
net worth to the PBGC. Therefore, authors have argued that risky ﬁrms should fund
their plans less to increase the value of the pension “put” option. Tests of this
implication have produced mixed results with Bodie et al. (1985) ﬁnding evidence
consistent with the risk argument and Francis and Reiter (1987) ﬁnding evidence
inconsistent with the risk argument.

Because the PBGC insurance is based on statutory funding and the discount rate
under FAS 87 has no direct effect on statutory funding, it is unlikely that this
traditional risk argument centering around the PBGC insurance should affect the
discount rate. However, if the pension plan is viewed as a form of corporate debt with

the beneﬁciaries as the providers of capital, it is reasonable to believe that some

63

component of the discount rate is a reﬂection of the riskiness of the ﬁrm. Feldstein
and Morck (1983) interpret their regression results with a similar argument, suggesting
that ﬁrms may choose interest rates under APB 8 to reduce the pension liability in
order to reduce the overall assessment of ﬁrm risk. Therefore, it is necessary to
examine the impact of ﬁrm risk on compliance.

Firm risk is measured as the Standard and Poor’s debt rating disclosed on
Compustat”. Similar measures have been used in Feldstein and Morck (1983), Bodie
et al. (1985), and Francis and Reiter (1987). If some component of the discount rate is
a reﬂection of ﬁrm risk, then the discount rate and the debt rating should be correlated.
Table 7 presents correlations between discount rates and debt ratings for the 288 (301)
ﬁrms with available data in 1992 (1993). As seen in the table, 1992 discount rates
correlate positively and signiﬁcantly (p= .045) with 1992 debt ratings, implying that
ﬁrms with higher risk tend to choose higher discount rates. However, 1993 discount
rates do not correlate signiﬁcantly (p= .738) with 1993 debt ratings. These two
correlations taken together, coupled with the fact that only 57 of the 278 ﬁrms with
rating data available for both years had a change in their debt ratings, suggest that as
ﬁrms moved to the SEC’s tight range of acceptable discount rates, ﬁrms may have
removed any risk component within their discount rates. This implies that the decrease
in the discount rate in 1993 was affected by ﬁrm risk because ﬁrm risk affected the
1992 discount rate but not the 1993 discount rate. To the extent that this reasoning is
true, however, the inclusion of the 1992 discount rate in the regression estimated to test
hypotheses one, two, and three should control for any impact that ﬁrm risk has on the
change in the discount rate. As a result, it appears as if any conclusions drawn

previously should not be altered because of the impact of ﬁrm risk.

 

‘9 The Standard and Poor’s debt rating works as follows: A ﬁrm with a rating of Aaa
(i.e., low risk) would be coded as a 2, a ﬁrm with a rating of Aa would be coded as a
3, and so on.

64

Table 7
Correlations between discount rates and debt ratings

 

 

DIS92 D1893 RATING92 RATING93
DIS92 l .00
DIS93 .24c 1 .00
RATING92 . 12" -.02 1 .00
RATIN G93 .08 .02 .98“ 1 .00

 

“(b)” signiﬁcant at less than the .10(.05)[.01] level

The variables in table 7 are deﬁned as follows.-

DIS92 1992 discount rate;

DIS93 1993 discount rate;

RATING92 Standard and Poor’s debt rating for 1992;
RATING93 Standard and Poor’s debt rating for 1993.

 

65

5.2.6 Analxsmfﬁnnslesﬂnsamnleseleetion

As mentioned in chapter four, ﬁrms with 1992 discount rates below eight
percent are eliminated from the sample in order to leave only those ﬁrms that have a
decision to comply or not with the SEC’s recommendation. This results in a loss of
151 ﬁrms. An issue that arises as a result of this elimination is whether the eliminated
ﬁrms differ systematically from the ﬁrms that remain in the sample and whether this
difference impacts the testing of hypotheses one, two and three. For example, one
could argue that those ﬁrms that would face a high potential for SEC scrutiny given an
SEC recommendation could have already reduced their discount rates in a prior year to
avoid such future scrutiny. If this is the case, then the ability to detect a relation
between compliance and the potential for scrutiny is limited, for those ﬁrms with a
high potential for scrutiny have already complied. Table 8 presents a comparison of
ﬁrms grouped by 1992 discount rates.

The ﬁrst three variables in Table 8 indicate that ﬁrms with discount rates equal
to or above eight percent are similar to ﬁrms with discount rates below eight percent
with respect to the size of the accumulated beneﬁt obligation, the assets contributed to
the plan, and the funding ratio. However, the two groups are not similar with respect
to the adjusted funding ratio. Firms choosing higher discount rates have signiﬁcantly
smaller adjusted funding ratios (p < .01). This result is consistent with the signiﬁcantly
negative correlation (p < .01) between independent regression variables AFR and DR
found in Table 4 and further corroborates the claims of Feldstein and Morck (1983)
that ﬁrms manipulate their funded status through discount rate choice.

Although the absolute sizes of the average pension liability and the average plan
assets are similar across groups, the pension plans of ﬁrms choosing higher discount
rates are signiﬁcantly more material to their overall ﬁnancial positions. Pension
expense divided by net income before pension expense and the adjusted liability divided
by total assets are both signiﬁcantly greater (p < .01 and p < .10, respectively) for ﬁrms

 

66

Table 8

Descriptive statistics for ﬁrm groups based on 1992 discount rate

mean (median) [std dev]

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Variable Firms 2 8.0% Firms < 8.0%
n=600 n = 151
499 618
ABC (53) (39)
[1859] [1713]
643 771
Plan assets (69) (58)
[2537] [2123]
1.30 1.34
Reported funding ratio (1.25) (1.30)
[.439] [.179]
1.09“ 1.29
AFR (1.05) (1.25)
[.378] [.389]
.24“ .13
PE (.08) (.07)
[.812] [.179]
. 16" .13
Psize (.11) (.10)
[.170] [.146]
3596 3370
Total assets (723) (703)
[11228] [62551
30476 24588
Scrutiny (5940) (5170)
[97685] [457031
.28“ .25
leverage (.28) (.26)
[.207] [.200]
-41 -25
Capital (-1) (-1)
[403] [186]
5.55“ 5.40
Salary rate (5.50) (5.25)
[.802] [.865]
9.19“ 8.50
Rate of return (9.00) (8.50)
[.891] [.927]
atb)l<:I

signiﬁcantly different across groups at less than the .10(.05)[.01] level

 

67

with higher discount rates. This result supports the inclusion of control variables in the
regression testing hypotheses one through three.

Firms with discount rates above eight percent also have higher leverage
(p < .10) than ﬁrms with lower discount rates. This difference is consistent with
hypothesis two, which predicts that ﬁrms with high leverage will be hesitant to comply
with the SEC’s recommendation. It appears that ﬁrms with lower leverage have
already reduced their discount rates, thus leaving a subset of ﬁrms for the ﬁnal sample
that face similar incentives with respect to leverage. The resulting sample of relatively
highly leveraged ﬁrms may therefore inhibit the ability to detect a relation between
leverage and compliance and could explain the weak support for hypothesis two.

5.3 Results for tests of smoothing

The ﬁnal empirical question arising from the SEC's recommendation that was
developed into hypothesis four is whether ﬁrms that complied with the recommendation
smoothed the negative effect by decreasing their salary rates. As discussed in section
5.1, descriptive statistics are consistent with ﬁrms using the salary rate in this manner.
Table 2 reports that the mean salary rate decreased from 5.55% in 1992 to 4.93% in
1993. In addition, Table 3 illustrates that this decrease in the salary rate was much
more pronounced than in prior years, both in the magnitude of the decrease and in the
percentage of ﬁrms making a change. The following paragraphs discuss the results of
the multivariate regression estimated to answer hypothesis four.

Table 9 presents the correlation matrix of independent variables included in the
multivariate regression speciﬁed in chapter four. As seen in the table, multicollinearity
does not appear to be an issue with respect to the majority of variables as most
correlations are insigniﬁcantly different from zero. However, there are highly
signiﬁcant correlations between total assets and both the variance of earnings (r= .56,
p< .001) and the reduction in employees (r=-.42, p< .001). In addition, the

individual t—statistics for both variables are less than .21 when the regression is

 

68

estimated, and the overall results of the regression are unchanged when the variables
are excluded. Thus, given that total assets may proxy for both variables, where larger
ﬁrms have a higher variance in earnings and larger reductions in employees, the
regression testing hypothesis four is estimated and reported without the variance of
earnings and the reduction in employees as independent variables.

Table 10 reports the results of the multivariate regression estimated for
hypothesis four. Predicted signs and deﬁnitions of all variables included in the
regression are summarized in Figure 3. There are two items to note before discussing
the results of the regression. First, Table 10 reports regression results excluding
industry dummy variables. Results including such dummies are similar to those
reported in the table. Second, only 373 of the possible 600 ﬁrms are used in the
regression estimation because institutional ownership data from Moody’s is available
for only 37820 ﬁrms. To the extent that the ﬁrms included in Moody’s differ
systematically from ﬁrms not included in Moody’s with respect to other ﬁnancial
characteristics such as discount rates, pension plans and ﬁrm size, the remaining
sample may inhibit correct inferences about hypothesis four. A comparison of values
for each of the regression variables for ﬁrms with and without institutional ownership
data reveals no signiﬁcant differences across any variable except for ﬁrm size, which is
expected given that Moody’s tends to follow larger ﬁrms. Thus, the reduction in the
sample size does not appear to be problematic for test of hypothesis four.

The results reported in Table 10 are consistent with hypothesis four”. The
coefﬁcient on the change in the discount rate is positive and highly signiﬁcant,

indicating that ﬁrms that decreased their discount rates in compliance with the SEC’s

 

20 Only 373 of the 378 ﬁrms with institutional ownership data are used in the
regression estimation due to missing data for other regression variables.

21 The null of no heteroskedasticity for White’s (1980) test is not rejected at the .10
level.

69

Table 9

Correlation matrix

 

 

ADR
SIZE
PE
UEARN
VAR
DOWN
SIZE
INS

OWN
SR

ADR

1.00
-.10"
-.02
-.O3
-.08’I
.03
.02

.02

SIZE

1.00

-.01

.03

.56c

-.42c

.06

.02

PE

1.00

-.01

.00

.00

.00

UEARN VAR

1.00

.01

.06

.14“

-.05

1.00

-.48c

.06

-.OO

DOWN
SIZE

1.00

-.03

-.02

INS SR
OWN

1.00

.08 1.00

 

"MM signiﬁcant at less than the .10(.05)[.01] level

The variables in table 9 are deﬁned as follows:
Change in 1993 discount rate;

Total assets;
Absolute value of pension expense divided by absolute value of net income

ADR
SIZE
PE
UEARN

VAR

DOWNSIZE

INSOWN
SR

before pension expense;

Dummy variable taking the value of one if the 1994 earnings surprise is

negative, zero otherwise;

Standard deviation of the past ﬁve years of earnings;

Change in number of employees;

Common shares held by institutions divided by total common shares;

1992 salary rate.

 

7O

 

 

 

 

 

 

 

 

 

 

 

Variable Deﬁnition Predicted sign
ADR Change in the discount rate +
Common shares held by institutional investors
INSOWN divided by total common shares outstanding +
Dummy variable taking the value of one if the
UEARN 1994 earnings surprise is negative, zero +
otherwise
SIZE Total assets -
Absolute value of pension expense divided by
PE the absolute value of net income before -
pension expense
SR 1992 salary rate -
Figure 3

Predicted signs for hypothesis four regression coefﬁcients

 

 

71
Table 10

Regression results for hypothesis four

ASR, = a, + more. + BZINSOWN, + B3UEARN, +
5431213,. + 13,1313, + 9,81; + e,-

 

 

 

 

 

 

 

 

 

Adjusted R2 = .37 e
n=373
Independent Coefﬁcients Standard T-statistics P-values‘

Variables errors
Intercept 2.08 0.22 9.65 .001
ADR 0.46 0.06 7.80 .001
INSOWN 0.15 0.12 1.29 .098
UEARN 0.09 0.06 1.55 .061
SIZE -0.00 0.00 -1.24 .109
PB -0.04 0.02 -1.85 .033
SR -0.43 0.04 -12.02 .001

 

 

 

 

 

 

 

‘ P-values are one-sided for all variables

The variables in table 10 are deﬁned as follows:

ASR
ADR
INSOWN
UEARN

SIZE
PE

SR
i

Change in the salary rate;

Change in the discount rate;

Common shares held by institutional investors divided by total shares;
Dummy variable taking the value of one if the 1994 earnings surprise is
negative, zero otherwise;

Total assets;

Absolute value of pension expense divided by the absolute value of net
income before pension expense;

1992 salary rate;

Firm index.

72

recommendation decreased their salary rates to offset the negative effect from
compliance. All control variables also have their predicted signs and are signiﬁcant at
p= . 109 or lower for one-sided tests. In other words, fnms with smaller institutional
ownership (INSOWN), negative future transitory earnings (UEARN), larger total
assets (SIZE), more material pension plans relative to overall ﬁnancial position (FEW,
and higher 1992 salary rates (SR) were more likely to smooth the effect of compliance
with the SEC’s recommendation.

The dependent variable in the above regression, ASR, has a large number of
observations with the value of zero. Of the 373 sample ﬁrms used in the estimation,
133 did not change their salary rate during 1993. Ordinary Least Squares (OLS)
regression estimates can be biased if the dependent variable is limited in some way
[Kennedy (1993)]. To ascertain the sensitivity of the above OLS results to this
potential bias, the regression is also estimated using Tobit. The Tobit procedure is
normally used for data with either censored or truncated dependent variables that arise
from the unobservability of the dependent variable above or below some threshold or
the unobservability of the independent variable above or below some threshold,
respectively [Kennedy (1993), Maddala ( 1991)]. The procedure is therefore not always
appropriate for data sets whose dependent variable contains a large number of observed
values, in this case zero, at one extreme of its range of values [Maddala (1991)].
However, estimation of the above regression using Tobit provides a reasonable check
against the possibility that the OLS regression results are biased and misleading.

Table 11 contains the Tobit results. As seen in the table, the results are similar
to the OLS results in Table 10. The coefﬁcient on the change in the discount rate

continues to be positive and highly signiﬁcant, and the coefﬁcients on the control

 

22 Of the 373 ﬁrms employed in the regression estimation, 73 reported negative pension
expense. Estimating the regression excluding these ﬁrms produces qualitatively similar
results.

 

73
Table 11

Tobit regression results

ASR, = a, + BIADR, + BZINSOWN, + B3UEARN, +
[3,31213, + BSPE, + 9651:. + e,

 

 

 

 

 

 

 

 

 

Independent Coefﬁcients Standard T-statistics P-values‘
Variables errors
Intercept 3 .69 0.36 10.10 .001
ADR 0.55 0.09 5.77 .001
INSOWN 0.08 0.18 0.47 .321
UEARN 0.20 0.09 2.19 .014
SIZE -0.00 0.00 -2.59 .005
PE -0.05 0.03 -1.56 .059
SR -0.62 0.06 -10.34 .001

 

 

 

 

 

‘ P-values are one-sided for all variables

The variables in table 11 are deﬁned as follows:

ASR Change in the salary rate;

ADR Change in the discount rate;

INSOWN Common shares held by institutional investors divided by total shares;

UEARN Dummy variable taking the value of one if the 1994 earnings surprise is
negative, zero otherwise;

SIZE Total assets;

PE Absolute value of pension expense divided by the absolute value of net
income before pension expense;

SR 1992 salary rate;

i Firm index.

 

 

74

variables continue to be directionally consistent with predictions. Thus, it appears that
conclusions drawn from the OLS regression are appropriate. Firms that reduced their
discount rates in compliance with the SEC’s recommendation smoothed the effect of
compliance by reducing the salary rate.

5.3.1 ResultsﬁemadditienaLsmoethinuests

As discussed in chapter four, an alternative hypothesis to the claim that the
coefﬁcient on the change in the discount rate in the above regression is evidence of
smoothing behavior is that the coefﬁcient simply represents the positive correlation
inherent in the two rates due to their common component. Data for the analysis of this
alternative hypothesis is restricted to a randomly selected 277 ﬁrms from the total
sample of 600 ﬁrms. The process of random selection is intended to avoid the
existence of biases in the subset that might result in erroneous inferences about the
whole sample. Table 12 presents a comparison of the 277 firms in the subset with the
remaining 323 ﬁrms omitted from the analysis to conﬁrm that the randomization
worked. As seen in the table, the subset of ﬁrms is not signiﬁcantly different from the
remaining ﬁrms with respect to any ﬁnancial variables of interest. Thus, the subset
appears to be representative of the whole sample and should allow for valid inferences
about the whole sample.

The alternative hypothesis is ﬁrst addressed by examining the yearly
correlations between changes in the discount rate and salary rate from 1988 to 1992.
Table 13 presents these correlations. Each correlation in the table is based only on data
from those ﬁrms with a non-zero change in either the discount rate or the salary rate
for the given year. This restriction is imposed because inference is sought concerning
the relation between changes in the two rates when some change is made in at least one
rate. As seen in the table, correlations for three of the ﬁve years are not signiﬁcantly
different from zero at the p < .10 level. Further, the correlation in 1991 has a negative

sign. As a whole, analysis of the history of these correlations does not generate the

75

Table 12

Comparison of subsample and non-subsample ﬁrms

mean (median) [std dev]

 

 

 

 

 

 

 

 

 

 

 

 

 

Variable Firms in subsample Firms not in subsample
n=277 n=323
546 593
ABC (71) (61)
[2013] [2145]
655 724
Plan assets (76) (72)
[2619] [281 1]
1.16 1.20
Funded status (1.13) (1.16)
[.358] [.362]
.61 .37
PE (.07) (.08)
[5.21] [1.22]
4313 3351
Total assets (706) (745)
[17633] [7145]
7.45 7.45
Discount rate (7.50) (7.50)
[.425] [.377]
-.96 -.93
Change in discount rate (-1.00) (-1.00)
[.485] [.465]
4.92 4.93
Salary rate (5.00) (5.00)
[.836] [.776]
-.63 -.63
Change in salary rate (-50) (-.50)
[.692] [.643]

 

"hm significantly different across groups at less than the .10(.05)[.01] level

 

 

76

Table 13

Yearly correlations between changes in discount rates and salary rates

 

 

 

1988-1992
Year Correlation P-value N
1988 0. 12 .257 91
1989 0. 18 .073 99
1990 0. 15 . 132 97
1991 -0.04 .650 122
1992 0.24 .013 107

 

 

 

 

 

 

77

expectation of a highly signiﬁcant coefﬁcient on the change in the discount rate in the
original smoothing regression.

The alternative hypothesis is further addressed by estimating the second
regression speciﬁed in section 4.3.4 of chapter four. Table 14 presents the results of
this estimation. The results ﬁrst provide evidence on whether the relation between the
change in the discount rate and the change in the salary rate was stronger in 1993 than
in prior years. The coefﬁcient on the change in the 1993 discount rate (ADR*D,993) is
signiﬁcantly greater than zero at p < .0001 , indicating that the relation was indeed more
positive in 1993 than in prior years. This result rules out the alternative hypothesis
that the coefﬁcient on the change in the discount rate in the original smoothing
regression was due solely to the normal relation between the two rates, thus allowing
for the conclusion that the salary rate was reduced to smooth the negative effect of
compliance with the SEC’s recommendation.

The results in Table 14 also provide evidence on the reasonableness of the
expectation of the change in the salary rate developed in chapter four for purposes of
measuring the dependent variable for the original smoothing regression. The
coefﬁcient on the change in the Consumers' Price Index (ACPI) is positive and
signiﬁcant, indicating that changes in the salary rate from 1987 to 1992 were positively
related to changes in the CPI over the same period. Because the change in the CPI was
zero for 1993, the positive relation in prior years suggests that the change in the salary
rate in 1993 should tend towards zero as well. Thus, the data is consistent with the
expectation that the change in the salary rate for 1993 should be zero in the absence of

smoothing behavior.

 

78

Table 14

Regression results for alternative hypothesis for hypothesis four

 

 

 

 

 

 

 

 

ASR“ = 01..., + B,ADR,, + [3le993 + B3ADR,, * D1993 + B4ACPI, + 3,
Adjusted R2 = .22
n=1409
h
Independent Coefﬁcients Standard T-statistics P-values '
Variables errors
Intercept -0.07 0.01 -4.91 .001
ADR 0.19 0.03 5.59 .001
13,99, 007 0.07 -1.08 .281 I "
ADR’IDw93 0.31 0.07 4.41 .001
ACPI 0.06 0.02 3.35 .001

 

 

 

 

 

 

The variables in table 14 are deﬁned as follows:

ASR
ADR
D1993
ACPI
i

t

Change in the salary rate;

Change in the discount rate;

Dummy variable taking the value of one if 1993, zero otherwise;
Change in the Consumers’ Price Index;

Firm index;

Year index, 1988-1993.

79

5.3 Summary

This chapter contains the results of tests of hypotheses one through four.
Hypothesis one predicts that compliance with the SEC’s recommendation is increasing
in the potential for being a target of SEC scrutiny. The data are consistent with this
hypothesis. Hypothesis two predicts that compliance is decreasing in leverage. The
data are also consistent (weakly) with this hypothesis. Hypothesis three predicts that
compliance is increasing in the adjusted funding ratio. The data are inconsistent with
this hypothesis. However, to the extent that the adjusted funding ratio is a proxy for
political sensitivity, the results for hypothesis three are consistent with the political cost
hypothesis. Hypothesis four predicts that those ﬁrms that comply with the
recommendation will smooth the effect by reducing the salary rate. The data are

consistent with this hypothesis. Concluding remarks are contained in chapter six.

Chapter Six
CONCLUSIONS

This study examines ﬁrm response to a September, 1993 letter written by the
Chief Accountant of the Securities and Exchange Commission (SEC) recommending
that ﬁrms reduce their assumed pension discount rates for 1993. The study provides
answers to the following empirical questions that arise as a result of the
recommendation. First, what are the implications of an SEC recommendation? Does a
recommendation from the SEC have enough inﬂuence to alter an accounting choice that
negatively impacts reported expenses and liabilities. Second, why is there variation, if
any, in the degree of compliance across ﬁrms? Did factors suggested by existing
theory and prior results from pension research inﬂuence the degree of compliance?
Third, given that ﬁrms complied with the recommendation to some degree, did some
ﬁrms alter other accounting choices to mitigate the effect of compliance?

Section 6.1 summarizes the research ﬁndings to the above questions. Section
6.2 lists the contributions of the study to existing accounting research. Suggestions for
future research are contained in section 6.3.
6.1 Summary of research ﬁndings

As stated above, one of the ﬁrst empirical questions arising from the SEC's
recommendation is whether such a recommendation contained enough inﬂuence to alter
accounting choice. The data in the study are consistent with the SEC’s
recommendation having a substantial impact on accounting choice. The decrease in the
discount rate for 1993 for the average ﬁrm is approximately one percent. This
decrease translates into a signiﬁcant decrease in the reported funded status for the
average ﬁrm. In addition, the decrease in the discount rate is much more pronounced
than in prior years, both in the magnitude of the change and in the percentage of ﬁrms

making a change. This result illustrates that the change in the discount rate for 1993
80

 

81

was well outside the historical trend, suggesting the introduction of an outside inﬂuence
in 1993, namely the SEC recommendation, which signiﬁcantly altered accounting
choice.

The second empirical question is whether the extent of compliance with the
recommendation can be explained by factors suggested by existing theory and pension
research. This question is addressed by hypotheses one, two and three. Hypothesis
one predicts that compliance with the SEC's recommendation is positively associated
with the potential for being a target of SEC investigation. The data are consistent with
this hypothesis. Larger ﬁrms appear to have complied with the recommendation to a
greater extent to avoid being a target of SEC investigation. This result is consistent
with the political cost hypothesis of positive accounting theory.

Hypothesis two predicts that compliance with the SEC’s recommendation is
negatively associated with leverage. The data are weakly consistent with this
hypothesis. Firms with higher leverage appear to have complied to a lesser extent to
avoid the reduction to earnings and the subsequent costs of tightening debt covenants.
This result is consistent with the debt—equity hypothesis of positive accounting theory.

Hypothesis three predicts that compliance with the SEC's recommendation is
positively associated with the adjusted funding ratio. The data are inconsistent with
this hypothesis. Firms with low adjusted funding ratios appear to have complied to a
greater extent than ﬁrms with high adjusted funding ratios. This result is interpreted as
consistent with the political cost hypothesis. Given that the adjusted funding ratio
proxies for political sensitivity to the SEC in the pension area, ﬁrms with greater
sensitivity, as measured by a low adjusted funding ratio, appear to have complied with
the SEC’s recommendation to a greater extent to reduce the political costs associated
with the SEC scrutiny.

The ﬁnal empirical question of whether those ﬁrms that complied with the

recommendation altered other accounting choices to mitigate the effect from

82

compliance is addressed by hypothesis four. Hypothesis four predicts that ﬁrms with
incentives to report smooth earnings who lower their discount rate in compliance with
the SEC are more likely to lower their salary progression rate. For 1993, the average
decrease in the salary rate is approximately one—half percent. This decrease is much
more pronounced than in prior years, both in the magnitude of the change and in the
percentage of ﬁrms making a change. This break from the historical trend is consistent
with hypothesis four. Regression results are also consistent with hypothesis four.
Firms appear to have smoothed the effect of compliance by reducing the salary
progression rate. This result is consistent with the smoothing hypothesis.

6.2 Contributions

This study contributes to the accounting research literature in several ways.
First, the study highlights the difference between the discount rate chosen under
Statement of Financial Accounting Standards N o. 87, "Employers' Accounting for
Pensions" (PAS 87) and the funding interest rate chosen under the Employee
Retirement Income Security Act of 1974 (ERISA). An important aspect of this
difference is that a change in the discount rate has no direct cash ﬂow effects. Cash
ﬂows are affected by statutory funding requirements, which are determined by the
funding interest rate, not the discount rate. As a result, the study illustrates that care
must be taken by both academic researchers and the popular press in characterizing the
effect on cash ﬂows from a change in actuarial assumptions under FAS 87.

Second, the study illustrates that the SEC has a substantial impact on accounting
choice. While this result is perhaps not surprising, the means to the end result is
somewhat surprising. Accounting choice studies are usually conducted around events
such as pronouncements from the Financial Accounting Standards Board, Financial
Reporting Releases by the SEC, or some other statement emerging from the ofﬁcial
processes of a regulatory body. This study illustrates that accounting choice can also

be affected by non-ofﬁcial statements such as a letter from the Chief Accountant of the

83

SEC. As a result, the study should impel researchers to investigate accounting choice
in a much wider array of settings, thus providing a richer understanding of how
accounting choices are made.

Third, the study provides further evidence consistent with both the political cost
and debt-equity hypotheses of positive accounting theory as well as the smoothing
hypothesis. As the study employs a unique setting, a recent time period, and an
accounting choice in the pension area, the study helps to establish the robustness of
these three hypotheses across different settings, different time periods, and different
accounting choices.

6.3 Suggestions for future research

One of the objectives of the SEC’s letter was to tie the discount rate to the rate
of return on high quality ﬁxed income investments. An extension of this research is to
examine the movement of discount rates subsequent to 1993 to ascertain how well they
correspond to the SEC’s benchmark rate. Deviations from the benchmark may be
systematically related to ﬁnancial and/or non-ﬁnancial aspects of ﬁrms.

Another direction for future research is to examine the reaction of the market to
the changes in the discount rates brought on by the SEC’s recommendation. Feldstein
and Morck (1983) ﬁnd that pension liabilities are impounded into market values at a
common discount rate. However, this result is found when pensions are governed by
Accounting Principles Board Opinion 8. Does the market continue to impound pension
liabilities at a common rate given the SEC’s recommendation and is that common rate
equal to the SEC’s benchmark?

Finally, searches for other statements made by the SEC outside its normal
policy-making process can be conducted to identify additional ways in which the SEC
affects accounting choice. Gaining a better understanding of the SEC’s impact on
accounting choice may provide a better understanding of the political cost hypothesis
and lead to stronger tests of the hypothesis.

LIST OF REFERENCES

LIST OF REFERENCES

Accounting Principles Board, 1966, Opinion N o. 8: Accounting for the cost of pension
plans.

Ali, A., and KR. Kumar, 1993, Earnings management under pension accounting
standards: SFAS 87 versus APB 8, Journal of Accounting Auditing and
Finance, 427-446.

American Institute of Certiﬁed Public Accountants, 1993, Audit Risk Alert-1993, 11-3.

Ball, R. and G. Foster, 1982, Corporate ﬁnancial reporting: A methodological review
of empirical research, Journal of Accounting Research, 161-234.

Bange, M .M. and W. De Bondt, 1995, R&D-budgets and corporate earnings targets,
working paper.

Barth, M.E. , 1991, Relative measurement errors among alternative pension asset and
liability measures, The Accounting Review, 433-463.

Barth, M.E., J.A. Elliott and M.W. Finn, 1992, Market rewards for increased
earnings patterns, working paper.

Beattie, V., S. Brown, D. Ewers, B. John, S. Manson, D. Thomas and M. Turner,
1994, Extraordinary items and income smoothing: A positive accounting
approach, Journal of Business Finance and Accounting, 791-811.

Bernard, V., 1989, Capital markets research in accounting during the 1908’s: A critical
review, in The State of Accounting Research as We Enter the 1990’s, University
of Illinois, 72-120.

Blankley, AI. and ER Swanson, 1995, A longitudinal study of SFAS 87 pension rate
assumptions, Accounting Horizons, 1-21.

85

Bodie, Z., J. Light, R. Morck and RA. Taggart, 1985, Corporate pension policy: An
empirical investigation, Financial Analysts Journal, 10-16.

Buck Consultants, Inc. , 1995, Economic Pension Actuarial Assumptions Survey -
1993.

Buck Consultants, Inc., 1994, The Health of America's Pension Plans - 1993.

Christie, A.A. , 1990, Aggregation of test statistics: An evaluation of the evidence on
contracting and size hypotheses, Journal of Accounting and Economics, 15-36.

Coopers & Lybrand, 1994, Audit Committee Update 1994, 17.

Duke, J.C. and HG. Hunt 111, 1990, An empirical examination of debt covenant
restrictions and accounting-related debt proxies, Journal of Accounting and
Economics, 45-63.

Dye, R.A. , 1988, Earnings management in an overlapping generations model, Journal
of Accounting Research, 195-235.

Employee Retirement Income Security Act, 1974 (Washington, DC).

Feldstein, M. and R. Morck, 1983, Pension-funding-decisions, interest rate
assumptions, and share prices, In Financial Aspects of the United States Pension
System, (Z. Bodie and J. Shoven, eds.), (Chicago, University of Chicago
Press), 177-207.

Feroz, E.H., K. Park and VS. Pastena, 1991, The ﬁnancial and market effects of the
SEC's accounting and auditing enforcement releases, Journal of Accounting
Research, 107-142.

Financial Accounting Standards Board, 1980, Statement of Financial Accounting
Standard No. 35: Accounting and reporting by deﬁned beneﬁt pension plans
(FASB, Stamford, CT).

Financial Accounting Standards Board, 1980, Statement of Financial Accounting
Standard No. 36: Disclosure of pension information (FASB, Stamford, CT).

Financial Accounting Standards Board, 1985a, Statement of Financial Accounting
Standard No. 87: Employers' accounting for pensions (FASB, Stamford, CT).

86

Financial Accounting Standards Board, 1985b, Statement of Financial Accounting
Standard No. 88: Employers’ accounting for settlements and curtailrnents of
deﬁned beneﬁt pension plans and for termination beneﬁts (FASB, Stamford,
CT).

Fogarty, T.J. and J. Grant, 1995, Impact of the actuarial profession on ﬁnancial
reporting, Accounting Horizons, 23-33.

Francis, LR. and SA. Reiter, 1987, Determinants of corporate pension funding
strategy, Journal of Accounting and Economics, 35-59.

Hand, J. , 1989, Did ﬁrms undertake debt-equity swaps for an accounting paper proﬁt
or true ﬁnancial gain?, The Accounting Review, 587—623.

Harper, RM. and JR. Strawser, 1993, The effect of SFAS 87 on corporate funding of
deﬁned beneﬁt pension plans, Journal of Business Finance and Accounting,
815-833.

Healy, RM. and KC. Palepu, 1990, Effectiveness of accounting-based dividend
covenants, Journal of Accounting and Economics, 97-123.

Hepworth, SR, 1953, Periodic income smoothing, The Accounting Review, 32—39.

Holthausen, R.W. and R.W. Leftwich, 1983, The economic consequences of
accounting choice, Journal of Accounting and Economics, 77-117.

Howe, IS. and G.G. Schlarbaum, 1986, SEC trading suspensions: Empirical evidence,
Journal of Financial and Quantitative Analysis, 323-334.

Jensen, MC. and W.H. Meckling, 1976, Theory of the ﬁrm: Managerial behavior,
agency costs and ownership structure, Journal of Financial Economics, 305-
360.

Jereski, L. “SEC is challenging funding for plans,” The Wall Street Journal,
November 7, 1993.

Karr, AR. and EB. Schultz, “Stealth legislation: Pension rules tacked quietly on trade
bill portend vast changes,” The Wall Street Journal, March 15, 1995.

Kennedy, P., 1993, A guide to econometrics - Third edition (The MIT Press,
Cambridge, MA).

Key, K., 1995, Political cost incentives for earnings management in the cable television
industry, working paper.

87

Kile, C. , 1993, An analysis of the impact of omitted evaluative information on
ﬁnancial reporting: Evidence from disclosures to amend management's
discussion and analysis, working paper.

Maddala, GS, 1991, A perspective on the use of limited-dependent and qualitative
variable models in accounting research, The Accounting Review, 788-807.

Moses, O.D., 1987, Income smoothing and incentives: Empirical tests using
accounting changes, The Accounting Review, 358-377.

Nourayi, M.M. , 1994, Stock price responses to the SEC's enforcement actions,
Journal of Accounting and Public Policy, 333-347.

Press, EC. and J .B. Weintrop, 1990, Accounting-based constraints in public and
private debt agreements, Journal of Accounting and Economics, 65-95.

Ronen, J. and S. Sadan, 1981, Smoothing income numbers (Addison-Wesley)

Schipper, K., 1989, Commentary on earnings management, Accounting Horizons, 91-
102.

Skousen, KR, 1991, An introduction to the SEC: Fifth edition (South-Western
Publishing, Cincinnati, Ohio).

Stone, M. , 1982, A survey of research on the effects of corporate pension plan
sponsorship: Implications for accounting, Journal of Accounting Literature, 1-
33.

Stone, M. and R.W. Ingram, 1988, The effect of Statement No. 87 on the ﬁnancial
reports of early adopters, Accounting Horizons, 48-61.

Suh, Y.S. , 1990, Communication and income smoothing through accounting method
choice, Management Science, 704-723.

Thomas, J.K., 1988, Corporate taxes and deﬁned beneﬁt pension plans, Journal of
Accounting and Economics, 199-237.

Trueman, B. and S. Titman, 1988, An explanation for accounting income smoothing,
Journal of Accounting Research, 127-139.

US. Department of Labor, Bureau of Labor Statistics, Bulletin 2426, 1993.

Watts, R.L. and LL. Zimmerman, 1978, Towards a positive theory of the
determination of accounting standards, The Accounting Review, 112-134.

88

Watts, R.L. and J .L. Zimmerman, 1986, Positive accounting theory (Prentice Hall,
Englewood Cliffs, NJ).

Watts, R.L. and IL. Zimmerman, 1990, Positive accounting theory: A ten year
perspective, The Accounting Review, 131-156.

Winklevoss, HE, 1993, Pension mathematics with numerical illustrations, second
edition, (Irwin, Homewood, IL).

Wong, J. , 1988, Political costs and intraperiod accounting choice for export tax
credits, Journal of Accounting and Economics, 37-51.

"111111111111“