UNIVERSITY OF
ILLINOIS LIBRARY
AT URBANA-CHAMPAIGN
BOOKSTACKS
THE HECKMAN BINDERY, INC.
North Manchester, Indiana
JUST FONT SLOT TITLE
H CC 1W
H CC 1W
H CC 1W
H CC 7W
3R
aCULTY
VORKING
PAPER
8 1988
7 NO. 1468-1485
330
B385<»,CV">
no. 1468-1485
cop. 2
<IMPRINT>
U. of ILL.
LIBRARY
SANA
DMF
BINDING COPY
PERIODICAL: D CUSTOM D STANDARD □ ECONOMY □ THESIS
BOOK □ CUSTOM
Di
Di
ACCOUNT LIBRARY NEW
RUBOR TITLE ID
SAMPLE
□
NO VOLS.
THIS TITLE
LEAD ATTACH.
ACCOUNT NAME
V OP ILLINOIS
ACCOUNT INTERNAL I.D.
FOIL MATERIAL
6632 WHI 488
ISSN.
BG19124
I D. #2 NOTES
STX4
COLLATING
35
ADDITIONAL INSTRUCTIONS
BINDING WHEEL
FREQUENCY
SYS. ID.
1 3
7 6
Dept=STX'5 1 Item= . !M=[ZY#
1CR2ST3CR MARK BY # B4 91]
SEP SHEETS PTS. BD PAPER TAPE STUBS CLOTH EXT
INSERT MAT
PRODUCT TYPE
__
HBQK.
1
/] FILLER
GUM STUB
SPECIAL PREP.
LEAF ATTACH.
ACCOUNT LOT NO
ACCOUNT PIECE NO
GROUP CARD
vol mis
COVER SIZE
,/ x
47
001247981
FACULTY WORKING
PAPER NO. 1471
Accounting Measures of Unfunded Pension Liabilities
and Bond Risk Premiums (Pension Accounting and Bond
Risk Premiums)
Sara Ann Reiter
SEP 131988
NOIS
College of Commerce and Business Administration
Bureau of Economic and Business Research
University of Illinois. Urbana-Champaign
BEBR
FACULTY WORKING PAPER NO. 1471
College of Commerce and Business Administration
University of Illinois at Urbana- Champaign
July 1988
Accounting Measures of Unfunded Pension Liabilities
and Bond Risk Premiums
(Pension Accounting and Bond Risk Premiums)
Sara Ann Reiter, Assistant Professor
Department of Accountancy
Digitized by the Internet Archive
in 2011 with funding from
University of Illinois Urbana-Champaign
http://www.archive.org/details/accountingmeasur1471reit
Accounting Measures of Unfunded Pension Liabilities
And Bond Risk Premiums
Abstract
The research issue investigated in this paper is whether accounting
information on unfunded pension obligations is associated with bond market
risk measures. The study provides further evidence linking pension numbers
with bond market risk measures and assesses the usefulness of SFAS 87 pension
obligation measures.
The study uses a sample of 209 electric utility new issues between 1981
and 1984. A risk premium model is used to test the incremental explanatory
power of net pension obligation measures for bond risk premiums. The
association of pension information with bond risk premiums has not previously
been investigated. The study introduces bond risk premium methodologies
commonly used in studying finance issues into accounting research.
The principal findings of the study are that net pension obligations
measured on a termination basis as well as net pension obligation measures
which take expected future benefit increases into account are associated with
bond market risk measures. The conclusion is that SFAS 87 measures of net
pension obligation appear to be an adequate representation of the market's
assessment of future cash flow obligations represented by defined benefit
pension plans.
Key Words: Risk premiums, pensions.
1.0 INTRODUCTION
The research issue investigated in this paper is whether accounting
information on unfunded pension obligations is associated with bond risk
premiums. Accounting disclosures of unfunded pension obligations are
required to assist financial statement users in assessing the risks
associated with future cash flows of the firm. It is an important empirical
issue whether such disclosures are associated with market risk measures.
This study provides further evidence linking pension numbers with bond market
risk measures and assesses the usefulness of SFAS 87 pension obligation
measures. The study is an extension of the Selling and Stickney (1986) paper
which examines the associations of alternative measures of unfunded pension
obligations with a measure that best reflects the present value of future
expected cash flows. Instead of determining associations with a measure of
"true" pension liability, this study examines the association of alternative
net pension obligation measures with a bond market risk measure.
In order to understand the contribution of this study, it is necessary
to review what is known about the association of pension measures with firm
risk and return. Table 1 presents previous studies of the association of net
pension liabilities with measures of firm risk and return. The table is
organized by issue studied and by pension benefit measure used. In all the
studies, it is a maintained hypothesis that true net pension liabilities are
associated with the market value of the firm or various risk measures. The
issue tested is whether accounting measures of net pension liabilities are
sufficiently associated with true pension liabilities to be associated with
market return and risk measures. A number of empirical studies have explored
the association of unfunded pension liabilities with the market price of a
firm's common stock. Several studies examine the association of unfunded
pension benefits with risk measures such as common stock beta and bond
ratings. The general conclusion is that unfunded pension benefits are
reflected in common equity prices and risk measures such as beta and bond
ratings .
insert Table 1 here
This study demonstrates the association of unfunded pension obligations
with bond risk premiums. Several of the reasons for extending research to
risk premiums as the market risk measure are that risk premiums represent a
bond market assessment of risk as opposed to common stock prices or betas and
that risk premiums are continuous variables and therefore offer a finer
measurement of risk than bond ratings which are categorical variables. Bond
risk premium models have been widely used in finance to study a variety of
bond pricing issues. The sample consists of 209 new issue electric utility
bonds, debentures and long-term notes issued between 1981 and 1984. New
issues are used in order to study risk premiums because thin bond market
trading and issue characteristics which change over time make modeling of
risk premiums on outstanding bonds quite difficult. There are three
advantages to the use of utilities. First, an adequate sample size can be
attained since there are many more utility straight debt issues than
industrial straight debt issues during the early 1980' s. Second, since many
utilities routinely rely on the credit markets as a source of funds, there is
less likelihood of a self -selection bias in terms of the types of firms which
come to the credit markets during a period of high interest rates. Finally,
electric utilities represent a relatively homogeneous operating risk group
and the effects of debt equivalent items, such as unfunded pension
obligations, on Che risk premium may be more clearly observed. Unfunded
pension obligations should be associated with bond risk premiums since there
are definite risks to bondholders associated with the future cash flow
commitments represented by unfunded pension promises . The study develops a
risk premium model and investigates the association of unfunded pension
obligations with risk premiums. Measures of pension obligations using SFAS
87 measurement principles are tested for association with bond market risk
measures so that the potential usefulness of SFAS 87 disclosures is
evaluated.
One contribution of the study is that it provides further evidence
confirming the association of accounting measures of net pension obligations
with market measures of risk and return. Cross -sectional studies of the
association between accounting disclosures and market risk and return
measures are important in validating the relevance of accounting numbers.
The association of pension information with bond risk premiums has not
previously been investigated. The second contribution is the use of bond
market risk measures. This study introduces bond risk premium methodologies
commonly used in studying finance issues into accounting research and
provides some insights into choice of appropriate models and solutions to
methodological problems. The use of a bond market risk measure is an
important contribution of this study. There is very little empirical
evidence on the association of accounting numbers with bond market parameters
despite the importance of the bond markets in new capital generation and the
equal status of creditors with investors in The Statement of Financial
Accounting Concepts No. 1. "Objectives of Financial Reporting by Business
Enterprises" (FASB 1978).
Different measures of pension benefits are explained in Section 2.
Measurement of the pension variables is discussed. The research hypothesis
is stated and principal implications are introduced. The risk premium model
and the sample used in the study are described in Section 3. Model
estimation and results are presented in Section 4. Several estimation issues
relating to model specification, collinearity and cross-sectional
correlations are evaluated. Significant associations are found between risk
premiums and each of the measures of net pension obligation. Conclusions and
implications are discussed in Section 5.
2.0 PENSION MEASUREMENT
2.1 Explanation of pension benefit measures
Pension benefits are measured as the present value of future benefits
expected to be paid. The estimate of future benefits expected to be paid can
reflect plan benefit formulas applied to current salary levels using service
accumulated to date or, at the other extreme, can be based on expected salary
levels at retirement, expected future service and expected plan amendments.
Selling and Stickney (1986) and Schipper and Weil (1982) conclude that a
measure based on future salaries and expected future service provides the
best information about the future cash flow commitments of the firm. There
is a basic trade-off between relevance and reliability in that using future
salary and service estimates may decrease the reliability of pension benefit
estimates . ■*
The basic research issue in the pension association studies listed in
Table 1 is whether reported pension benefit measures are sufficiently
relevant and reliable to be reflected in market return and risk measures.
Essentially, the issue tested is whether accounting measures of net pension
benefits are sufficiently correlated with "true" net pension benefits to be
associated with market risk and return measures. Pension benefit measures
used in the studies come from pension disclosures prepared under APB 8, SEC
or SFAS 36 requirements. The most recent pension accounting standard, SFAS
87, becomes effective in 1987 for most firms (FASB 1985). No studies to date
have used SFAS 87 disclosures, although Selling and Stickney (1986) test SFAS
87 pension measures in a simulation setting. Selling and Stickney (1986) use
simulation to test directly which measure of net pension liability is most
closely correlated with the "true" net pension liability. They find that
although pension benefit measures based on different assumptions are highly
correlated with each other, the correlation between net pension benefits is
much lower. They find that net pension benefits measured as projected
benefit obligations are more highly correlated with "true" net pension
liabilities than net pension benefits measured as accumulated benefit
obligations. This finding is important because the minimum net obligation to
be recognized on the balance sheet under SFAS 87 is based on the accumulated
benefit obligation. It may be, however, that the accumulated benefit
obligation is not the most relevant measure of pension obligation. This
study extends Selling and Stickney by testing the association of various
pension benefit measures with a market risk measure rather than a simulated
"true" pension liability.
Actual SFAS 87 pension benefit disclosures cannot be used because it is
necessary to use a sample of new bond issues to study risk premiums. SFAS 87
disclosures have not been available for a sufficient number of years (firms
are not required to adopt SFAS 87 before 1987 annual reports) to develop a
new issue sample using actual SFAS 87 numbers. Approximations of SFAS 87
numbers are therefore calculated using conservative transformation techniques
and the basic measurement principles underlying SFAS 87. In addition, an
estimate of economic pension liability, which takes additional expected
future cash flow commitments into account, is calculated.
Two different pension benefit measures are used in SFAS 87 - the
accumulated benefit obligation and the projected benefit obligation. The
accumulated benefit obligation represents benefits earned to date with no
future salary growth discounted at a current annuity settlement rate. In
other words, accumulated benefit obligations represent termination benefits.
The accumulated benefit obligation forms the basis for determination of
whether a minimum pension obligation must be recognized on the balance sheet.
The projected benefit obligation takes future salary growth into account.
The projected benefit obligation is used in the determination of pension
expense and is presented in footnote disclosures. The projected benefit
obligation may not fully reflect expected future cash flows since firms do
adjust benefits of retired workers and workers with fixed benefit formulas
for inflation and it can be argued that pension benefits involve an implicit
contract with workers to make such adjustments in the future (Ippolito 1985,
1986a, b). Economic benefits represent pension benefits expected to be paid
rather than benefits contractually due to workers and are therefore more
representative of expected future cash flows (Selling and Stickney 1986 and
Schipper and Weil 1982) . An economic benefit measure is included in the
study for comparison with the projected benefit measure in order to see if
the market regards projected benefits as an adequate representation of future
cash flow commitments.
insert Table 2 here
2.2 Estimation of pension benefit measures
The pension benefit measures used in this study are (1) pension benefits
reported under SFAS 36, pension benefit measures reported under SFAS 87 - (2)
the accumulated benefit obligation and (3) the projected benefit obligation
and (4) economic pension benefits reflecting the probable future amounts to
be paid given implicit contracts with the workers. Table 2 presents a
comparison of pension benefit measures in terms of salary and service
assumptions, actuarial methods and discount rates. The benefit measures are
calculated using the Bulow and Ippolito transformation methods described
below.
A simple linear transformation procedure is suggested by Bulow (1979).
The underlying rationale is a comparison of pension benefit promises to a
consol, which is an infinite series of future cash flows. The discount rate
for a consol is one over the interest rate. Since the duration of the
pension benefit stream is less sensitive to changes in interest rates than
the duration of the consol, the assumption that the change in value of
pension benefits with regard to changes in interest rates is proportional to
one over the interest rate is a conservative assumption (Bulow, 1979, p. 49).
The Bulow transformation is a simple linear adjustment with the following
formula:
bp - LR x (iR / iT)
where Lp - Transformed liability
Lr - Reported liability
i-p - Transformed interest rate
iR - Reported interest rate
The Bulow transformation method is employed in the pension studies which use
SFAS 36 footnote data (Maher 1987, Feldstein and Morck 1982 and Landsman 1986
(uses a similar method)).
Ippolito (1986b) develops an approximation of the sensitivity of pension
benefits to variation in interest rates from empirical data using Department
of Labor reports. Ippolito derives a model of pension liabilities and
estimates an equation using data from over 4,000 plans in 1978. The results
appear to be quite reasonable since values for the constants in his equation
conform to realistic assumptions about time to retirement and average
retirement period and the sensitivity of pension benefits to changes in
interest rates for retired workers is not as great as for active workers.
Ippolito uses this transformation to estimate economic liabilities which
reflect implicit promises to adjust future benefits for inflation. Ippolito
finds that the estimated economic liabilities conform with actual wage -tenure
profiles (Ippolito 1985) and with stock market values (Ippolito 1986a).
Francis and Reiter (1987) use the Ippolito adjustment to estimate economic
pension benefits. The formulas which are used to transform pension benefits
are:
for active participants: Lt/Lr ~ exp( - .077(i-j; - iR))
for retired participants: Lj/Lr - exp( - .057(i-j' - iR) )
The Ippolito model assumes that all plans have the same average
retirement period for retired workers and the same average time to retirement
for active workers. If these assumptions have changed since 1978, it seems
likely that retirement periods are longer and that work forces are younger
(due to the effects of forced early retirements etc.) so that actual
sensitivities to interest rate changes may be greater than the model
8
indicates. The Ippolito model, therefore, represents a conservative
adjustment process.
Total benefits (vested plus nonvested) are used in the calculations
since accumulated benefits approximate termination liabilities and all
pension benefits are considered vested in a voluntary termination. Benefit
measures which assume future salary growth such as projected benefits and
economic benefits implicitly assume that benefits will become vested. The
discount rates used to determine the SFAS 36 disclosures are reported in
footnote disclosures and the average of the high and low rates is used when a
range of rates is reported. The appropriate discount rate for SFAS 87
pension measures is the annuity settlement rate. The average settlement rate
published by the Pension Benefit Guarantee Corporation (PBGC) is used for
determining SFAS 87 benefits and economic benefits. Estimates of the
percentages of retired and active workers (202 and 802 respectively) are
determined using a sample of utility pension plans from the Blue Book of
Pension Funds.
The accumulated benefit obligation is estimated by using the Bulow
method and the PBGC annuity settlement rates for each year. The projected
benefit obligation and the economic benefit obligation are estimated using
the Ippolito method. The projected benefit obligation is basically the same
as the accumulated benefit obligation for benefits belonging to retired
workers and to workers whose benefits are not tied to final pay. Therefore,
the retiree's benefits (20Z) are adjusted to the PBGC annuity rate and the
benefits of active workers are adjusted to the average of the spread between
the discount rate and rate of salary growth and the PBGC rate. This assumes
that about 50Z of the active workers have benefits tied to final pay. In
order to determine the appropriate spread between the discount rate and
assumed salary growth rate, 1986 annual reports available on NAARS for the
firms in the sample which use SFAS 87 in 1986 (N-20) are analyzed. The
average difference between the discount rate and the compensation growth rate
in this sample is about 2Z.
The accuracy of the transformation process for projected benefit
obligations is verified by comparison with the subsample of actual SFAS 87
benefit measures for 1986. The ratio of actual projected benefits to
accumulated benefits is 1.35 to 1 while the ratio of estimated accumulated
benefits to estimated projected benefits for the entire sample is 1.48 to 1.
Since interest rate levels are higher during 1981 through 1983 than in 1986,
a slightly larger spread between accumulated benefits and projected benefits
would be expected for the sample period so that the transformation procedure
is verified.
Ippolito (1986b) provides a conservative proxy for economic benefits.
Benefits for retired workers are adjusted to a rate of 1.52 plus half of
inflation to reflect the experience of the 1970 's when retired workers
received increases in benefits representing about half of inflation.
Benefits for active workers are adjusted to a real rate of 3Z.
2.3 Comparisons of benefit measures
Table 2 presents descriptive statistics for the pension variables and a
correlation matrix of pension measures. The measures of estimated pension
benefits from smallest to largest are accumulated benefit obligations,
reported benefits, projected benefit obligations and economic benefit
obligations. Accumulated benefit obligations represent termination benefits,
projected benefit obligations take part of expected future benefit increases
10
in account and economic benefit obligations reflect expected future benefits
assuming that firms adjust benefits for inflation. The average funded status
of firms (pension assets divided by pension benefits) is 1.2 using reported
benefits, 1.69 using accumulated benefit obligations, 1.14 using projected
benefit obligations and .91 using economic benefit obligations. The
correlations between the benefit measures are high (ie. .99). Correlations
between net pension assets (liabilities) are not as high. For instance, the
correlation between net pension assets based on accumulated benefit
obligations and net pension assets based on economic benefit obligations is
.62. These results coordinate with the simulation results of Selling and
Stickney (1986) which show a high correlation between different pension
benefit measures and lower correlations between net pension assets
(liabilities). Since the ranking of firms by net pension assets
(liabilities) differs between measures, there is the possibility that some
measures produce cross -sectional rankings which are more highly correlated
with market risk measures.
2.4 Research Issue
In studies of the association of unfunded pension benefits with market
measures, the real issue is whether accounting measures of pension obligation
are sufficiently relevant and reliable to be reflected. The research issue
in this study is whether net pension assets (liabilities) reported in SFAS 36
footnote disclosures and net pension assets (liabilities) measured to
approximate SFAS 87 pension measures and economic pension liabilities are
associated with bond risk premiums. Theoretically, Merton (1974) has shown
that liabilities of the firm increase the risk premium required on new debt
issues. Risk premiums are measures of the default risk of firms issuing
11
bonds. Pension obligations represent liabilities of the sponsoring firm.
The research issue is whether current accounting measurements of unfunded
pension benefits are reflected in bond risk premiums. This is basically a
measurement issue. The research hypothesis, stated in the alternate form,
is:
HI: Unfunded pension benefits information is associated with
bond risk premiums. Specifically, the coefficient of net pension
assets (liabilities) is inversely associated with risk premiums.
The research issue is tested by adding net pension asset (liability)
variables to a base or control model and testing for increases in explanatory
power. The sign and significance level of the coefficient for the pension
variable is also evaluated. Various diagnostics on the proper specification
of the model and the validity of the tests are presented in Section 4.
The second research issue, which is addressed only in an exploratory
manner, is whether different measures of net pension asset (liability) have
different degrees of association with bond risk premiums. No formal tests of
differences in association are proposed, however, informal comparisons of F
statistics will be made. Selling and Stickney (1986) find that projected net
assets (liabilities) are more highly correlated than accumulated net pension
assets (liabilities) with "true" net pension liabilities. Furthermore,
Schipper and Weil (1982), Selling and Stickney (1986) and Ippolito (1985,
1986a, b) claim that economic pension liabilities are more relevant than
termination measures. For these reasons, we would expect that economic net
pension assets (liabilities) will be most highly associated with bond risk
measures with projected net pension assets (liabilities) next most associated
and accumulated net pension assets (liabilities) least associated.
12
The results of the study may have several policy implications. The
association of net pension assets (liabilities) with bond market risk
measures provides additional research evidence that the funded position of
defined benefit pension plans has an impact on market parameters even before
the SFAS 87 requirement to recognize a minimum pension obligation. In
addition, conclusions about the usefulness of various SFAS 87 requirements
may be possible. Evidence from this study can help evaluate the usefulness
of SFAS 87 requirements that (1) pension asset and liability amounts are
separately disclosed, (2) a termination liability measure is used in
determining the minimum pension liability to be recognized and (3)
realization of probable future benefit increases is limited to future salary
growth assumptions and excludes probable other future benefit increases.
3.0 RISK PREMIUM MODEL AND SAMPLE
3.1 Risk Premium Model Development
The risk premium on corporate bonds can be defined as the difference
between the yield on a risky security and that on a security that is
risk- free but identical in all other respects. The classic study on the
determinants of bond yields is Fisher (1959) which hypothesizes that bond
risk premiums are a function of the default risk of the firm and of the
marketability of the bond issue. Appendix A summarizes a number of models
used in various studies to explain bond yields or risk premiums. Factors
found important by researchers include indenture provisions (such as term to
maturity, sinking funds and secured status) , call risk, macroeconomic factors
(such as business cycle effects) and marketability.
The dependent variable (DYIELD) is formed by subtracting the yield to
maturity of a U.S. Treasury issue from the offering yield (OFYLD) of a new
13
utility issue. Fung and Rudd (1986) indicate that it is important to use the
previous day's treasury issue closing yield to match with the offering yield
on new securities. The independent variables are chosen to proxy for
maturity and indenture characteristics, call risk, macroeconomic factors and
default risk. '
Term to maturity is expected to be directly related to risk premiums due
to the increased exposure to interest rate risk with increased time to
maturity. A variable for the presence of a sinking fund is added to the
model since the necessity of entering into complex sinking fund agreements
for the enhanced security of the borrower is related to the perceived quality
of the issuer.
The period of the study, from 1981 through 1984, is a period of high
market interest rates so that call risk is an important factor in pricing the
bonds sold. Future refinancing at lower interest rates seems probable for
many of these issues and investors are willing to pay extra for call
protection to lock in the high yields. Degree of call protection is proxied
by the difference between the yield to first call or refunding and the
offering yield. Effects from both the length of the deferment period and the
amount of the call premium are taken into account by this measure.
It is necessary to control for macroeconomic factors since the sample
period spans three years. Previous studies (Jaffee 1975 and Cook and
Hendershott 1978) find evidence that risk premiums vary with the business
cycle. A number of economic indicators are used in these studies and the
variable with the most consistent significant explanatory power is the index
of consumer sentiment. The index of consumer sentiment, which is based on
data collected by the University of Michigan and is described in detail by
14
Fair (1971), is used in this model to control for macroeconomic effects.
Since the risk premium rises as overall interest rates rise (Cook and
Hendershott 1978), the level of interest rates is also included as an
independent variable.
Financial ratios are used to proxy for default risk. Evidence of the
connection between various financial ratios and default risk of utilities is
gathered from Standard & Poor's Rating Guide (1979), Melicher's (1974) factor
analysis of utility ratios and Altman and Katz ' s (1976) bond rating
prediction study. The following categories of factors are found to be
important: cash flow adequacy, asset protection, capitalization and earning
stability. Variables representing cash flow adequacy, capitalization and
earnings protection are cash flow to construction expenditures, the
debt- equity ratio and the property funding ration (long term debt to
property, plant and equipment). The coefficient of variation of return on
equity for five years represents earning stability. Pretax interest coverage
is one of the most important financial ratios used by bond raters (Standard &
Poor's 1979). One potential drawback in using a utility sample is that
during the 1980' s, factors which are not reflected in the financial ratios of
utilities, such as potential problems with bringing new plants on-line, begin
to significantly and rapidly alter the risk of several utilities. A dummy
variable NUKE is included for utilities which are experiencing problems
connected with their nuclear generating facilities at the time of the bond
o
issue . °
Table 3 summarizes the risk premium model variables and expected signs.
Sample descriptive statistics are presented in Table 4.
insert Tables 3 and 4 here
15
3.2 Sample
The sample consists of new issues of public utility bonds between
February 23, 1981 and February 29, 1984. The starting date of the study
coordinates with the earliest availability of pension footnote disclosures
mandated by SFAS No. 36. Issues between February 23, 1981 and February 29,
1984 are included in the sample if the issuers are considered to be electric
utilities by Moody's Public Utility Manual and a full set of pension and
financial information is available. Lack of publicly available pension
footnote information causes 22 observations to be dropped. Because it is not
comparable with other bond issues, one deep discount issue is not included in
the sample. The final sample consists of 209 issues.
The offering date, offering yield and other terms of each issue,
including indenture terms, are gathered from Moody's Bond Survey. The
Investment Dealer's Digest, and Moody's Public Utility Manual. Descriptive
information necessary to code the NUKE variable comes from Standard & Poor's
CreditWeek analysis of new issues. Treasury yields are from the Wall Street
Journal.
The primary source for financial variables is Standard & Poor's
CreditWeek and secondary sources are annual reports and Moody's Public
Utility Manual. One advantage of using CreditWeek data is timeliness. In
many cases the financial data is reported up to the nearest quarter to the
issue date and capitalization data are pro forma. The information used to
form the pension variables is collected from the FASB 36 pension data bank
(Version2, Columbia University) and from annual reports.
16
4. ESTIMATION AND RESULTS
4.1 Estimation Issues
insert Table 5 here
The risk premium model is estimated using Ordinary Least Squares
regression. Results are reported in Table 5. The increase in explanatory
power for the addition of pension variables is evaluated using the general
linear test. The formula is:
F* - SSE (R) - SSE (?) / SSE (F)
d.f.R - d.f.F d.f.F
Where SSE (R) and SSE (F) and d.f.R and d.f.F are the sum of squared errors
and degrees of freedom for the reduced and full models respectively. F* is
distributed by the F distribution with ((d.f.F - d.f.R), d.f.F) degrees of
freedom (Neter and Wasserman 1974) .
There are three potential problems which are important in evaluating
results. First, spurious results could arise if the model is not specified
properly. Second, results could be influenced by severe collinearity .
Finally, cross-sectional correlations could affect the statistical
significance of the results. These three potential problems are evaluated
and I find that the model appears to be well-specified, that collinearity
between pension and other variables is not a problem and that the
significance of the results is not generated by cross-sectional correlations.
4.1.1 Model Specification
One facet of model fit is explanatory power. The control model, without
the net pension asset (liability), has an adjusted R-squared of 60. 5X which
is typical for a risk premium model. All variables have the expected signs
except for the debt-to-equity ratio and all coefficients except for
17
coefficient of variation of return on equity are significantly different from
zero at a 10X significance level. The unexpected sign of debt-to-equity
appears to be due to a collinearity problem between debt-to-equity and the
property funding ratio which is discussed further in Section 4.1.2.
Since the sample period spans three years, I test to see if different
values of the financial ratios and pension variables would be expected in
different years. The financial ratios and pension variables are calculated
at December 31, 1980, 1981 and 1982 for the 22 electrical utilities in the
Standard & Poor's 40 utilities index. T- tests are performed to see if the
levels of the financial ratios and pension variables are different for this
group of firms between the three years. No significant differences in
financial ratios or pension variables is found. Therefore, no bias is
introduced by including financial ratios and pension variables of issues
spanning this three year period in the same model.
Ordinary least squares assumptions of normality of the dependent
variable and residuals are met. Tests for normality (Stevens 1974) are
performed for the dependent variable and residuals. The null hypothesis of
normality cannot be rejected in either case. A Goldfeld-Quandt test
(Goldfeld and Quandt 1965) is performed to test for heteroscedasticity . The
resulting F statistic is not significant (1.25 for degrees of freedom 71,
71). The conclusion is that the model appears to be well-specified and that
results are, therefore, not caused by the pension variables proxying for the
effects of incorrect model specification.
4.1.2 Collinearity Problems
Another concern when evaluating results is that severe collinear
problems may affect the results. Collinearity diagnostics (Belsley, Kuh and
18
Welsch 1981) indicate that there are strong collinear associations in the
sample between three groups of variables: the intercept, the index of
consumer sentiment, the level of interest rates and interest coverage; the
debt- to -equity ratio and the property funding ratio; and the level of
interest rates, the property funding ratio and interest coverage. No strong
collinear associations involve the pension variables, however. Another
diagnostic for collinear problems, the adjusted R-square of a regression of
the pension variable on the other independent variables, is reported in Table
6. The pension variables are not highly associated with the other
independent variables. The simple correlations between pension variables and
the other independent variables are presented in Table 6. The highest simple
correlation is the .29 correlation between the debt-to-equity ratio and the
reported net pension asset (liability). This level of correlation is well
below the threshold level needed to cause collinearity problems (Belsley, Kuh
and Welsch 1981) . Collinearity between the pension variables and the other
independent variables is not, therefore, biasing the results.
insert Table 6 here
4.1.3 Cross -sectional Correlations
Finally, it is possible that cross-sectional correlations within the
sample lead to an overstatement of statistical significance of the
coefficients. The sample observations span the time period between February
24, 1981 and February 22, 1984 so that there is no concentration in calendar
time. Another problem may arise due to multiple issues by the same firm.
The 209 issues included in the sample represent 72 separate issuers. This is
because many utilities routinely come to the bond market on a yearly basis.
If the model is not well-specified, individual issuer financial condition may
19
not be well controlled for and correlations between the residuals of issues
by the same firm could result in overstatement of the statistical
significance of the results. In order to see is this is an important factor,
the model is run on a subsample consisting of only one issue per issuer.
Significance levels are similar for the subsample and the full sample. This
indicates that multiple issues do not lead to overstatement of significance
levels. F statistics for the increase in explanatory power in the single
issue sample are reported in Table 6.
4.2 Results
The results of the risk premium model tests are presented in Table 5.
When the reported net pension asset (liability) is added to the control
model, there is a significant increase in explanatory power (F-10.25). In
addition, the coefficient is negative and significant as expected. Similar
results are obtained when the accumulated net pension asset (liability)
(F-6.89), projected net pension asset (liability) (F-7.75) and economic net
pension asset (liability) (F-4.17) variables are added to the control model.
In conclusion, all the pension measures are significantly associated with
risk premiums.
Counter to expectations, economic and projected net assets (liabilities)
are not more highly associated with risk premiums than accumulated net assets
(liabilities). In fact, it seems that economic net assets (liabilities) are
the least closely associated. This result is not consistent with Selling and
Stickney (1986), which finds that projected net assets (liabilities) are more
highly correlated with "true" net pension assets (liabilities) than
accumulated net pension assets (liabilities). Since the economic net asset
(liability) measure takes more expected future cash flows into account, it
20
was expected to be more closely related to a market risk measure than
accumulated or projected benefit obligations.
5.0 CONCLUSIONS
The principal finding of this study is that accounting measures of net
pension asset (liabilities) (or simple transformations of accounting
measures) are associated with bond market risk measures. This indicates that
market risk and return measures reflect net pension assets (liabilities) even
before balance sheet recognition is required. The different pension benefit
measures, accumulated, projected and economic benefits, are highly correlated
but net pension assets (liabilities) formed with the different measures are
less highly correlated. Therefore, the SFAS 87 requirements for separate
disclosure of pension assets and liabilities appear to be justified. The use
of accumulated net assets (liabilities) as the basis for liability
recognition appears to be justified also, since accumulated net assets
(liabilities) are as closely associated with bond risk premiums as net asset
measures which take future benefit increases into account. Since economic
net assets (liabilities) are less closely associated with risk premiums than
projected net assets (liabilities), it appears that SFAS 87 pension measures
are an adequate representation of the market's assessment of future cash flow
obligations despite the fact that the projected benefits measure only takes a
portion of expected future benefit increases into account. Therefore, SFAS
87 disclosures appear to provide optimal information to users while taking a
conservative position on premature realization of obligations.
One limitation of the study is the use of estimated SFAS 87 pension
measures. Results should, therefore, be considered preliminary in nature.
The principal qualification of the research methodology is that specification
21
of an appropriate model is extremely important in achieving internal
validity. Although the risk premium modeling approach used is not common in
accounting research, it is a widely used methodology in finance studies.
Furthermore, diagnostics of model fit do not indicate any problems with
misspecif ication. Finally, results using a utility sample may not be fully
generalizable to industrial firms. Creditors may view pension obligations of
utilities in a different manner and may not be as interested in evaluating
long-term cash flow commitments as when examining non- regulated firms. It is
possible, therefore, that creditors might evaluate termination benefits,
projected benefits and economic benefits differently for regulated and
industrial firms.
The author gratefully acV^oviodees the support of the Peat Marwick Research
Fellowship program. :es to thank members of her dissertation
committee, particularly Jert Francis and Doug Emery, and participants of the
workshop at the University of Illinois for their helpful comments on earlier
drafts of this paper.
22
1. Lys (1984) finds that the debt-equity ratio has little power to explain
debt default risk unless variables to control for total firm risk are
included in the model. In addition, capital structure research indicates that
there are different typical debt levels for firms across industry groups
(DeAngelo and Masulis 1980 and Bowen, Daley and Huber 1982).
2. The fact that utilities are regulated industries does not invalidate
their use in this study. Public utility regulation does not guarantee
returns to bondholders or payment of employee pensions. Rate-making is often
not particularly timely, a phenomena known as regulatory lag. In times of
inflation and rising fuel prices, utilities suffer from problems of attrition
(replacement costs of plant and equipment exceed historical costs) and
erosion (actual operating expenses exceed those embedded in the rates) . In
many ways, utilities face an environment not very different from that of
competitive firms (Howe and Rasmussen 1982) .
3. In addition, accountants may be constrained by the concept of
realization from using future service estimates in determining the present
value of benefits to be paid in the future.
4. Economic benefits differ from the "true" pension liability in Selling
and Stickney (1986) in that the present value of future expected service is
not incorporated. The economic benefit measure simply reflects expectations
that benefits will be adjusted for expected future inflation.
5. Statistically significant increases in explanatory power do not imply
that there is a large economic benefit to be earned by considering the
additional factor. It is basically interesting to know that accounting
numbers which are designed to be helpful in assessing risk are associated
with market measures of risk and return. The bulk of what we know about
23
accounting numbers is based on association tests. It is not reasonable to
expect that addition of an incremental piece of accounting information to any
but an extremely misspecified model would result in a dramatically large
increase in explanatory power.
6. Gonedes and Dopuch (1974) assert that associations of alternative
accounting disclosures with market measures cannot determine which disclosure
is "best" because of market imperfections. As Lev and Ohlson (1982) point
out, however, there is an intrinsic value in knowing that accounting measures
which are designed to be helpful is assessing risk are correlated with market
risk and return measures. Therefore, it is of interest to note which measure
of pension benefits is most closely associated with bond risk measures.
7. Two factors often mentioned in other studies are not controlled for
explicitly in this model: coupon tax effects and marketability. When bonds
are purchased at a substantial discount, a portion of the expected return is
the capital gain on the difference between maturity value and purchase price
and this capital gain advantage is priced by the market. In a study using
new issues, coupon tax effects are not important, however. Although
marketability does play a role in bond pricing, there has been little support
for the marketability proxies used in previous studies.
8. An association between regulatory climate and bond ratings has been
demonstrated (Pinches, Singleton and Jahankhani 1978). Various agencies, for
example Value Line, provide ratings of regulatory climate by state. Use of
these rankings would provide a more objective measure of regulatory climate
but because of the speed with which circumstances surrounding the
construction of nuclear facilities change within the time period of this
study, the more timely CreditWeek information is used.
24
REFERENCES
Abdel-khalik, A. R. , Thompson, R. B. and Chen, R. D. 1981. Effects on bond
pricing and on bond rating. In The Economic Effects on Leases of FASB No.
13. Accounting for Leases (Abdel-khalik, A.R. , principal researcher)
Stamford, Connecticut: Financial Accounting Standards Board, pp. 163-191.
Altman, E. I. and Katz , S. 1976. Statistical bond rating classification using
financial and accounting data. In Proceedings of the Conference on Topical
Research in Accounting (Schiff, M. and G. Sorter, eds . ) New York: New York
University, pp. 205-239.
Barrett, W. B., Heuson, A. J. and Kolb, R. W. March 1986a. The effect of
Three Mile Island on utility bond risk premia: A note. The Journal of
Finance 41(1): 255-261.
. Winter 1986b. The differential effects of sinking funds on bond
risk premia. The Journal of Financial Research 9(4): 303-312.
Belsley, D. A., Kuh, E. and Welsch, R. E. 1980. Regression Diagnostics
New York: John Wiley & Sons.
Berndt, E. R. , Sharp, K. C. and Watkins , G. C. December 1979. Utility bond
rates and tax normalization. Journal of Finance 34(5): 1211-1220.
Bowen, R. M. , Daley, L. H. and Huber, C. C, Jr. Winter 1982. Evidence on
the existence and determinants of inter- industry differences in leverage.
Financial Management 11(4): 10-20.
Bulow, J. I. 1979. Analysis of pension funding under ERISA. National Bureau
of Economic Research Working Paper No. 402.
Cook, T. Q. and Hendershott, P. H. September 1978. The impact of taxes, risk
and relative security supplies on interest rate differentials. The Journal
of Finance 33(4): 1173-1186.
Daley, L. A. April 1984. The valuation of reported pension measures for
firms sponsoring defined benefit plans. The Accounting Review 59(2):
77-198.
DeAngelo, H. and Masulis, R. W. March 1980. Optimal capital structure under
corporate and personal taxation. Journal of Financial Economics 8(1):
3-29.
Dhaliwal, D. S. October 1986. Measurement of financial leverage in the
presence of unfunded pension obligations. The Accounting Review 61(4):
651-661.
Ederington, L. , Yawitz, J. B. and Roberts, B. H. Fall 1987. The
informational content of bond ratings. The Journal of Financial Research
10(3): 211-226.
25
Fair, R. C. 1971. A Short-run Forecasting Model of the United States Economy
Lexington, Massachusetts: Heath.
Feldstein, M. and Morck, R. 1982. Pension funding decisions, interest rate
assumptions and share prices. National Bureau of Economic Research
Working Paper No. 938.
Feldstein, M. and Seligman, S. September 1981. Pension funding, share
prices, and national savings. The Journal of Finance 36(4): 801-824.
Financial Accounting Standards Board. 1978. Statement of Financial Accounting
Concepts No. 1: Objectives of Financial Reporting by Business
Enterprises . Stamford, Connecticut: Financial Accounting Standards Board.
. 1985. Statement of Financial Accounting Standards No. 87:
Employers' Accounting for Pensions. Stamford, Connecticut: Financial
Accounting Standards Board.
Fisher L. June 1959. Determinants of risk premiums on corporate bonds.
The Journal of Political Economy 67(3): 217-237.
Francis, J. R. and Reiter, S. A. April 1987. Determinants of corporate
pension funding strategy. Journal of Accounting and Economics 9(1):
35-59.
Fung, W. K. H. and Rudd, A. July 1986. Pricing new corporate bond issues:
An analysis of issue cost and seasoning effects. The Journal of
Finance 41(3): 633-643.
Goldfeld, S. and Quandt, R. June 1965. Some tests for homoskedasticity .
Journal of the American Statistical Association 60(310): 539-547.
Gonedes, N. J. and Dopuch, N. 1974. Capital market equilibrium, information
predictions, and selecting accounting techniques: Theoretical
framework and review of empirical work. Studies On Financial Accounting
Objectives: 1974 Supplement to Journal of Accounting Research 12: 48-129.
Howe, K. M. and Rasmussen, E. F. 1982. Public Utility Economics and Finance.
Englewood Cliffs, New Jersey: Prentice-Hall, Inc.
Ippolito, R. A. December 1985. The labor contract and true economic pension
liabilities. American Economic Review 75(5): 1031-1043.
. January- February 1986a. The economic burden of corporate pension
liabilities. Financial Analysts Journal 42(1): 22-34.
, 1986b. Pensions. Economics and Public Policy. Homewood, Illinois:
Dow- Jones - Irwin.
Jaffee, D. M. July 1975. Cyclical variations in the risk structure of
interest rates. Journal of Monetary Economics 1(3): 309-325.
26
Kidwell, D. S., Marr , M. W. and Thompson, G. R. April 1987. Shelf
registration: Competition and market flexibility. Journal of Law and
Economics 30(1): 181-206.
Landsman, W. October 1986. An empirical investigation of pension and
property rights. The Accounting Review 61(4): 662-691.
Lev, B. and Ohlson, J. A. 1982. Market-based empirical research in
accounting: A review, interpretation, and extension. Studies on Current
Research Methodologies in Accounting: A Critical Evaluation Supplement to
Journal of Accounting Research 20: 249-322.
Lys , T. April 1984. Mandated accounting changes and debt covenants. Journal
of Accounting and Economics 6(1): 39-65.
Maher, J. J. October 1987. Pension obligations and the bond credit market:
An empirical study. The Accounting Review 62(4): 785-798.
Martin, L. J. and Henderson, G. V., Jr. December 1983. On bond ratings and
pension obligations, A note. Journal of Financial and Quantitative
Analysis 18(4): 463-470.
Melicher, R. W. March 1974. Financial factors which influence beta
variations within an homogeneous industry environment. Journal of
Financial and Quantitative Analysis 9(1): 231-242.
Merton, R. C. May 1974. On the pricing of corporate debt: The risk
structure of interest rates. The Journal of Finance 29(2): 449-470.
Neter, J. and Wasserman, W. 1974. Applied Linear Statistical Models
Homewood, Illinois: Richard D. Irwin, Inc.
Oldfield, G. S. February 1977. Financial aspects of the private pension
system. Journal of Money. Credit, and Banking 9(1): 48-54.
Pinches, G. E. , Singleton, J. C. and Jahankhani, A. Summer 1978. Fixed
coverage as a determinant of electric utility bond ratings. Financial
Management 7(2): 45-55.
Rogowski, R. J. and Sorensen, E. H. Spring 1985. Deregulation in investment
banking: Shelf registrations, structure, and performance. Financial
Management 14(1): 5-15.
Schipper, K. and Weil, R. L. October 1982. Alternative accounting
treatments for pensions. The Accounting Review 57(4): 806-824.
Selling, T. I. and Stickney, C. P. Winter 1986. Accounting measures of
unfunded pension liabilities and the expected present value of future
pension cash flows. Journal of Accounting and Public Policy 5(4):
267-285.
27
Smith, R. L. July 1987. The choice of issuance procedure and the cost of
competitive and negotiated underwriting: An examination of the impact of
Rule 50. The Journal of Finance 42(3): 703-720.
Standard & Poor's Rating Guide. 1979. New York: McGraw-Hill, Inc.
Stevens, M. A. September 1974. EDF statistics for goodness of fit and some
comparisons. Journal of the American Statistical Association 69(347):
730-737.
28
Appendix A
Yield Studies
Author (s)
Saiple
and data
Abdel-
1975-79
khalik,
178 bonds
Thompson
i Chen
[1981]
Barrett,
1977-B2
Meuson,
Public
Kolb
Utilities
[1986a]
76 bonds
Issue investigated
Association of risk
prenuis with
capitalized lease
obligations
Effect of Three Nile
Island on utility
bond risk premiums
Dependent Factors represented by independent variables
variable
Risk Log of coefficient of variation in earnings(t),
preiiui Log of larket value of equity to book value of
(log) debt it), Log of market value of all traded
debt (I)
R
square
.48 -
.90
Risk Ten to maturity!!), First Mortgage!!), Sinking .63
premiui fund it), Log of issue size, Bond rating
duMies(t), Discount factor (t), Level of
interest rates(t), Change in industrial
production (1), Change in lonetary policy (t),
Shape of yield curved), Three Nile Island
vanables(t)
Barrett,
1977-83
Effect of sinking
Risk
Heuson i
76 public
funds on bond
preiiui
Kolb
utility
yields
[1986b]
issues
Berndt,
1962-76
Effect of rate
Tax
Sharp,
Electric
normalization for
Adjusted
Hatkins
Utilities
deferred taxes
Yields
[1979]
90-93
Cook i
1961-73
Deteriinants of the
Risk
Hendershott Yield series risk preiiui
preiiui
[1978]
Utility Aa
Ederington,
2728/79 t
Deter ii runts of bond
Yield to
Yawitz fc
2/27/81
yields
■aturity
Roberts 176 and
[1987] 180 bonds
Ten to taturity(t), First aortgage(t), Log of .63
issue size, Bond rating dunies(t), Discount
factor (t), Call preiiui(t), Level of interest
rates(t), Shape of yield curved), Change
in industrial production (t), Change in monetary
policy (t) , Sinking fund variables(t)
Debt to equity(t), Current assets, Coupon (t), .48-. 85
Exchange listing, Rate normalization (!) , Change
in loney supply, Change in 6NP growth (!),
Market power, Location
Index of consumer sentiment (t), Employment
index, Call(t), Level of interest rates(t),
Relative supply
Financial ratios!!), Bond ratings(t), .80-. 87
Subordinate status! t), Call and capital
gain(t), Period of call protection
Fisher
1927,1932,
Deteriinants of the
Risk
[19591
1937,1949
risk preiiui
preiiui
1933
(log)
45-88 firms
Fung fc
1983-83
Effects of shtlf
Risk
Rudd
123 -em
registration
premium
[1986]
issues
Jaffee
1954-69
Cyclical variations
Risk
[1975]
Quarterly
in the risk structure
1 spreads
between
ratings
Log of earnings variability!!), Log of period .73-
of solvency (t), Log of equity to debt ratio(t),
Log of market value of bonds outstanding!!)
.81
Level of interest rates!!), Log of ten to .79
maturity!!), Log of years to first call, Log
of issue size, Quality!!), Industrial sector!!),
Financial sector, Shelf registration
Index of consumer sentiment!!), Unemployment .65-. 75
rate, Growth rate of retained earnings!!),
6rowth of capital investment (I), 6rowth
of output. Baa interest level!!), Term
to laturity(l), Total float, Coupon!!)
Appendix A
Yield Studies
KidMell ,
1982-83
Effects of shelf
Harr &
111 new
registration
Thoapson
issues
(1987]
Rogoaski &
1981-83
Effect of shelf
Sorensen
307 new
registration
C1985]
issues
Risk Rating duaaies(t), Level of rates(t), Interest .78
prenui volatility (t) , Sinking fund ( t) , Call feature,
Log of issue size(t), Nuaber of bids(t),
Sale iet hod
Offering
yields
Saith
1977-85
Choice of under-
Yield to
[1987]
380 new
issues
utilities
writing aethods
issuer
Level of interest rates(t), Trend of rates, .88
Supply of new issues! I), Log of issue size,
Bond rating duaaies(t), Average aaturity(t),
Call protection aeasure, Interaction of call
and aaturity, Shelf registration!!)
Rating duaaies(t), Log of issue size, Log of
nuaber of issues by fira(t), Call protection,
First call preaiua, Years to first call,
Characteristics of investaent banker, Variance
in interest rate level, Level of interest rates(t)
Table 1
PENSION STUDIES
Association of unfunded
pension obligations with:
Theoretical Pension Liability
Measures
Selling and Stickney (1986)
Market Value of Equity and/or
Stockholders' Equity
Landsman (1986)
Daley (1984)
Feldstein and Morck (1982)
Feldstein and Seligman (1981)
Oldfield (1977)
Systematic Risk - Common Stock Beta
Dhaliwal (1986)
Bond Ratings
Maher (1987)
Martin and Henderson (1983)
Pension Measures Used
SFAS 87 - Simulated Benefits
SFAS 36 disclosures: Total and
Standardized
APB 8 disclosures - Unfunded Vested
Benefits and Pension Expense and SEC
disclosures - Unfunded Past Service Cost
SFAS 36 disclosures - Vested and
Standardized
APB 8 disclosures - Unfunded Vested
Benefits and SEC disclosures - Unfunded
Past Service Cost
APB 8 disclosures - Unfunded Vested
Benefits
APB 8 disclosures - Unfunded Vested
Benefits
SFAS 36 disclosures - Vested, Total and
Standardized
SEC disclosures - Unfunded Past Service
Cost
31
Table 2
COMPARISON OF PENSION BENEFIT MEASURES
Benefit measures:
Reported under SFAS 36
Assumptions :
Salary
Current
salaries
Service
Accumulated
to date
Actuarial
method
Unit credit
Discount
rate
Various
Accumulated benefits
SFAS 87
Current
salaries
Accumulated
to date
Unit credit
Annuity
settlement
rates
Projected benefits
SFAS 87
Expected
future
salaries
Accumulated
to date
Unit credit
Annuity
settlement
rates
Economic benefits
Expected
future
salaries
and benefit
increases
Accumulated
to date
"True" pension benefits
Expected
future
salaries
and benefit
increases
Expected
DEFINITION OF PENSION VARIABLES
Reported net asset (liability)
(Fair market value of plan assets -
Reported pension benefits) / Permanent
capitalization
Accumulated net asset (liability) (Fair market value of plan assets
Accumulated benefit obligation) /
Projected net asset (liability)
Economic net asset (liability)
Permanent capitalization
(Fair market value of plan assets -
Projected benefit obligation) /
Permanent capitalization
(Fair market value of plan assets -
Economic benefit obligation) / Permanent
capitalization
32
Table 2 - continued
DESCRIPTIVE STATISTICS
Reported pension benefits
Accumulated benefit obligation
Projected benefit obligation
Economic benefit obligation
Reported funding ratio
Accumulated funding ratio
Projected funding ratio
Economic funding ratio
Reported net asset (liability)
Accumulated net asset (liability)
Projected net asset (liability)
Economic net asset (liability)
Mean Standard
Minimum
Maximum
Deviation
173.828
197
.783
5.620
1
,103.000
122.412
140
.705
4.738
779.899
180.263
205
.192
6.539
1
,143.695
226.622
259
.362
7.926
1
,438.250
1.193
.364
.602
2.645
1.698
.677
.722
5.103
1.139
.375
.580
2.769
.908
.298
.461
2.146
.006
.014
-.031
.053
.020
.015
-.020
.062
.004
.014
-.039
.049
-.009
.017
-.069
.041
PEARSON CORRECTION COEFFICIENTS
Reported benefits
Accumulated benefits
Projected benefits
Economic benefits
Reported net asset
(liability)
Accumulated net asset
(liability)
Projected net asset
(liability)
Economic net asset
(liability)
Reported
Accumulated
Projected
Economic
benefits
benefits
benefits
benefits
1.000
.987
.996
.997
1.000
.997
.995
1.000
.999
1.000
Reported
Accumulated
Projected
Economic
net asset
net asset
net asset
net asset
(liability)
(liability)
(liability)
(liability)
1.000
.783
.947
.859
1.000
.849
.617
1.000
.933
1.000
33
Table 3
Risk Premium Model
Variable Expected Description
Sign
Dependent Variable
DYIELD Risk premium
Maturity and Issue Characteristics
MATYR + Years to maturity
SF + Sinking fund
Political and Regulatory Risk
NUKE + Trouble with nuclear plant
Call Risk
DFYLD - Offering yield - Yield to first call
Macroeconomic Factors
MOOD - Index of consumer sentiment
TYIELD + Level of Treasury yields
Financial Variables
CONST - Cash flow to construction expenditure
DE + Debt- to -equity ratio
PROP + Property funding ratio
ROE + Coefficient of variation of return on equity
COV - Pretax interest coverage
Pension Variables
SUNB - (Pension plan assets - reported benefits)
to permanent capitalization
TUNB - (Pension plan assets - accumulated benefits)
to permanent capitalization
PUNB (Pension plan assets - projected benefits)
to permanent capitalization
EUNB - (Pension plan assets - economic benefits)
to permanent capitalization
34
Variable
Table 4
DESCRIPTIVE STATISTICS
Sample Descriptives N-209
Mean Standard Minimum Maximum
deviation
DATE
2/
'25/*
51
2/2?
>/84
ISSUE SIZE
84
.4928
43
.1312
10
250
(Million $)
COUPON RATE
14.
,7992
1,
.9368
10.
875
18.
75
%
OFFERING YIELD
14,
,8779
1.
,9467
10.
95
18.
75
X
TREASURY YIELD
12.
5356
1.
,5238
9.
45
15.
78
X
YEARS TO MATURITY
20,
,7034
10,
,2304
5
33
Years
PERIOD OF CALL
OR
REFUNDING PROTECTION
5.
2895
2,
0331
0
30
Years
NO CALL OR REFUND PROTECTION
CALL PROTECTION
REFUNDING PROTECTION
FIRST MORTGAGE
Number
Percent
Coded 1
in Sample
2
1.1%
21
10.0%
186
88. 9X
201
96. IX
Model Descriptives
Variable Mean
Standard Minimum Maximum Number Percent
Deviation Coded 1 in Sample
Dependent Variable
DYIELD 2.3422 .8879
.43
4.82
Independent Variables
MATYR
20.7034
10.2304
5
33
SF
87
41.63
NUKE
61
29.19
DFYLD
-1.1539
.6116
-3.57
0
MOOD
74.8584
10.3838
62
100.1
TYIELD
12.5356
1.5238
9.45
15.78
CONST
29.2928
28.3328
-96
132
DE
50.1239
5.3128
31.80
67
PROP
45.1029
5.7715
30.09
72.1
ROE
.1274
.0716
.01
.46
COV
2.6281
.6269
1.63
4.77
Distribution of Sample Issues by Year
Year
Number
Percent
1981
69
33. 0Z
1982
78
37.3%
1983
57
27. 3X
1984
5
2.4X
35
Table 5
REGRESSION RESULTS
N-209, Dependent variable - DYIELD
Reduce
d Model
Full Model
Full Model
Contrc
1 Model
Add Reported
Add Ac
cumulated
net pension
net pension
asset
(liability)
asset
(liability)
Variable
Pre-
Coeffi
T Prob.
Coeffi
T Prob.
Coeffi
T Prob
dicted
cient
Stat.
cient
Stat.
cient
Stat.
sign
Intercept
3.285
3.394 <.001
3.213
3.396 <.001
3.579
3.727 <.001
MATYR
+
.039
5.937 <.001
.039
6.129 <.001
.039
6.146 <.001
SF
+
.188
2.309 .011
.173
2.162 .016
.177
2.191 .015
NUKE
+
.218
2.294 .011
.189
2.024 .022
.217
2.312 .011
DFYLD
-
-.241
-2.455 .008
-.232
-2.422 .008
-.236
-2.438 .008
MOOD
-
-.035
-7.779 <.001
-.035
-7.901 <.001
-.036
-8.069 <.001
TYIELD
+
.176
4.999 <.001
.175
5.097 <.001
.173
4.981 <.001
CONST
-
-.009
-4.875 <.001
-.008
-4.689 <.001
-.008
-4.331 <.001
DE
+
-.019
-2.231 .013
-.012
-1.406 .081
-.016
-1.827 .035
PROP
+
.013
1.567 .059
.010
1.214 .113
.010
1.211 .114
ROE
+
.712
1.213 .113
.619
1.079 .141
.544
.935 .176
COV
-
-.463
-5.108 <.001
-.494
-5.540 <.001
-.500
-5.530 <.001
REPORTED NET
PENSION ASSET
(LIABILITY)
-
-9.293
-3.201 <.001.
ACCUMULATED NET
PENSION ASSET
(LIABILITY)
-
-7.206
-2.626 .009
Adjusted R- Square
60.50
62.27
61.65
F Statistic *
10.248
6.894
* The F statistics are from general linear tests of differential explanatory
power of the full models over the reduced model (without pension variables). F*
at a significance level of .10 Is approximately 2.75 for degrees of freedom (1,
196).
36
Table 5 - continued
Full Model
Full Model
Add Pr
ojected
Add Ec
onomic
net pension
net pension
asset
(liability)
asset
(liability)
Variable
Pre-
Coeffi
T Prob .
Coeffi
T Prob
dicted
cient
Stat.
cient
Stat.
sign
Intercept
3.297
3.464 <.001
3.141
3.263 <.001
MATYR
+
.040
6.181 <.001
.040
6.103 <.001
SF
+
.169
2.101 .018
.171
2.106 .018
NUKE
+
.188
1.991 .024
.180
1.876 .031
DFYLD
-
-.236
-2.448 .008
-.240
-2.468 .007
MOOD
-
-.036
-8.044 <.001
-.035
-7.900 <.001
TYIELD
+
.175
5.047 <.001
.178
5.101 <.001
CONST
-
-.008
-4.858 <.001
-.009
-5.089 <.001
DE
+
-.014
-1.572 .059
-.015
-1.712 .044
PROP
+
.011
1.255 .106
.012
1.371 .086
ROE
+
.657
1.138 .128
.744
1.278 .101
COV
-
-.489
-5.462 <.001
-.476
-5.279 <.001
PROJECTED NET
PENSION ASSET
(LIABILITY)
-
-7.794
-2.784 .003
ECONOMIC NET
PENSION ASSET
(LIABILITY)
-
-4.866
-2.049 .021
Adjusted R- Square
61.81
61.13
F Statistic *
7.752
4.169
* The F statistics are from general linear tests of differential explanatory
power of the full models over the reduced model (without pension variables). F*
at a significance level of .10 is approximately 2.75 for degrees of freedom (1,
196).
37
Table 6
REGRESSION DIAGNOSTICS
CORRELATION BETWEEN PENSION VARIABLES AND OTHER INDEPENDENT VARIABLES
Pearson correlation coefficients -
* significant at .05 level
Reported Accumulated Projected Economic
net asset net asset net asset net asset
(liability) (liability) (liability) (liability)
MAT
-.022
.042
.024
.014
SF
-.062
-.072
-.072
-.064
NUKE
-.119
-.043
-.132
-.195*
DFYLD
.032
-.026
-.014
-.021
MOOD
.029
-.015
-.035
-.063
TYIELD
.027
-.020
.032
.093
CONST
.107
.162*
.058
-.034
DE
.296*
.189*
.259*
.250*
PROP
.064
.008
.044
.048
ROE
.022
-.042
.022
.060
COV
-.117
-.064
-.109
-.125
MULTIPLE R2**
.073
.050
.057
.083
** Coefficient of determination between each pension variable and all other
independent variables.
F STATISTICS - SINGLE ISSUE MODEL
Reported Accumulated Projected Economic
net asset net asset net asset net asset
(liability) (liability) (liability) (liability)
One observation per
issuer N-71
10.438
4.938
8.944
6.389
* The F statistics are from general linear tests of differential explanatory
power of the full models over the reduced model (without pension
variables). F* at a significance level of .10 is approximately 2.79 for
(1,59).
38
HECKMAN
BINDERY INC.
JUN95
Bod •Ta-PW' N. MANCHESTER.
INDIANA 46963