Earnings management using pension rate estimates and the timing of adoption of SFAS 87.

James, Marianne L.

ABSTRACT

Concerns about earnings management and its effect on the usefulness of financial reporting continue to be pervasive and have prompted the Securities and Exchange Commission (SEC) to increase its enforcement activities. Flexibility in applying accounting provisions frequently provides opportunities for earnings management. Statement of Financial Accounting Standards No. 87, Employers' Accounting for Pensions, (SFAS 87), provides extensive flexibility in choosing required pension estimates, and in timing the adoption of the standard. Several studies suggested that early adopters of SFAS 87 were motivated by its positive income effect.

This study investigated the relationship between the timing of adoption of SFAS 87 and earnings management using pension rate estimates subsequent to adoption. Study of 1,035 firms found that early adopters tended to use higher estimates of the rate-of-return on pension plan assets (ROR) subsequent to adoption, than on-time adopters. In the absence of higher actual returns, these findings suggest that early adopters were using the ROR estimates to facilitate earnings management subsequent to adoption. This may provide important information for the Financial Accounting Standards Board in assessing the effect of multi-year transition periods on earnings management.

INTRODUCTION

Concerns about earnings management continue to be pervasive, and prompted the SEC to increase its regulatory enforcement activities to both prevent and detect earnings management. For example, in 1998, the SEC informed 150 firms that it may be reviewing their earnings reports and directed its Division of Corporate Finance to appoint a special task force to investigate potential earnings management problems (MacDonald, 1999, A2-6). In several of his speeches, Arthur Levitt, until recently chair of the SEC, and other SEC officials, such as former Chief Accountant, L.E. Turner have spoken out against earnings management that firms practice within the boundaries of Generally Accepted Accounting Principles (GAAP).

The potential for earnings management tends to arise when accounting standards provide for flexibility, require extensive estimates, and when long time horizons are involved. Accounting for defined benefit pensions combines all three of these criteria: it requires extensive estimates (e.g., of longevity, employee turn-over rates, and pension rates), permits flexibility in applying some of its provisions (e.g., in choosing pension rates), and typically involves an unusually long time horizon (i.e., employees for whom pensions benefits are currently accrued may not retire for several decades). In addition, SFAS 87, like a number of other recent accounting standards, provided for a long, multi-year transition period. In fact, the standard permitted firms to choose an adoption date from among three years.

Pension accounting evolved from the initial cash-as-you-go-basis to accrual accounting. SFAS 87 represents the newest in a series of accounting standards and was expected to improve the relevance and reliability of financial reporting related to defined benefit pension plans. However, Ali and Kumar (1993) found that the new standard provides more earnings management opportunities, than were available under the prior authoritative standard, APBO 8, Accounting for the Cost of Pension Plans. These enhanced earnings management opportunities arise in part from the provision that permits the discount rate to differ from the ROR, and due to income smoothing incentives (Ali & Kumar, 1993).

Prior research also found independent evidence of (1) an association between adoption timing of SFAS 87 and increased earnings at time of adoption (e.g., Langer & Lev, 1993; Tung & Weygandt, 1994), and (2) manipulation of pension rates subsequent to adoption (Blankley, 1992; Blankley & Swanson, 1995; Amir & Benartzi, 1998). Both of these findings are consistent with earnings management. A review of the related literature yielded no studies investigating the potential relationship between early adoption of the pension accounting standard and choice of pension rate estimates subsequent to adoption.

The purpose of this study was to investigate the relationship between timing the adoption of SFAS 87 and firms' choice of pension rate estimates subsequent to adoption. A sample of firms that adopted SFAS 87 were identified from the Standard and Poor COMPUSTAT industry files, and related financial statement data and disclosures were extracted. The relationships between adoption timing and pension rate estimates (expected ROR on pension plan assets, discount rate, and salary trend rate) were evaluated through multiple regression analyses.

This study found evidence of a significant positive relationship between early adoption of SFAS 87 and the ROR estimates utilized by the sample firms subsequent to adoption for each of the four years tested. That is, early adopters tended to use higher ROR estimates than on-time adopters, and these were not supported by higher actual returns. These findings are consistent with earnings management and thus may provide an alternate or additional explanation for the adoption timing decision. Thus, while previous studies linked SFAS 87 adoption timing to earnings management only at time of adoption, this study linked adoption timing to earnings management subsequent to adoption. These findings are important, as they may assist the Financial Accounting Standards Board (FASB) in assessing whether the intended purpose of long multi-year transition periods for standards requiring extensive estimates is being achieved.

REVIEW OF LITERATURE

Schipper defines "Earnings Management" as a "purposeful intervention in the external financial reporting process, with the intent of obtaining some private gain..." (1989, 92). The implication of this definition is that earnings may reduce the representational faithfulness and thus the usefulness of financial reporting. Arthur Levitt, in his now famous speech "The Numbers Game" asserted that earnings management "if not addressed soon, will have adverse consequences for America's financial reporting system" (Levitt, SEC, 1998). Firms that manage earnings usually do not restrict their efforts to a single accrual, instead they tend to use a portfolio of accruals, including pension accruals (Zmijewski and Hagerman, 1981). Using a portfolio approach, some firms manage earnings only when perceived necessary, or when a special opportunity arises (e.g., a new earnings management tool becomes available; or the adoption timing of a new standard). For others, earnings management may be a more ongoing practice (Schipper, 1989).

Standards with long transition periods may provide a one-time earnings management opportunity by permitting firms to choose the timing of the effect of the change on the financial statements. Standards may also provide earnings management opportunities by providing flexibility in choosing extensive required estimates.

During the past few decades FASB has issued a number of accounting standards with long, multi-year transition periods and flexibility in choosing extensive required estimates. For example, the transition period was four years for SFAS 106, Employers' Accounting for Postretirement Benefits Other than Pensions, and three years for SFAS 109, Accounting for Taxes. SFAS 87 requires extensive estimates, permits considerable flexibility in choosing those estimates, and provided for a three-year transition period.

The primary purpose of long transition periods for new accounting standards is to assist firms in generating reliable estimates. The purpose stated in SFAS 87 is: "to give time for employers and their advisors to assimilate the requirements ... [,] to obtain the information required....[and] to renegotiate or to obtain waivers of provisions of some legal contracts" (FASB 1985, par. 259-260). A number of studies have examined the timing of adoption of accounting standards with long transition periods, including the adoption of SFAS 87. Some of the major studies are discussed in the following paragraphs.

Recent studies found evidence that income smoothing, a form of earnings management, was associated with the adoption timing of new accounting standards. For example, Gujarathi and Hoskin (1992) found evidence of income smoothing in their study of 284 early adopters of SFAS 96, Accounting for Income Taxes. Their study showed that some entities that experienced significant income-increasing effect from the standard's early adoption, would otherwise have reported a loss or no earnings increase. Their follow-up study (Gujarathi & Hoskin,1995) provided additional evidence suggesting that the long transition period was utilized by early adopters to "present their financial picture in the best light" (p. 28), and that those firms with the largest positive income effect on adoption tended to adopt at the earliest possible date.

Studies on early adoption of SFAS 52 also found evidence of income smoothing. For example, Ayres (1986) investigated the relationships between choice of adoption date and proximity of debt and dividend-constraints, control of entities, size, and earnings. They found that early adopters tended to experience lower rates of income growth prior to adoption of the Standard, that they tended to be smaller, closer to debt and dividend-constraint violations, and tended to have smaller percentages of stock owned by directors and managers, than late adopters.

A number of studies also examined adoption timing of SFAS 87 and found evidence consistent with earnings management. SFAS 87 was issued in December 1985; its effective date, the date for which adoption was required, was for fiscal years ended after Dec. 15, 1986. Thus, for most firms on-time adoption was for their 1987 fiscal period. However, since early adoption of the standard was permitted, firms could choose to adopt the standard for fiscal years ending in 1985, 1986, or 1987.

Fifty-one percent of the entities surveyed in Accounting Trends and Techniques that disclosed defined benefit pension plans early-adopted SFAS 87 during 1985 and 1986 (based on information reported by the AICPA, 1987, p. 274). Most of these early adopters reported increased income from the adoption. A number of studies found a positive association between the positive income effect and the timing of adoption. For example, Tung and Weygandt (1994) investigated whether debt contracting and/or political cost influenced adoption timing, and found that early adopters reported significantly higher increases in income than on-time adopters, and that of 1,011 early adopters only 35 reported decreased income associated with their adoption of SFAS 87.

Langer and Lev (1993) investigated both compliance costs and investor perceptions and found that the motive to increase earnings consistently discriminated between early and late (on-time) adopters of SFAS 87; these results also held true regardless of the quarter during which the standard was adopted. Other studies such as Norton (1989), Stone & Ingram (1988), and Senteney & Strawser (1990) also found that early adopters reported income increases at time of adoption.

Norton (1989) found that early adopters also had better funded pension plans, and that those firms that adopted during the earliest possible year had better funded plans than those that adopted later; i.e., 1985 adopters had better funded plans than 1986 adopters, and 1986 adopters had better funded plans than 1987 adopters. Thus, early adopters generally were not required to record additional minimum liabilities that would have increased debt-to-equity ratios.

SFAS 87 not only provided a one-time opportunity for earnings management at time of adoption, but also provides continued opportunities for earnings management subsequent to adoption through the choice of extensive required estimates, including pension rate estimates. SFAS 87 requires that firms estimate three pension rates: the discount rate, the expected ROR on pension plan assets, and the salary trend rate, and utilize them in calculating pension expense and pension benefit obligations. The standard does not prescribe specific pension rates or set specific limits, and instead provides suggested guidelines. For example, it emphasizes (FASB 1985, par. 43) that the "best" estimate for each future event should be given for each rate.

Choice of these three rates tends to significantly affect pension expense and the pension benefit obligations. The ROR estimate is utilized in calculating the return on plan assets and pension expense, with a higher ROR estimate typically resulting in lower expense and higher income. The discount rate also affects the interest component of pension expense, the pension benefit obligations reported in the financial notes, and the minimum pension liability. The salary trend rate affects the current service cost component of pension expense and the projected benefit obligation disclosed in the notes.

Firms typically engage actuaries to assist in deriving estimates, including pension rate estimates. However, actuaries' primary responsibility is to ensure that pension plans are adequately funded (Fogarty & Grant, 1995), and management ultimately decides which actuary to hire, and what pension rate estimates to use in accounting for defined benefit pension plans. Thus, firms have considerable flexibility in choosing estimates, and as a result, the rates utilized by firms vary considerably. To illustrate, information reported by Accounting Trends and Techniques for fiscal year 1992 indicated that six percent of the entities disclosing discount rates utilized rates of 7.5 percent or less, while 18 percent utilized rates of 9.0 percent or greater (AICPA, 1993, p. 300). Furthermore, for fiscal year 1994, 19 percent utilized rates of 7.5 percent or less, while 7 percent utilized rates of 9 percent or more (AICPA, 1995, p. 342).

Pension rates provide an excellent tool for earnings management, as rates can be chosen to uniformly increase or decrease income, or one rate can be chosen to offset the effect of another. For example, the salary trend rate can be decreased, along with an increase in the estimated ROR on pension plan assets (both decreasing expense); or an increase in the salary trend rate can be offset partially or totally by an increase in the expected ROR.

While SFAS 87 was issued to improve the relevance and reliability of financial reporting, Ali and Kumar (1993) found that this new standard provides more opportunities for earnings management than APBO 8 did. Some of these opportunities may arise from a provision that permits the interest rate used for calculating the return on pension plan assets to differ from that used for calculating the interest cost component of pension expense; i.e., the ROR and the discount rate do not have to be the same (Ali & Kumar, 1993).

Several other studies found that the extensive pension rate estimates required under SFAS 87 may provide opportunities for earnings management. For example, Mittelstaedt (1989) and Thomas (1989) found that prior to pension plan terminations, actuarial assumptions, (such as the discount rate), were frequently manipulated. Ghicas (1990) found that firms tended to increase discount rates to reduce funding requirements in periods of cash shortages. Blankley (1992) and Blankley and Swanson (1995) linked the choice of pension rates to the opportunistic behavior of managers. Amir and Benartzi (1998) found that the actual ROR on pension plan assets was not associated with the estimated ROR.

Another recent study examined rate choices under SFAS 106, Employers Accounting for Postretirement Benefits Other than Pensions, which is very similar to and deals with similar issues as SFAS 87. Amir and Gordon (1996) investigated cross-sectional variations in firms' choice of the discount rate and healthcare cost trend rates, and found that firms that chose more conservative estimates tended to have lower debt-to-equity ratios, smaller postretirement benefit obligations, more extreme (high or low) earnings, and more significant postretirement plan amendments. They also found that investors based their valuations of an entity on reported, instead of actual rates; the authors interpreted this as another incentive for using less conservative estimates that facilitate earnings management goals. This last finding also supports the contention that using pension rates for earnings management (which is investigated in this study) also would not be detected easily.

HYPOTHESES DEVELOPMENT

Firms that manage earnings usually do not restrict their efforts to a single accrual, instead they tend to use a portfolio of accruals (Zmijewski & Hagerman, 1981). Several studies (e.g., Tung & Weygandt, 1994, Senteney & Strawser, 1990, and Langer & Lev, 1993) showed that most early adopters of SFAS 87 increased income upon adoption. This suggests that early adoptions may have been motivated by a desire to manage income.

However, SFAS 87 can be used to manage income not only upon adoption, but also subsequent to adoption through the choice of pension rate estimates. Ali and Kumar (1993) found that SFAS 87 provides more opportunities for earnings management than the previous accounting standard did. Because pensions typically have long time horizons, such earnings management cannot be detected easily.

Firms that were motivated to early-adopt a standard that increases opportunities for earnings management may be more likely to utilize this new standard to manage earnings subsequent to adoption. The following hypotheses were tested to investigate the relationships between the timing of adoption of SFAS 87 and pension rate estimates subsequent to adoption.

H1: Firms that adopted SFAS 87 early choose different ROR estimates
than firms that adopted on the effective date.

H2: Firms that adopted SFAS 87 early choose different discount
rates than firms that adopted on the effective date.

H3: Firms that adopted SFAS 87 early choose different salary trend
rates than firms that adopted on the effective date.

For testing Hypotheses H1, H2, and H3, the independent variable of interest was the adoption year, and the dependent variables were the pension rates. All variables are defined in the methodology section.

METHODOLOGY

A sample of 1,035 firms that adopted SFAS 87 and disclosed pension rate and other financial information for the 1991-1994 fiscal periods were chosen from the COMPUSTAT industry files. The relationships between adoption timing and pension rate estimates were evaluated utilizing multiple linear regression. The years 1991-1994 were chosen as study period to avoid compounding of too many additional factors with that of interest. Thus, the period surrounding the initial adoption was excluded, and the analysis restricted to four years to avoid the influence of new factors on the rates chosen (i.e., economic changes, new accounting standards and regulation that may affect choice of rates).

The dependent variables were the pension rates used and disclosed by the firms; that is, the ROR for testing H1, the discount rate for testing H2, and the salary trend rate for testing H3. The independent variable for testing all three hypotheses was the adoption year of SFAS 87 (YRADOPT). A dichotomous classification, similar to that used in most studies on timing choice (e.g., Tung & Weygandt,1994; Langer & Lev, 1993, Senteney & Strawser, 1990), was chosen; a value of one (1) was assigned to the independent variable for firms that adopted SFAS 87 early (in fiscal years 1985 or 1986), and a value of zero (0) for firms that adopted on the effective date.

Other factors may have influenced the timing of adoption, and may influence the choice of pension rate estimates. A number of studies have investigated these factors (e.g., Tung & Weygandt, 1994, Stone & Ingram, 1988, Senteney & Strawser, 1990, Norton, 1989). The variables found significant in one or more of pertinent prior studies were included as control variables; these were measured in the year prior to the year tested and scaled by beginning-of-year total assets. The control variables were:

1 Earnings variable: Prior-year earnings may create earnings
expectations and influence subsequent year accruals and pension rate
choices. Previous SFAS 87 studies (e.g., Tung & Weygandt, 1994, and
Langer & Lev, 1993) found that earnings influenced adoption timing.
In this study, earnings were measured by income before extraordinary
items (EARN).

2 Funding: Senteney and Strawser (1990) and Norton (1989) found that
firms with well-funded plans tended to adopt SFAS 87 early. Funding
status also may affect choice of pension rates, particularly choice
of the discount rate, which can influence the funding status. The
unfunded or over funded accumulated pension benefit obligation was
included as a control variable (FUND).

3 Debt Contracting: The more likely an entity is to violate a
debt-covenant ratio, the less likely it is to make income-decreasing
accounting choices (Watts & Zimmerman 1986, 216). Tung and Weygandt
(1994) found the debt/equity ratio (a measure of leverage frequently
used as proxy for debt-covenant tightness) to be significantly higher
for early adopters of SFAS 87, than for on-time adopters. In this
study, the debt/equity ratio (DE) was utilized to measure leverage.

4 Pension Benefit Obligation: The size of the pension benefit
obligation may affect the interest component of pension expense, and
may influence the choice of pension rates, particularly the discount
rate. Amir and Gordon (1996) found that the relative size of the
postretirement benefit obligation affects firms' selection of health
care and discount rates. This study used the scaled accumulated
pension benefit obligation to measure the pension benefit obligation
(ABO).

5 International or Domestic firm: The model included a dichotomous
variable to control for potential differences between domestic and
foreign firms. A value of 0 was assigned to domestic firms and a value
of 1 to foreign firms (FOREIGN).

6 Size variable: Firm size may influence firms' desire to
increase/decrease accruals (Watts & Zimmermann, 1986). The most
frequently used measures of size are the natural logs of sales revenue
and total assets; both have been found to be valid measures of size
(e.g., Tung & Weygandt, 1994). In this study, size was measured by the
natural logarithm of total assets (SIZE).

The statistical model tested for hypotheses H1, H2, and H3 was:

H1 H2 H3

[ROR.sub.it], [DISC.sub.it], [SAL.sub.it = [b.sub.o] + [b.sub.1][x.sub.it1] + [b.sub.2][x.sub.it2] + [b.sub.3][x.sub.it3] + [b.sub.4][x.sub.it4] + [b.sub.5][x.sub.it5] + [b.sub.6][x.sub.it6] + [b.sub.7][x.sub.it7] + [e.sub.it], where [x.sub.it1] was the independent variable (YRADOPT), and [x.sub.it2] ... [x.sub.it7] were the control variables (EARN, FUND, DE, ABO, FOREIGN, and SIZE); [e.sub.it] = stochastic error term.

Since the year of adoption (YRADOPT) was the variable of interest, this study tested [H.sub.o]: [b.sub.1] = 0, against [H.sub.1]: [b.sub.1] [not equal to] 0 for Hypotheses H1, H2, and H3.

EMPIRICAL RESULTS

Of the 1035 sample firms, 490 (47 percent) adopted the new pension accounting standard prior to its effective date. Some firms were omitted from the final sample due to missing data. The means, medians, and standard deviations of the pension rates, and the sample sizes used in the analysis are presented in Table 1. Table 2 presents the mean, median, and standard deviations for the control variables.

The mean estimated RORs utilized by the firms were 9.02, 8.96, 8.67, and 8.78 for years 1991, 1992, 1993, and 1994, respectively. The rates ranged from a minimum of 0 to a maximum of 13 percent.

The mean discount rates were 8.33, 817, 7.44, and 8.13 for years 1991, 1992, 1993, and 1994, respectively. The decrease between 1991 and 1993 was consistent with the overall decrease in interest rates during this period. The rates ranged from a minimum of 4.2 to a maximum of 13 percent.

For all years tested, the mean estimated RORs exceeded the mean discount rates. Differences in these two rates may facilitate smoothing of periodic pension expense and the pension liability/asset, since a higher (lower) ROR decreases (increases) periodic pension expense, while a lower (higher) discount rate increases (decreases) the present value of the pension benefit obligation and may increase (decrease) pension expense. The standard deviations for the estimated RORs were higher than those for the discount rate, indicating more extreme inter-firm variations in the estimated ROR, than in the discount rate.

The mean salary trend rates were 5.63, 5.44, 4.96, and 4.95 for the years 1991, 1992, 1993, and 1994, respectively. The year-to-year decreases were consistent with overall lower rates of salary and wage increases during this period of low inflation.

Table 3 presents Pearson correlation matrixes for the pension rates in 1991-1994. Since the sample sizes were large for each year, the coefficients of correlation obtained between the discount rate and the estimated ROR were highly significant. This suggests that estimated ROR and discount rates were correlated.

However, further analysis showed that the coefficients of correlation between the estimated and the actual RORs were not significant, and positive for some years and negative for others. These findings are consistent with those of Amir and Benartzi (1998) who also did not find a significant correlation between actual and estimated RORs, and interpreted their findings as a potential indication of earnings management.

Hypotheses H1, H2, and H3 tested the effects of SFAS 87 adoption timing on the pension rate estimates chosen by the firms subsequent to adoption.

Hypothesis H1 tested the effect of adoption timing on the ROR chosen by sample firms for fiscal periods 1991-1994. Regression analysis showed that the coefficients for YRADOPT were positive and highly statistically significant for each of the years 1991-1994 (p-values <0.01). These results suggest that early adopters tended to utilize higher ROR estimates, than on-time adopters.

Higher ROR estimates decrease pension expense and increase reported income; this may indicate an earnings management strategy if the higher ROR estimates are not supported by higher actual returns. To further investigate this, the relationship between the timing of adoption and actual RORs earned by the pension plan assets was analyzed. This analysis showed that the pension plans of early adopters did not tend to earn higher actual RORs, than those of on-time adopters for any of the years tested.

Hypothesis H2 tested the relationship between adoption timing and the estimated discount rate chosen by the firms subsequent to adoption.

Regression analysis found that the adoption year (YRADOPT) was statistically significant for years 1991 and 1992 (p-values < 0.01), but not for 1993 and 1994. The coefficients were positive, which means that early adopters tended to use significantly higher discount rate estimates during 1991 and 1992, than on-time adopters. Higher discount rates reduce the present value of the reported pension benefit obligations, which tends to reduce the minimum pension liability that must be recognized by firms sponsoring the pension plans, and may reduces the interest cost component of pension expense. Because of the typically high magnitude of the pension benefit obligations of many plans, even a small difference in the discount rate can significantly change reported expenses and liabilities.

Hypothesis H3 tested the relationship between adoption timing and the salary trend rate subsequent to adoption.

Regression analysis showed that the adoption year (YRADOPT) was not statistically significant for any of the years tested. Thus, the estimated salary trend rates utilized by the early adopters was not significantly different from those utilized by the on-time adopters.

Significant Control Variables

Funding status (FUND) was significantly related to the ROR chosen by firms subsequent to adoption (p-values <0.01). The regression coefficients were negative, which suggests that firms with underfunded pension plans were more likely to choose higher ROR estimates than those that had well-funded plans. Funding status did not appear to effect choice of the discount and salary trend rates.

The relative size of the accumulated pension benefit obligation (ABO) was statistically significantly related to all three pensions rates. The coefficients were positive for the ROR and the discount rate hypotheses, and negative for the salary trend rate hypothesis. The positive coefficient for the ROR regression suggests that firms with relatively high pension benefit obligations chose higher RORs, which decrease pension expense and increase income. The negative coefficient for salary trend rate suggests that firms with relatively large pension benefit obligations choose lower salary trend rates, which decrease the projected pension benefit obligation and pension expense. The positive coefficient for ABO in the discount rate regression suggests that firms with relatively large pension benefit obligations tend to choose higher discount rates, which tends to decrease the vested, accumulated, and projected pension benefit obligations and may also decrease pension expense for pension plans with long time horizons.

Firm size was significantly related to the estimated ROR and the discount rates chosen subsequent to adoption. The coefficients were positive which suggests that large firms tended to use higher ROR estimates and higher discount rates than small firms. Size was not significantly related to salary trend rate choice.

Thus, firms with underfunded pension plans, and large firms were more likely to choose higher ROR estimates. Firms with relatively large pension plans tended to use higher ROR and discount rate estimates, and lower salary trend rates. The other control variables (EARN, DE, FOREIGN) were not consistently statistically significant in any of the regression analyses.

CONCLUSIONS

Several studies suggested that firms that early-adopted SFAS 87 were motivated by the income increasing effect at time of adoption of the new pension standard (e.g., Langer and Lev, 1993; Tung & Weygandt, 1994). However, SFAS 87 can be used to manage earnings not only upon adoption, but also subsequent to adoption through the choice of pension rate estimates (the ROR, discount rate, and salary trend rate). This study investigated the relationship between adoption timing and choice of pension rate estimates subsequent to adoption.

This study found that early adopters tended to use higher ROR estimates than on-time adopters for all the years tested. In addition, early adopters tended to use higher discount rates during two of the years tested. In the absence of higher actual returns, this suggests that early adopters were utilizing the ROR estimates to manage earnings.

These findings may provide an additional explanation for the timing of adoption of SFAS 87. The findings may also provide useful input to FASB for assessing (1) whether its intended purpose for long multi-year transition periods is being achieved, and (2) whether future accounting standards requiring extensive estimates should combine the flexibility of choice of estimates with the flexibility of choice of the adoption year.

Acknowledgment

I am grateful to the members of my dissertation committee for their guidance, especially to Professor Harvey Hendrickson, my major professor, to whose memory this paper is dedicated, and Professor Arun Prakash, who also provided useful suggestions on this paper.

REFERENCES

Ali, A., and K.R. Kumar (1993). Earnings Management Under Pension Accounting Standards: SFAS 87 Versus APB 8. Journal of Accounting, Auditing & Finance, 8(4), 427-446.

American Institute of Certified Public Accountants (1966, November). Accounting for the Cost of Pension Plans. Accounting Principles Board Opinion No. 8.

American Institute of Certified Public Accountants (1986, 1987, 1988,1991, 1992, 1993, 1994, 1995). Accounting Trends and Techniques.

Amir, E, & Gordon, E. (1996, Summer). Firms' Choice of Estimation Parameters: Empirical Evidence from SFAS 106. Journal of Accounting, Auditing and Finance, 11(3), 427-448.

Amir, E., & S. Benartzi, S. (1998). The Expected Rate of Return on Pension Funds and Asset Allocation as Predictors of Portfolio Performance. The Accounting Review, 73(3), 335-352.

Ayres, F. L. (1986). Characteristics of Firms Electing Early Adoption of SFAS 52. Journal of Accounting and Economics, 64(3), 143-158.

Blankley, A. I. (1992). Incentives in Pension Accounting: An Empirical Study Investigation of Reported Rate Estimates. Dissertation, Texas A&M University.

Blankley, A.I., & E.P. Swanson (1995). A Longitudinal Study of SFAS 87 Pension Rate Assumptions. Accounting Horizons, 9(4), 1-21.

Financial Accounting Standards Board (1981, December). Foreign Currency Translation. Statement of Financial Accounting Standards No. 52.

Financial Accounting Standards Board (1985, December). Employers' Accounting for Pensions. Statement of Financial Accounting Standards No. 87.

Financial Accounting Standards Board (1987, December). Accounting for Income Taxes. Statement of Financial Accounting Standards No. 96.

Financial Accounting Standards Board (1990, December). Employers' Accounting for Postretirement Benefits Other Than Pensions. Statement of Financial Accounting Standards No. 106.

Financial Accounting Standards Board (1992, February). Accounting for Income Taxes. Statement of Financial Accounting Standards No. 109.

Fogarty, T. J., & J. Grant (1995) Impact of Actuarial Profession on Financial Reporting. Accounting Horizons, 9(3), 23-33.

Ghicas, D. C. (1990). Determinants of Actuarial Cost Method Changes for Pension Accounting and Funding. The Accounting Review, 65(2), 384-405.

Gujarathi, M. R., & R.E. Hoskin (1992). Evidence of Earnings Management by the Early Adopters of SFAS 96. Accounting Horizons, 6(4) 18-31.

Gujarathi, M. R., & R.E. Hoskin (1995). Managers' Use of Adoption Timing and Choice of Transition Method as a Strategic Financial Reporting Opportunity: Additional Evidence from the Early Adopters of SFAS 96. Working Paper. Bentley College, University of Connecticut.

Langer, R., & B. Lev (1993). The FASB's Policy of Extended Adoption for New Standards: An Examination of FAS No. 87. The Accounting Review, 68(3), 515-533.

Levitt, A. (1998, Sept. 28). The Numbers Game. Remarks of the Chairman of the Securities and Exchange Commission. Retrieved August 10, 2001 from www.sec.gov/news/ speech/speecharchive/1998/spch220.txt.

MacDonald, E. (1999, January 22). SEC Weighs Wide Review of Write-Offs. The Wall Street Journal, A2-6.

Mittelstaedt, H. F. (1989). An Empirical Analysis of the Factors Underlying the Decision to Remove Excess Assets from Overfunded Pension Plans. Journal of Accounting and Economics. (November), 11(4), 399-418.

Norton, C. (1989). Transition to New Accounting Rules: The Case of FAS 87. Accounting Horizons, 3(4), 40-48.

Schipper, K. (1989). Commentary on Earnings Management. Accounting Horizons, 3(4), 91-101.

Senteney, D. L., & J.R. Strawser (1990). An Investigation of the Association Between Financial Statement Effects and Management's Early Adoption of SFAS 87. Review of Business and Economic Research, 25(2), 12-22.

Stone, M.S., and R. Ingram (1988). The Effect of Statement No. 87 on the Financial Reports of Early Adopters. Accounting Horizons, 2(3), 48-61.

Thomas, J. K. (1989). Why Do Firms Terminate Their Overfunded Pension Plans. Journal of Accounting and Economics, 11(4), 361-398.

Tung, S. S., & J.J. Weygandt (1994). The Determinants of Timing in the Adoption of New Accounting Standard: A Study of SFAS No. 87, Employers' Accounting for Pensions. Journal of Accounting, Auditing, and Finance, 9(2), 325-337.

Watts, R. L., & J.L. Zimmerman (1986). Positive Accounting Theory. Englewood Cliffs, N.J.: Prentice-Hall, p. 216.

Zmijewski, M. E., & R.J. Hagerman (1981). An Income Strategy Approach to the Positive Theory of Accounting Standard Setting/Choice. Journal of Accounting and Economics, 3 (1), 129149.

Marianne L. James, California State University, Los Angeles

Table 1: Descriptive Statistics--Pension Rates in Percentage

Pension Rate 1991 1992 1993 1994
ROR
Mean 9.02 8.96 8.67 8.78
Median 9.00 9.00 8.83 9.00
Std. Deviation 1.29 1.34 1.25 1.21
Sample Size 743 757 766 752
Discount Rate
Mean 8.33 8.17 7.44 8.13
Median 8.50 8.21 7.50 8.25
St. Deviation 0.65 0.62 0.52 0.56
Sample Size 724 756 759 749
Salary Trend Rate
Mean 5.63 5.44 4.96 4.95
Median 5.95 5.50 5.00 5.00
Standard Deviation 1.05 1.03 0.98 0.98
Sample Size 668 681 676 655

Table 2: Descriptive Statistics--Control Variables
Scaled by Total Assets

Variable 1991 1992 1993 1994
EARN
Mean 0.03 0.02 0.02 0.03
Median 0.03 0.02 0.02 0.03
Std. Deviation. 0.03 0.03 0.03 0.03

FUND
Mean 0.03 0.03 0.02 0.02
Median 0.00 0.02 0.01 0.01
Std. Deviation 0.06 0.06 0.07 0.06
DE
Mean 2.18 2.30 2.28 2.23
Median 1.53 1.58 1.70 1.63
Std. Deviation 3.44 3.80 2.53 2.47
ABO
Mean 0.12 0.13 0.14 0.15
Median 0.08 0.08 0.08 0.09
Std. Deviation 0.13 0.17 0.17 0.19
SIZE
Mean 6.82 6.90 6.87 6.92
Median 6.85 6.90 6.97 7.03
Std. Deviation 1.89 1.91 1.91 1.91

YRADOPT = Adoption year

EARN = Income before extraordinary items,
divided by total assets
FUND = Accumulated pension benefit obligation minus
pension plan assets, divided by total assets
DE = Total debt divided by total equity
ABO = Accumulated pension benefit obligation
divided by total assets
SIZE = Natural Log of Total Assets
FOREIGN = Domestic or Foreign Company: for all years
tested, the mean percentage of foreign companies
was 3.58 percent.

Table 3: Correlation Matrix Using Pooled 1991-1994 Pension Rates
Numbers are Pearson Correlations

 DISC 1991 DISC 1992 DISC 1993 DISC 1994

DISC 1991 1 0.84 0.48 0.40
DISC 1992 1 0.53 0.47

 DISC 1991 DISC 1992 DISC 1993 DISC 1994

DISC 1993 100.00% 0.25
DISC 1994 1
EROR 1991
EROR 1992
EROR 1993
EROR 1994

 DISC 1991 DISC 1992 DISC 1993 DISC 1994

AROR 1991 0.03 0.00 0.13 -0.11
AROR 1992 -0.032 -0.01 -0.14 0.09
AROR 1993 -0.041 -0.04 -0.03 0.03
AROR 1994 -0.036 -0.05 -0.13 0.05
SAL 1991 0.09 0.06 0.09 0.04
SAL 1992 0.01 0.08 0.06 0.02
SAL 1993 -0.15 -0.13 0.17 -0.14
SAL 1994 -0.11 -0.06 0.25 -0.04

 AROR 1991 AROR 1992 AROR 1993 AROR 1994

AROR 1991 1 -0.06 0.28 -0.22
AROR 1992 1 0.11 0.13
AROR 1993 1 0.01
AROR 1994 1
SAL 1991
SAL 1992
SAL 1993
SAL 1994

 EROR 1991 EROR 1992 EROR 1933 EROR 1994

DISC 1991 0.47 0.48 0.43 0.40
DISC 1992 0.37 0.43 0.40 0.39

 EROR 1991 EROR 1992 EROR 1993 EROR 1994

DISC 1993 0.17 0.20 0.24 0.22
DISC 1994 0.25 0.30 0.34 0.42
EROR 1991 1 0.91 0.78 0.62
EROR 1992 1.00 0.88 0.75
EROR 1993 1 0.88
EROR 1994 1

 EROR 1991 EROR 1992 EROR 1993 EROR 1994

AROR 1991 0.04 -0.06 -0.13 -0.14
AROR 1992 -0.05 -0.06 -0.06 -0.07
AROR 1993 -0.12 -0.15 0.04 0.05
AROR 1994 -0.05 -0.05 0.01 0.02
SAL 1991 0.01 -0.01 0.02 0.05
SAL 1992 -0.02 -0.02 0.01 0.04
SAL 1993 -0.13 -0.13 -0.08 -0.07
SAL 1994 -0.09 -0.09 -0.05 -0.05

 SAL 1991 SAL 1992 SAL 1993 SAL 1994

AROR 1991 -0.01 -0.01 -0.01 -0.05
AROR 1992 0.25 0.02 0.00 0.00
AROR 1993 0.00 0.02 -0.05 -.05
AROR 1994 -0.07 -0.05 -0.05 -.01
SAL 1991 1 0.81 0.63 0.61
SAL 1992 1 0.75 0.69
SAL 1993 1 0.78
SAL 1994 1

DISC = Discount rate used in calculating pension benefit
obligations and interest component of expense

EROR * = Estimated ROR used in calculating pension expense

AROR * = Actual ROR achieved

SAL = Estimated salary trend rate used in calculating
projected pension benefit obligation and service
component of pension expense

* E and R was added to the ROR variable to distinguish
between the estimated ROR (which is used in calculating
pension expense), and the actual ROR achieved by the
pension plan assets.

Table 4: Relationship Between Estimated ROR and Choice of
Adoption Date of SFAS 87
Model: [ROR.sub.it] = [b.sub.o] + [b.sub.1] [x.sub.it1] +
[b.sub.2] [x.sub.it2] + [b.sub.3] [x.sub.it3] + [b.sub.4]
[x.sub.it4] + [b.sub.5] [x.sub.it5] + [b.sub.6] [x.sub.it6]
+ [b.sub.7] [x.sub.it7] + [e.sub.it]

Year Yradopt Earn Fund De

 [B.sub.1] [B.sub.2] [B.sub.3] [B.sub.4]
1991
coeff. 0.43 0.77 -4.43 0.00
t-stat. 6.45 1.58 -7.52 -0.05
p-val. 0.00 0.11 0.00 0.96
1992
coeff. 0.36 0.55 -3.26 0.00
t-stat. 5.58 1.26 -6.94 -0.34
p-val. 0.00 0.21 0.00 0.74
1993
coeff. 0.35 0.87 -2.75 0.00
t-stat. 5.74 1.85 -5.71 -0.33
p-val. 0.00 0.07 0.00 0.74
1994
coeff. 0.26 0.29 -2.11 0.00
t-stat. 4.28 1.05 -4.89 0.19
p-val. 0.00 0.29 0.00 0.85

Year Abo Foreign Size Sample

 [B.sub.5] [B.sub.6] [B.sub.7] Size
1991
coeff. 1.19 -0.26 0.10 743
t-stat. 4.60 -1.25 5.48
p-val. 0.00 0.21 0.00
1992
coeff. 1.07 -0.25 0.11 757
t-stat. 4.71 -1.18 6.19
p-val. 0.00 0.24 0.00
1993
coeff. 0.61 -0.71 0.12 766
t-stat. 2.92 -3.65 6.91
p-val. 0.00 0.00 0.00
1994
coeff. 0.43 -0.90 0.10 752
t-stat. 2.24 -4.40 5.88
p-val. 0.03 0.00 0.00

Year [R.sup.2]

 Adj [R.sup.2]
1991
coeff. 0.15
t-stat. 0.15
p-val.
1992
coeff. 0.15
t-stat. 0.14
p-val.
1993
coeff. 0.15
t-stat. 0.14
p-val.
1994
coeff. 0.12
t-stat. 0.11
p-val.

YRADOPT = Adoption year (1=early, 0=on-time); EARN = Income
before extraordinary items, divided by total assets; FUND
= accumulated pension benefit obligation minus pension plan
assets, divided by total assets; DE = total debt divided by
total equity;; ABO = accumulated benefit obligation divided
by total assets; FOREIGN = Domestic or Foreign Company;
SIZE = Natural Log of Total Assets

Table 5: Relationship Between Discount Rate and Choice of
Adoption Date of SFAS 87

Model: [DISC.sub.it] = [b.sub.o] + [b.sub.1] [x.sub.it1] +
[b.sub.2] [x.sub.it2] + [b.sub.3] [x.sub.it3] + [b.sub.4]
[x.sub.it4] + [b.sub.5] [x.sub.it5] + [b.sub.6] [x.sub.it6]
+ [b.sub.7] [x.sub.it7] + [e.sub.it]

Year Yradopt Earn Fund De
 [B.sub.1] [B.sub.2] [B.sub.3] [B.sub.4]
1991
coeff. 0.18 0.18 0.00 -0.01
t-stat. 3.97 0.55 -1.67 -0.97
p-val. 0.00 0.58 0.95 0.33
1992
coeff. 0.11 0.05 0.00 0.00
t-stat. 2.53 0.19 -1.61 0.24
p-val. 0.01 0.85 0.11 0.81
1993
coeff. 0.04 0.26 0.00 0.00
t-stat. 1.12 0.99 -0.81 -0.77
p-val. 0.26 0.32 0.42 0.44
1994
coeff. 0.03 -0.29 0.00 0.00
t-stat. 0.74 1.71 -2.18 0.46
p-val. 0.46 0.09 0.03 0.64

Year Abo Foreign Size Sample
 [B.sub.5] [B.sub.6] [B.sub.7] Size
1991
coeff. 0.55 0.20 0.05 724
t-stat. 3.15 1.33 3.65
p-val. 0.00 0.18 0.00
1992
coeff. 0.41 0.19 0.04 756
t-stat. 2.82 1.39 2.81
p-val. 0.01 0.16 0.01
1993
coeff. 0.35 0.43 -0.02 759
t-stat. 3.05 3.83 -1.66
p-val. 0.00 0.00 0.10
1994
coeff. 0.33 -0.49 0.94 749
t-stat. 2.87 -3.82 8.47
p-val. 0.00 0.00 0.00

Year [R.sup.2]
 Adj [R.sup.2]
1991
coeff. 0.062
t-stat. 0.053
p-val.
1992
coeff. 0.036
t-stat. 0.027
p-val.
1993
coeff. 0.040
t-stat. 0.032
p-val.
1994
coeff. 0.112
t-stat. 0.104
p-val.

YRADOPT = Adoption year (1=early, 0=on-time); EARN = Income
before extraordinary items, divided by total assets; FUND
= accumulated pension benefit obligation minus pension
plan assets, divided by total assets; DE = total debt
divided by total equity; ABO = accumulated benefit
obligation divided by total assets; FOREIGN = Domestic
or Foreign Company; SIZE = Natural Log of Total Assets.

Table 6: Relationship Between Estimated Salary Trend Rate and
Choice of Adoption Date of SFAS 87

Model: [SAL.sub.it] = [b.sub.o] + [b.sub.1] [x..sub.it1] + [b.sub.2]
[x.sub.it2] + [b.sub.3] [x.sub.it3] + [b.sub.4] [x.sub.it4] + [b.sub.5]
[x.sub.it5] + [b.sub.6] [x.subit6] + [b.sub.7] [x.sub.it7] + [e.sub.it]

Year Yradopt Earn Fund
 [B.sub.1] [B.sub.2] [B.sub.3]
1991
coeff. 0.07 0.31 0.00
t-stat. 0.96 0.60 -1.21
p-val. 0.34 0.55 0.23
1992
coeff. 0.01 1.24 0.00
t-stat. 0.16 2.99 -0.59
p-val. 0.87 0.00 0.56
1993
coeff. 0.69 1.26 0.00
t-stat. 1.13 2.51 0.48
p-val. 0.26 0.01 0.63
1994
coeff. 0.07 1.41 0.00
t-stat. 1.13 3.83 2.12
p-val. 0.26 0.00 0.03

Year De Abo Foreign
 [B.sub.4] [B.sub.5] [B.sub.6]
1991
coeff. 0.00 -0.45 0.41
t-stat. -0.28 -1.47 1.74
p-val. 0.78 0.14 0.08
1992
coeff. 0.00 -0.48 0.59
t-stat. 0.87 -2.08 2.73
p-val. 0.38 0.04 0.01
1993
coeff. 0.00 -0.59 0.73
t-stat. -0.08 -2.65 3.40
p-val. 0.94 0.01 0.00
1994
coeff. 0.00 -0.58 0.47
t-stat. 0.08 -2.74 1.95
p-val. 0.94 0.01 0.05

Year Size Sample [R.sup.2]
 [B.sub.7] Size Adj [R.sup.2]
1991
coeff. 0.04 668 0.02
t-stat. 1.67 0.00
p-val. 0.10
1992
coeff. 0.01 681 0.03
t-stat. 0.54 0.02
p-val. 0.59
1993
coeff. -0.04 676 0.04
t-stat. -2.31 0.03
p-val. 0.02
1994
coeff. -0.02 655 0.05
t-stat. -1.24 0.04
p-val. 0.22

YRADOPT = Adoption year (1=early, 0=on-time); EARN = Income
before extraordinary items, divided by total assets; FUND
= accumulated pension benefit obligation minus pension
plan assets, divided by total assets; DE = total debt
divided by total equity; ABO = accumulated benefit obligation
divided by total assets; FOREIGN = Domestic or Foreign
Company; SIZE = Natural Log of Total Assets.