Read Changes%20in%20the%20Funded%20Status%20of%20Retirement%20Plans%20after%20the%20Adoption%20of%20SFAS%20No%20%20158-%20Economic%20Improvement%20or%20Balance%20Sheet%20Management.pdf text version

Changes in the Funded Status of Retirement Plans after the Adoption of SFAS No. 158: Economic Improvement or Balance Sheet Management?

DENISE A. JONES, College of William & Mary

May 2011

Please address correspondence to Denise A. Jones, Mason School of Business, The College of William & Mary, P.O. Box 8795, Williamsburg, VA, 23187; 757-221-2876; [email protected]

Changes in the Funded Status of Retirement Plans after the Adoption of SFAS No. 158: Economic Improvement or Balance Sheet Management?

Abstract Statement of Financial Accounting Standards No. 158 (SFAS 158) requires all companies to report the funded status of defined benefit pension plans and other postretirement plans, such as retiree healthcare plans, on the balance sheet for fiscal years ending after December 15, 2006. Prior to this, underfunded retirement plans were often not recorded as a liability on companys balance sheets. At the end of 2005, the average defined benefit pension plan of the companies in this study was underfunded by $324 million and the average other postretirement plan was underfunded by $760 million. After the adoption of SFAS 158 in 2006, this average underfunding decreased. After controlling for changes in market conditions, I find that companies with debt contracting incentives to manage their balance sheets had a larger increase in their defined benefit pension plan funded status subsequent to the adoption of SFAS 158. This increase was at least partially due to the choice of assumptions used to measure the pension obligation. Keywords Defined benefit pension plans, other postretirement plans, pension assumptions, balance sheet management

Changes in the Funded Status of Retirement Plans after the Adoption of SFAS No. 158: Economic Improvement or Balance Sheet Management? 1. Introduction Statement of Financial Accounting Standards No. 158 (SFAS 158) requires all companies to report the funded status of defined benefit pension plans and other postretirement plans (hereafter referred to as pension plans and postretirement plans respectively, and collectively as retirement plans) on the balance sheet for fiscal years ending after December 15, 2006. The funded status is defined as the fair value of the retirement plan assets less the benefit obligation (expected liability). Prior to this, underfunded retirement plans were often not recorded as a liability on companys balance sheets due to accounting rules designed to smooth the effect of changes in interest rates and volatile stock market returns on retirement plan assets and liabilities. In 2005, the average pension plan in this studys sample was underfunded by $324 million, and the average postretirement plan was underfunded by $760 million. Interestingly, after the mandatory adoption of SFAS 158 in 2006, this declined to an underfunding of $117 and $695 million for pension plans and postretirement plans, respectively. In this study, I investigate three related research questions. First, why is there an improvement in the funded status of retirement plans after the adoption of SFAS no. 158? Second, is the improvement in the funded status of retirement plans after the adoption of SFAS 158 associated with managements contracting and other incentives to manage the balance sheet by reducing the recognized retirement liability? Third, did companies with contracting and other incentives manage the assumptions used to estimate the retirement obligation to a greater extent after the adoption of SFAS 158? For a sample of 433 companies with defined benefit pension plans in existence over the period 2003 to 2006, I find that the funded status increased in part due to the overall improvement in the economy leading to large positive returns on pension plan assets. However, the return on pension plan assets in the post-SFAS 158 period is not statistically different from the return in the

1

previous three years. The funded status also increased due to changing actuarial assumptions, plan amendments, and plan freezes. The increase due to both actuarial changes and plan amendments is significantly larger in the post-SFAS 158 period than in the previous three years. In addition, I find that companies with debt contracting incentives to manage their balance sheet, companies with a higher unrecognized liability, and smaller companies had a larger increase in their pension plan funded status after the adoption of SFAS 158. Both changes in the funded status due to changes in actuarial assumptions and plan amendments are associated with debt contracting incentives and the magnitude of the unrecorded liability in the post-SFAS 158 time period. Finally, I directly examine the assumptions selected by managers to measure the retirement obligations. After the adoption of SFAS 158, I find that companies with debt contracting incentives used a higher discount rate and a lower rate of compensation increase, which would reduce the retirement obligation. However, I find no relation between the magnitude of the unrecorded asset or liability and the discount rate or rate of compensation increase. I also examine a subsample of 355 companies with postretirement plans and I find mixed results. Similar to pension plans, I find that after the adoption of SFAS 158, smaller companies and companies with a large underfunding not previously recognized on the balance sheet had a larger increase in the postretirement plan funded status attributable to actuarial changes. However, I find no relation between changes in the funded status and debt contracting incentives. When I directly examine the postretirement assumptions, I find that companies with debt contracting incentives had a larger decrease in their assumed healthcare cost trend rate, which would decrease the benefit obligation and increase the funded status. This paper contributes to the literature in several important ways. First, while there is extensive evidence that managers make accounting choices related to managing earnings (see Fields et al. 2001), there has been very little research on whether managers attempt to manage the balance sheet. One exception is Gramlich et al. 2001, who document that companies reclassify

2

short-term debt to long-term in years when the current ratio before the reclassification would have been lower than the prior year and industry benchmarks. In this study, I examine a situation where a new accounting pronouncement affected only the balance sheet, and provide evidence that accounting choices were related to contracting and other incentives to manage the balance sheet after the adoption of SFAS 158. Second, prior to SFAS 158, even when the retirement plan funded status was not recognized on the balance sheet, it was still disclosed in the footnotes along with other extensive retirement plan disclosures. Although recognition versus disclosure is an important issue, there has been relatively little research in this area, primarily because there are very few circumstances where the same item is both recognized and disclosed. Prior research on two situations where a disclosed item was subsequently recognized, capital leases and employee stock options, has documented that managements actions may be influenced by whether an item is recognized versus disclosed (see, e.g., Johnston 2006; Brown and Lee 2007; Carter et al. 2007; Aboody et al. 2004; Imhoff and Thomas 1988). In this study, I extend this literature by providing evidence that recognition versus disclosure of retirement liabilities may affect managements accounting choices. In addition, the adoption of SFAS 158 takes place in a setting that overcomes a couple of difficulties faced by prior studies. First, when there is a switch from disclosure to recognition, there is often a concurrent change in measurement. 1 The adoption of SFAS 158 provides a unique opportunity where information measured in the exact same way was disclosed one year and recognized the next year. Second, it is often difficult to compare accounting choices across companies because different accounting choices may be due to real economic differences. Retirement plan assumptions are expected to be homogeneous across companies (Amir and Gordon 1996), providing a good setting to examine managements incentives across companies. Finally, one reason that recognition versus disclosure of retirement liabilities may matter to managers is that accounting based debt contract provisions, such as performance pricing

3

provisions, dividend restrictions or debt covenants, might be affected by the recognition of a new retirement plan liability. The empirical evidence on whether accounting choices are motivated by debt contracting concerns is inconclusive (Fields et al. 2001). Managers often have conflicting motivations for making accounting choices, and managers can also take multiple actions to accomplish the same goal (e.g., accrual manipulation, changing accounting methods, etc.), making it difficult to document manipulative behavior (Fields et al. 2001). In this study, I extend the debt contracting literature by documenting managers reaction to a mandatory accounting change that affected a large number of companies, and where the magnitude of the liability is potentially large. The adoption of SFAS 158 affected the balance sheet only, and potentially conflicting earningsrelated motivations, such as maximizing bonuses, did not change. In addition, managers choices to mitigate the impact of this event were limited, making it easier to design a test of the association between managements incentives and accounting choices. The remainder of this paper is organized as follows: section 2 discusses the background of the accounting for retirement plans, trends in retirement plans over time, and incentives to manage retirement plan obligations; section 3 describes the research design and sample selection; section 4 reports the results; and section 5 concludes the paper.

2. Background and literature review Accounting for retirement plans For all fiscal years ending after December 15, 2006, SFAS 158 requires companies to recognize on the balance sheet the funded status of pension plans and postretirement plans, such as retiree health care plans (FASB 2006). The funded status is defined as the fair value of retirement plan assets less the expected liability or projected benefit obligation (PBO) for pension plans and accumulated postretirement benefit obligation (APBO) for postretirement plans. Prior to this, an amount different from the funded status was reported on the balance sheet due to accounting rules

4

designed to smooth fluctuations in the market return on retirement plan assets and changes in actuarial assumptions. For underfunded retirement plans, this delayed recognition of economic events could result in a balance sheet liability that was significantly less than the underfunding of the plan, and in some situations resulted in an asset being reported on the balance sheet. The measurement of retirement related items did not change under SFAS 158. In addition, the measurement of retirement expense on the income statement remains the same and any adjustment necessary to record the funded status on the balance sheet is recorded directly to stockholders equity. The PBO is defined the actuarial present value of future pension benefits attributed to service rendered to date, and should be based on estimated future events such as compensation increases, turnover, and mortality (FASB 1985). The APBO is defined the actuarial present value of future postretirement benefits attributed to service rendered to date, and should be based on the expected future costs of providing the benefits to the employee and other covered dependents, and the extent to which the costs are expected to be shared by the employee (FASB 1990). For the PBO and APBO, management must decide on a discount rate to compute the present value of expected future benefits. The discount rate should be based on the rate at which the retirement benefit could be effectively settled and should reflect the return on high quality fixed-income investments (FASB 1990; FASB 1985).

Incentives to manage the retirement liability A manager concerned about recognizing a large retirement liability on the balance sheet has multiple options to reduce that liability. The manager could take real actions such as reduce retirement benefits, freeze retirement plans, or redirect cash flows to fund the plans (see, e.g., Beaudoin et al. 2010; Campbell et al. 2010; Mittelstaedt et al. 1995). These actions are all costly to the firm either directly by requiring cash or indirectly by hurting the motivation of the workforce. A manager could instead choose the less costly route of changing the assumptions used to measure the retirement liability. This section discusses four incentives that managers have to manage the

5

retirement plan liability: debt contracting incentives, the magnitude of the unrecorded retirement liability, the ability to fund the retirement plan, and the size of the company. There are debt-contracting incentives to minimize the retirement liability recognized on the balance sheet (Watts and Zimmerman 1990). Debt contract provisions can take many forms. Performance pricing provisions make interest rates charged on a bank loan a function of current credit ratings or debt-related financial ratios (see Asquith et al. 2005). In addition, debt contracts often contain dividend restrictions that limit the amount of dividends that can be paid out using a formula typically based on accumulated retained earnings (see Healy and Palepu 1990). Finally, affirmative debt covenants often require a firm to meet certain accounting-based financial ratios, stocks or flows (see Duke and Hunt 1990; Press and Weintrop 1990). Companies with high leverage reduced retiree health care benefits after adopting SFAS 106 (Mittelstaedt et al. 1995) and companies with higher liabilities were more likely to choose prospective adoption of SFAS 106, which postpones the recognition of a liability (DSouza et al. 2001). In addition, previous studies have documented that in the pre-SFAS 158 time period, companies with higher leverage tend to make income-increasing retirement plan related actuarial choices (Asthana 1999; Amir and Gordon 1996; Godwin et al. 1996). Although the opportunistic choice of retirement assumptions has been previously documented, I expect the period after the adoption of SFAS 158 to be different. In the pre-SFAS 158 time period, management had incentives to use the retirement assumptions as an earnings management tool. Indeed, some of the strongest evidence in the pre-SFAS 158 period is related to the expected rate of return assumption (see Bergstresser et al. 2006; Comprix and Muller 2006), which affects pension expense but not the PBO. In addition, the strongest incentive for early adoption of SFAS 87 was the ability to increase reported earnings on the income statement (see Tung and Weygandt 1994; Langer and Lev 1993; Scott 1991). SFAS 158 changed the reporting of retirement amounts on the balance sheet but did not change how retirement expense is reported on

6

the income statement, leading to incentives to manage the potential new balance sheet liability that are unrelated to the prior profitability based incentives. Even small changes to the balance sheet can make a difference to managers as debt covenants are set tightly and violations are common (Dichev and Skinner 2002). 2 In addition, financial ratios are often used by debt rating agencies in determining credit ratings, which may affect performance pricing provisions or future debt contract negotiations. In summary, the adoption of SFAS 158 has the potential to have a large impact on the ratios commonly used in debt contracting if previously unrecorded retirement liabilities are recognized on the balance sheet. This leads to the following hypotheses: HYPOTHESIS 1a. After the adoption of SFAS 158, companies with debt contracting incentives will have a larger increase in the retirement plan funded status. HYPOTHESIS 1b. After the adoption of SFAS 158, companies with debt contracting incentives will make larger PBO or APBO decreasing changes to the assumptions used to estimate the retirement plan obligations. The second incentive relates to the magnitude of the unrecorded retirement liability. Larger retirement obligations are more likely to make a material difference to the financial statements. Asthana (1999) documents that the more underfunded a firms pension plan is, the more likely managers are to make PBO decreasing actuarial pension choices. Similarly, companies with large postretirement obligations reduced retiree health care benefits after adopting SFAS 106 (Mittelstaedt et al. 1995) and tend to select more obligation decreasing postretirement assumptions (Amir and Gordon 1996). As discussed above, SFAS 158 requires recognition of the retirement plan funded status on the balance sheet, whereas previously this information had been partly disclosed and partly recognized. Prior research has documented that the capital market assigns a greater weight to recognized items than disclosed items (see, e.g., Picconi 2006; Davis-Friday et al. 2004; Chen and Schoderbek 2000; Davis-Friday et al. 1999; Aboody 1996).3 Because of this and for other

7

contracting reasons, managers perceive recognized amounts to be different from disclosed amounts and this affects their actions (see, e.g., Bens and Monahan 2008; Brown and Lee 2007; Carter et al. 2007; Johnston 2006; Aboody et al. 2004; Imhoff and Thomas 1988). Therefore, I expect managements incentives to be different after adoption of SFAS 158, when previously disclosed retirement obligations became recognized on the balance sheet and the larger the previously unrecorded amounts the stronger the incentives to minimize their impact. This leads to the following hypotheses: HYPOTHESIS 2a. After the adoption of SFAS 158, companies with a larger unrecorded retirement liability will have a larger increase in the retirement plan funded status. HYPOTHESIS 2b. After the adoption of SFAS 158, companies with a larger unrecorded retirement liability will make larger PBO or APBO decreasing changes to the assumptions used to estimate the retirement plan obligations. Managements incentives are also related to the ability to fund the retirement plan through either operating profits or returns on existing retirement assets. Companies whose retirement plan assets are earning a poor return on investment and that are unable to fund the retirement plan due to weak operating cash flows have incentives to manage the actuarial assumptions. Asthana (1999) finds that companies with lower cash from operations tend to make PBO decreasing actuarial pension choices. Similarly, Healy and Palepu (1990) find that firms adopt income increasing (PBO decreasing) pension accounting assumptions during the four years surrounding a near violation of a dividend covenant, but attribute this to changes in operating cash flows, not the dividend covenant. Finally, Eaton and Nofsinger (2004) document that government sponsors of public pension plans facing financial constraints are more likely to make obligation decreasing retirement assumptions. As discussed above, the incentives to reduce the retirement obligation are expected to be stronger after the adoption of SFAS 158 because previously disclosed obligations will be recognized on the balance sheet potentially triggering debt contract provisions. Companies

8

without the ability to fund the retirement plan through operating cash flows or returns on existing assets have a stronger incentive to minimize the liability through changing the actuarial assumptions. This leads to the following hypothesis: HYPOTHESIS 3. After the adoption of SFAS 158, companies with less ability to fund the retirement plan will make greater PBO or APBO decreasing changes to the assumptions used to estimate the retirement plan obligations. It is unclear what the relation between the ability to fund the retirement plan and the change in the funded status will be. If hypothesis 3 is correct then there would be a negative relation between the ability to fund the plan and the funded status, due to managing the assumptions to decrease the obligation. However, there may also be a positive relation because companies with strong operating cash flows have the resources to fund the retirement plan and strong returns on plan assets also increase the funded status. Therefore, I make no predictions about the relation between the ability to fund the retirement plan and the change in the funded status. Finally, Watts and Zimmerman (1990) propose that size is a proxy for political attention and, therefore, larger companies are more likely to make accounting choices that reduce profits. In recent years some larger companies, such as Ford and General Motors, have come under SEC scrutiny for possibly changing their retirement plan assumptions to make the retirement plans look healthier (Solomon and Hawkins 2005; Schultz 2004). Therefore, managers of larger companies have incentives to make PBO/APBO increasing (net income decreasing) retirement assumptions to avoid political scrutiny. In addition, smaller companies might be at greater risk of having their debt ratings downgraded if they were to recognize a liability. Regulators are sometimes less stringent with larger companies (OHara and Shaw 1990) and DSouza et al. (2001) find that larger companies were more likely to elect immediate recognition of a liability upon adoption of SFAS 106. This leads to the following hypotheses:

9

HYPOTHESIS 4a. After the adoption of SFAS 158, smaller companies will have a larger increase in the retirement plan funded status. HYPOTHESIS 4b. After the adoption of SFAS 158, smaller companies will make larger PBO or APBO decreasing changes to the assumptions used to estimate the retirement plan obligations. 3. Research Design Empirical models and variable definitions This study investigates three related research questions. First, why is there an improvement in the funded status of retirement plans after the adoption of SFAS no. 158? Second, is the improvement in the funded status of retirement plans after the adoption of SFAS 158 associated with managements contracting and other incentives to manage the balance sheet by reducing the recognized retirement liability? Third, did companies with contracting and other incentives manage the assumptions used to estimate the retirement obligation to a greater extent after the adoption of SFAS 158? To answer the first research question, I collect data from the sample firms retirement plan footnote on the reasons why the funded status changed. The empirical models used to investigate the second and third research question are discussed below.

Changes in the funded status of defined benefit pension plans To answer the second research question, I begin with a model explaining the changes in the pension plans funded status: FS_PENit = 0 + 1MKTRETit + 2DEBTit-1 + 3NWit + 4UNREC_PENit-1 + 5CFOit + 6SIZEit + 7EMPit + 8DURATIONit +9FIRMAGEit + 10FREEZE_PENit + 11FAS158 + 12DEBTit-1*FAS158 + 13NWit*FAS158 + 14UNREC_PENit-1*FAS158 + 15CFOit*FAS158 + 16SIZEit*FAS158 + 17 j IND j +

j 0 7

y 0

2

24 y

YEARy + it

(1)

The dependent variable (FS_PENit) is the change in the pension plan funded status deflated by total assets, where the funded status is defined as pension plan assets less the PBO. 4 As discussed in section 2, there are several reasons why the funded status of the pension plan can

10

change, and in their footnote disclosures companies typically break down the change in the PBO into several standard categories. In addition to the total change in the funded status (FS_PEN_TOTit), I also examine the change in the funded status due to changes in all actuarial assumptions (FS_PEN_ACTit) and due to plan amendments (FS_PEN_AMDit). I estimate model (1) three separate times for each of the dependent variables. 5 Each dependent variable is defined such that a decrease in the PBO, which increases the funded status, is a positive number. This information was hand-collected from the retirement plan footnote for all companies and all years. All data requiring hand collection are noted in the variable description below; the remainder of the data are obtained from Compustat. The first independent variable (MKTRETit) controls for the change in the funded status due to changing market conditions. Retirement plans typically invest in a mix of stocks and fixed income investments. As the stock market grows or declines and fixed income investment yields change, retirement plans realize gains or losses on the retirement plan assets. MKTRETit is defined as the return on large company stocks (from Ibbotsons Stocks, Bills, Bonds, and Inflation) weighted by the percentage of pension plan assets held in equity investments plus the yield on high quality corporate bonds (average of the yield on corporate bonds rated Aaa and Baa by Moodys as reported in the monthly Federal Reserve statistical release) weighted by the percentage of pension plan assets held in debt and other investments. The percentage of pension plan assets held in equity and debt investments is from Compustat and hand-collected if missing. The next two variables (DEBTit-1 and NWit) are proxies for the existence of debt contracting incentives.6 Performance pricing provisions are often based on debt-related financial ratios or on credit ratings, which are affected by debt-related financial ratios (Asquith et al. 2005). In addition, Duke and Hunt (1990) and Press and Weintrop (1990) find that leverage ratios capture the existence of common debt covenant restrictions. 7 I use the ratio of total liabilities to total assets (DEBTit-1) as a measure of leverage.8 Beginning of the year leverage is used to capture the level of

11

leverage before any changes to the pension liability recorded that year. In addition to a leverage variable, I also include a proxy for one of the most commonly violated debt covenants--net worth. Press and Weintrop (1990) find that 46% of their sample companies had net worth restrictions, and Sweeney (1994) finds that net worth covenants are the most commonly violated debtcovenants (60% of sample companies). Following Sweeney (1994), I use the change in net worth (deflated by total assets) as a proxy for changes in the tightness of a firms net worth debtcovenants (NWit). The magnitude of any unrecognized pension assets or liabilities is captured by UNREC_PENit-1, which is the beginning of the year pension plan funded status less the amount recognized on the balance sheet, deflated by total assets. Similar to the leverage variable, I use the beginning of the year funded status to capture the extent of any unrecognized amounts before changes to the balance sheet that year. As UNREC_PENit-1 would be negative for companies with unrecognized liabilities (and positive for unrecognized assets), I expect a negative relation between UNREC_PENit-1 and the change in the funded status. As a proxy for the ability to fund the pension obligation out of operating income, I use the average change in cash from operations over the past three years, deflated by total assets (CFOit). Finally, firm size (SIZEit) is defined as the log of the market value of equity. I include several control variables (EMPit, DURATIONit, FIRMAGEit, FREEZE_PENit, INDj, and YEARy). The change in the number of employees deflated by total assets (EMPit) controls for changes in the growth or decline in the pension plan due to events like expansion or restructuring. Managers of older companies and companies with pension plans with a shorter duration might have different incentives. For example, for pension plans with a shorter duration, the interest cost component of pension expense is often larger than the service cost component and managers may find it beneficial to decrease the discount rate. I control for both duration and the age of the firm. DURATIONit is defined as the accumulated benefit obligation divided by the

12

projected benefit obligation. Both the accumulated benefit obligation and the projected benefit obligation are from Compustat and are hand-collected if missing. FIRMAGEit is defined as the number of years that Compustat has reported information on the company. There has been an increase in the number of companies freezing their pension benefits in recent years and companies with a higher unfunded pension liability prior to the adoption SFAS 158 in 2006 are more likely to freeze their pension benefits (Beaudoin et al. 2010; Campbell et al. 2010). I control for pension benefit freezes with FREEZE_PENit, which is an indicator variable that is 1 if there was a hard freeze (complete freeze of all future benefits), soft freeze (freeze of some but not all future benefits), or closed freeze (plan closed to new employees) during the year, and 0 otherwise. Benefit freezes were identified by reading the retirement footnote disclosures for each sample company each year. Finally, included in each regression, but not reported in the tables, are dummy variables to control for industry differences at the one-digit SIC code level (INDj) as well as year dummy variables to control for year-specific effects (YEARy). To test whether the post-SFAS 158 time period is different, I interact each of the incentive-related variables (DEBTit-1, NWit, UNREC_PENit-1, CFOit, and SIZEit) by a dummy variable that is 1 after the adoption of SFAS 158, and 0 otherwise (FAS158). I estimate model (1) with and without the interaction terms.

Changes in the funded status of other postretirement plans For postretirement plans, I estimate the following models: FS_OPEBit = 0 + 1MKTRETt + 2MEDACTit + 3DEBTit-1 + 4NWit + 5UNREC_OPEBit-1 + 6CFOit + 7SIZEit + 8EMPit + 9FIRMAGEit + 10FREEZE_OPEBit + 11FAS158 + 12DEBTit-1*FAS158 + 13NWit*FAS158 + 14UNREC_OPEBit-1*FAS158 + 15CFOit*FAS158 + 16SIZEit*FAS158 + 17 j IND j +

j 0 5

j 0

2

22 y

YEARy + it

(2)

This model is very similar to model (1) with a couple of differences. FS_OPEBit is the change in the funded status of the postretirement plan, deflated by total assets and

13

UNREC_OPEBit-1 is the beginning of the year postretirement plan funded status less the amount recognized on the balance sheet. Similar to pension plans, I examine the total change in the funded status (FS_OPEB_TOTit), the change in the funded status due to changes in actuarial assumptions (FS_OPEB_ACTit), and the change in the funded status due to plan amendments (FS_OPEB_AMDit). Similar to model (1), I estimate model (2) three separate times for each of the dependent variables, as well as with and without interaction terms. In addition, I include the variable MEDACTit to capture the effect of the Medicare, Prescription Drug, Improvement, and Modernization Act of 2003. For most companies this Act reduced their retiree health care costs and therefore reduced the APBO. Finally, the duration variable is excluded from the model as it relates to the duration of pension plans only. All data specific to postretirement plans were handcollected from the retirement plan footnote. 9

Retirement assumptions To answer the third research question, I directly investigate the assumptions made by management when measuring the pension PBO and postretirement APBO. Three commonly disclosed assumptions are the discount rate, the rate of compensation increase, and the initial and ultimate healthcare cost trend rate. Note that the change in the PBO and APBO due to changes in actuarial assumptions includes these assumptions as well as other assumptions that are not disclosed such as employee turnover, mortality rates, years to retirement, etc. In addition, there is very little variation in the ultimate healthcare cost trend rate (for most companies it is 5 percent) and the following analyses focus on the initial trend rate. I estimate the following models: DISCit = 0 + 1YIELDt + 2DEBTit-1 + 3NWit + 4UNREC_PENit-1 + 5CFOit + 6ASSETS_PENit + 7SIZEit + 8EMPit + 9DURATIONit + 10FIRMAGEit + 11FREEZE_PENit + 12FAS158 + 13DEBTit-1*FAS158 + 14NWit*FAS158 + 15UNREC_PENit-1*FAS158 + 16CFOit*FAS158 + 17ASSETS_PENit*FAS158 + 18SIZEit*FAS158 +

j 0

7

19 j

IND j +

y 0

2

26 y

YEARy + it

(3)

14

RCIit = 0 + 1DEBTit-1 + 2NWit + 3UNREC_PENit-1 + 4CFOit + 5ASSETS_PENit + 6SIZEit + 7EMPit + 8DURATIONit + 9FIRMAGEit + 10FREEZE_PENit + 11FAS158 + 12DEBTit-1*FAS158 + 13NWit*FAS158 + 14UNREC_PENit-1*FAS158 + 15CFOit*FAS158 + 16ASSETS_PENit*FAS158 + 17SIZEit*FAS158+ +

j 0

7

18 j

IND j

(4)

y 0

2

25 y

YEARy + it

HTRit = 0 + 1DEBTit-1 + 2NWit + 3UNREC_OPEBit-1 + 4CFOit + 5ASSETS_OPEBit + 6SIZEit + 7EMPit + 8FIRMAGEit + 9FREEZE_OPEBit + 10FAS158 + 11DEBTit-1*FAS158 + 12NWit*FAS158 + 13UNREC_OPEBit-1*FAS158 + 14CFOit*FAS158 + 15ASSETS_OPEBit*FAS158 + 16SIZEit*FAS158 + +

j 0

5

17 j

IND j

(5)

y 0

2

22 y

YEARy + it

The dependent variable in model (3) is the industry adjusted change in the discount rate used to estimate the PBO (DISCit). The dependent variable in model (4) is the industry adjusted change in the rate of compensation increase used to estimate the PBO (RCIit). The dependent variable in model (5) is the industry adjusted change in the initial healthcare cost trend rate used to estimate the APBO (HTRit). The healthcare cost trend rate is hand-collected from the retirement plan footnote, and both the discount rate and the rate of compensation increase are obtained from Compustat and checked against the retirement plan footnote.10 The retirement plan assumptions should be relatively homogeneous across companies in the same industry (Hann et al. 2007; Amir and Gordon 1996). In addition, companies in the same industry face similar pressures to attract and retain employees by providing compensation levels commiserate with market rates of employee compensation. Following Hann et al. (2007), I derive the discretionary component of the pension assumptions as the deviation from the industry median. The independent variables are the same variables used to estimate models (1) and (2) with two exceptions. First, for model (3) a variable is added to control for current market discount rates. As discussed in section 2, the SEC requires companies to use the yield on high quality corporate bonds as the discount rate for calculating the PBO and APBO. YIELDt is defined as the

15

yield on high quality corporate bonds (average of the yield on corporate bonds rated Aaa and Baa by Moodys as reported in the monthly Federal Reserve statistical release). Second, I added an additional variable to capture the ability to fund the pension plan. When the retirement plan assets earn a higher return, the need to fund the plan is lower and therefore the incentive to manage the PBO and APBO is lower. I define the variable ASSETS_PEN (OPEB)it as the percentage change in the pension (postretirement) plan assets.

Sample selection and descriptive statistics To select the final sample of companies, I begin with all companies incorporated in either the United States or Canada for which Compustat reports the necessary pension and other data required to calculate the variables over the period 2003 to 2006, except for the variables noted in the previous section as hand-collected. The sample period begins with 2003 as this is when SFAS No. 132 (R), which specified the disclosure of some of the required retirement data, became effective (FASB 2003). Because I am comparing the period pre and post-SFAS 158, I exclude companies that do not have the necessary data for the entire 2003 to 2006 time period. Otherwise, the results could be related to the sample having different characteristics in the pre and post-SFAS 158 periods. For example, if companies who terminate their pension plan or companies who are acquired by other companies tend to have fully funded plans, then this may cause the post-SFAS 158 time period to look worse than it really is. It may also cause the interaction variables to be significant in the post-SFAS158 period because the sample changed over time, not because managements incentives were different. Managers of companies for which the retirement obligation is an insignificant portion of their total liabilities have different incentives and might behave differently than managers of companies where the retirement plan is a material part of the balance sheet. Therefore, I exclude from the sample all companies where the combined PBO and APBO (as reported by Compustat) is less than ten percent of total liabilities. This results in a final sample of 433 companies, 355 of

16

which have postretirement plans. In each regression, outliers were removed by excluding the top and bottom 1% of all variables. Table 1 provides some descriptive information about the sample companies and the regression variables. Any data hand-collected from the retirement plan footnotes are noted above in the variable definition; all other data are from Compustat. Table 1, panel A reports information for the entire sample of companies with pension plans. The average sample firm is large with a mean market value of equity of $7,484 million. The average firm has also been in business a long time--35 years. This is expected as it is the companies in older, established industries that tend to have defined benefit pension plans. Many newer companies offer defined contribution plans. The mean duration of 91% is also very high. This reflects the growing number of retirement plans where the benefits have been frozen in recent years. In addition, the average firm reports increasing net worth and debt levels that are 64% of total assets. Finally, retirement obligations are a material part of the sample companies balance sheets, with the retirement obligation as a percentage of total liabilities ranging from 17.1% for the 10 th percentile to 93.4% for the 90th percentile. Table 1, panel B reports information on the postretirement plan variables for the subsample of companies with postretirement plans. This sub-sample is also large, with a mean market value of equity of $8,984 million. The change in operating cash flows, change in net worth, and leverage ratios (untabulated) are similar to the full sample. [Insert Table 1]

4. Results Descriptive information about the funded status and retirement assumptions over time Table 2, panel A reports the average funded status over the period 2003 to 2006 for the sample of 433 companies with defined benefit pension plans. As expected, pension plans were underfunded during the entire sample time period, which follows a sharp decline in the stock

17

market from 2000 to 2002. The funded status remained at a pretty constant underfunding of $324 to $343 million from 2003 to 2005. In 2006, there was a $207 million increase in the average funded status, which is significantly larger than the average change in the previous three years. This is partly due to a 15.8% average stock return leading to a return on pension plan assets of $289 in 2006. Although this explains part of the increase in the funded status in the post-SFAS 158 time period, the return is not statistically different from the return in the previous three years. In addition, the funded status increased due to contributions made by the firm and decreased due to the accrual of additional service cost and interest cost, all of which are not statistically different in the pre and post-SFAS 158 time periods. The change in funded status could also be related to management incentives to improve the funded status and avoid recording a pension liability under SFAS 158. During the 2003 to 2005 time period the funded status decreased on average by $125 to $128 million due to changes in actuarial assumptions. In contrast, in 2006, there was an increase in the funded status of $44 million on average due to changes in actuarial assumptions, which is significantly larger than the previous three years.11 Finally, the funded status improved due to both plan amendments and plan freezes, although only the change due to plan amendments is statistically significant. [Insert Table 2]

Table 2, panel B reports the average funded status over the period 2003 to 2006 for the sub-sample of 355 companies with postretirement plans (as well as pension plans). The Employee Retirement Income Security Act of 1974 (ERISA) established minimum funding requirements for pension plans, which is why all companies have a substantial amount of pension plan assets. In contrast, only a small portion of the postretirement plans are funded. Similar to pension plans, there was an improvement in the funded status in 2006 (increase of $62 million) and at least a portion of this improvement is related to a $41 million increase in the funded status on average due to changes in actuarial assumptions. In addition, many postretirement plans were amended

18

over the period 2003 to 2006 to reduce benefits and cap rising healthcare costs. This is reflected in the increase in the funded status each year due to amendments, with the largest average increase of $68 million in 2006. Table 3 reports the distribution of discount rates (panel A), rates of compensation increase (panel B), and the initial health care cost trend rate (panel C) over the period 2003 to 2006 for all sample companies. In 2006, the median discount rate increased from 5.65% to 5.86% and the discount rate also increased across all percentiles. The median rate of compensation increase has remained steady at 4% over the sample time-frame. The rate of compensation increase in all percentiles (10th, 25th, 75th, and 90th) decreased over the period 2003 to 2005 and then increased in 2006. Finally, the median healthcare cost trend rate has declined over time from 10% in 2003 to 9% in 2006, and has also declined across all percentiles. [Insert Table 3]

Table 4 reports the distribution of the mean changes in the assumptions across companies for the pre-SFAS 158 period (2003-2005) and the post-SFAS 158 period (2006). The sample is broken down into companies with low, medium, and high levels of debt (panel A), unrecorded retirement assets or liabilities (panel B), actual return on retirement plan assets (panel C), and small, medium, and large companies (panel D). Because many postretirement plans are unfunded, there is not much distribution in the return on postretirement plans and nothing is presented for the change in the initial healthcare trend rate in panel C. In 2006, the discount rate increased on average and this increase was higher for companies with high levels of debt (0.21 vs. 0.18) and companies with the smallest return on pension plan assets (0.22 vs. 0.17). Although the magnitude of the increase in the discount rate is small, it can still have a large impact on the PBO. In addition, the change in the initial healthcare cost trend decreased on average in both the period before and after SFAS 158, and the decrease was the largest for companies with high levels of debt in 2006 (0.41) and highest unrecorded liabilities (-0.36; low category in panel B).

19

[Insert Table 4]

Regression results for changes in the funded status of defined benefit pension plans The results from estimating model (1) for all sample companies with pension plans are reported in Table 5. Columns (1) and (2) report the results for the total change in the funded status ((FS_PEN_TOTit). The first column reports the results with no interaction terms and the second column reports the results of the expanded model that includes interaction terms to test whether the post-SFAS 158 time period is different. The results for the change in the funded status due to changes in actuarial assumptions (FS_PEN_ACTit) and plan amendments (FS_PEN_AMDit) are reported in columns (3) and (4), and (5) and (6), respectively. [Insert Table 5]

Columns (1) and (2) report a positive and significant relation between the total change in the pension plan funded status and the market return (MKTRETit). At least a portion of the change in the funded status can be explained by changes in market conditions. The variable controlling for expected growth or decline in the size of the retirement plan (EMPit) is negatively related to FS_PEN_TOTit (p<0.01). As the number of employees decreases, the funded status of the pension plan improves, perhaps because the retirement obligation is lower. Finally, there is a positive relation between FS_PEN_TOTit and DURATIONit (p<0.01). This relation is not surprising as many companies have a high duration due to freezing or reducing plan benefits and you would expect to see a corresponding increase to the pension plan funded status. Contrary to expectations, in the pre-SFAS 158 sample period there is a negative relation between FS_PEN_TOTit and DEBTit-1 (p<0.05) and a positive relation between FS_PEN_TOTit and NWit (p<0.01). This is consistent with companies having better financial prospects, as seen by lower debt and higher net worth, increasing funding to the pension plan. However, if the reasons for the change in the funded status are considered, it is apparent that this is not the only

20

explanation. The results in columns (3) and (4) show that the negative relation with DEBTit-1 and the positive relation with NWit are present when the dependent variable is the change in the funded status due to changes in actuarial assumptions (FS_PEN_ACTit). In the pre-SFAS 158 time-period when the entire retirement obligation was not on the balance sheet, other incentives may have taken precedence over debt contracting incentives. For example, companies with high rates of collective bargaining use more financial leverage to improve union negotiating positions (Matsa 2010) and DSouza et al. (2001) find that more unionized companies were more likely to adopt SFAS 106 immediately as the larger retirement obligation gave the company bargaining power in future labor negotiations. The negative relation between DEBTit-1 and the change in the funded status in the pre-SFAS 158 time period is consistent with the findings in these studies. As reported in columns (2), (4), and (6), the coefficient on the DEBTit-1*FAS158 interaction variable is positive and significant when the dependent variable is total changes in the funded status (p<0.01), actuarial changes (p<0.01), and changes due to amendments (p<0.05). After the adoption of SFAS 158, companies with high levels of debt at the beginning of the year are more likely to have an increase in the pension plan funded status due to both changes in actuarial assumptions and amendments to the pension plan. This supports hypothesis 1a; companies with debt contracting incentives had a larger improvement in the funded status of the pension plan after they were required to move pension assets and liabilities from the footnotes to the balance sheet. In addition, supporting hypothesis 2a, the coefficient on the UNREC_PENit1*FAS158

interaction variable is negative and significant (p<0.01) when the dependent variable is

total changes in the funded status, actuarial changes, and changes due to amendments. This variable is negative if there is an unrecorded pension liability, and the negative coefficient indicates that companies with a higher unrecorded pension liability tend to have increases in the pension funded status in the following year due to both changes in actuarial assumptions and plan amendments. Finally, the negative coefficient on the SIZEit*FAS158 interaction variable (p<0.01)

21

supports hypothesis 4a. Smaller companies had a larger increase in the total pension funded status and the portion of the change in the funded status due to actuarial assumptions.

Regression results for changes in the funded status of other postretirement plans The results from estimating model (2) for the subsample of companies with postretirement plans are reported in Table 6. Columns (1) and (2) report the results for the total change in the funded status ((FS_OPEB_TOTit). The first column reports the results with no interaction terms and the second column reports the results of the expanded model that includes interaction terms to test whether the post-SFAS 158 time period is different. The results for the change in the funded status due to changes in actuarial assumptions (FS_OPEB_ACTit) and plan amendments (FS_PEN_AMDit) are reported in columns (3) and (4), and (5) and (6), respectively. [Insert Table 6]

As expected, there is a positive and significant relation between the variable controlling for the effect of the Medicare Prescription Drug, Improvement and Modernization Act (MEDACTit) and both the total change in the postretirement plan funded status (FS_OPEB_TOTit) and the portion of the change due to changes in actuarial assumptions (FS_OPEB_ACTit). Similar to the results with pension plans, the change in the number of employees (EMPit) is negatively related to FS_OPEB_TOTit (p<0.01). In contrast to the results with pension plans, the relation between the change in the postretirement plan funded status and market returns (MKTRETit) is not statistically significant. This is not surprising as many of these plans are unfunded. For the entire sample period, columns (1) and (2) report a positive relation between DEBTit-1 and FS_OPEB_TOTit (p<0.01) and a negative relation between UNREC_OPEBit-1 and FS_OPEB_TOTit (p<0.01). It can be seen from columns (5) and (6) that this relation is due to changes in the funded status due to plan amendments, which is consistent with the findings of

22

Mittelstaedt et al. (1995). Similar to the results for pension plans, columns (3) and (4) report a negative relation between (FS_OPEB_ACTit) and DEBTit-1, and a positive relation with UNREC_OPEBit-1, indicating that incentives other than debt contracting incentives took precedence in the pre-SFAS 158 time period. As reported in columns (2) and (4), the coefficient on the UNREC_OPEBit-1*FAS158 interaction variable is negative and significant when the dependent variable is total changes in the funded status (p<0.01) and actuarial changes (p<0.01). After the adoption of SFAS 158, companies with unrecorded postretirement liabilities are more likely to have an increase in the postretirement plan funded status due to changes in actuarial assumptions. This supports hypothesis 2a. In addition, supporting hypothesis 4a, after adoption of SFAS 158 there is a negative relation between company size and the change in the funded status due to actuarial assumptions. Finally, with respect to postretirement plans, there is no support for hypothesis 1a. In the post-SFAS 158 time period, there is a positive relation between NWit and FS_OPEB_TOTit (p<0.01). This does not support the debt contracting hypothesis, which would predict a negative sign on the coefficient. However, it is consistent with companies in better financial condition, allocating more resources to fund the postretirement plan.

Regression results for changes in the retirement plan assumptions Table 7, columns (1) and (2) present the results from estimating model (3), where the dependent variable is the industry adjusted change in the discount rate (DISCit). The first column reports the results with no interaction terms and the second column reports the results of the expanded model that includes interaction terms to test whether the post-SFAS 158 time period is different. Columns (3) and (4) of table 7 report the results from estimating model (4), where the dependent variable is the industry adjusted change in the rate of compensation increase (RCIit), and columns (5) and (6) report the results from estimating model (5), where the dependent variable is the industry adjusted change in the initial healthcare cost trend rate (HTRit). Models (3) and (4)

23

are estimated using the entire sample of companies with pension plans and model (5) is estimated using the subsample of companies with retirement plans. [Insert Table 7]

Columns (1) and (2) show that the coefficient on YIELDit is positive and significant (p<0.05). This is consistent with discount rates changing each year to reflect the yield on high quality corporate bonds as mandated by the SEC. In the pre-SFAS 158 time period, all variables capturing managers incentives to manage the balance sheet are not significantly different from zero, which is not surprising as a good portion of the retirement obligation was not recorded on the balance sheet during this time period. The coefficient on the DEBTit-1*FAS158 interaction term is positive and significant (p<0.05). This suggests that after the adoption of SFAS-158 companies with debt contracting incentives were more likely to increase the discount rate, which would decrease the PBO and increase the funded status. In addition, there is a negative and significant relation between the ASSETS_PENit *FAS158 interaction term and the industry adjusted change in the discount rate. After the adoption of SFAS 158, managers of pension plans with assets realizing poor returns are more likely to increase the discount rate. This is a particularly interesting result given the recent stock market losses and the resultant steep decline in pension plan assets. Many companies have recently requested from Congress that they be allowed to delay making their required pension contributions (see Walsh 2008). The findings of this study indicate that these same companies have incentives to minimize any recognized liability. As can be seen in columns (3) and (4), RCIit is positively associated with the change in operating cash flows (CFOit) for the entire sample period (p<0.05). This is not surprising as these companies have less incentive to manage their balance sheet as they have a greater ability to fund the retirement plans out of operations. Alternatively, this might not be related to balance sheet management but might simply reflect the fact that companies that are performing better are more likely to compensate their employees better. For the entire sample period, there is a negative

24

relation between RCIit and the duration of the pension plan (p<0.05). A high duration can be due to having an older workforce and can also be due to freezing pension plans (which causes the ABO to equal the PBO in the case of a hard freeze). Taken together, the negative relation between RCIit and DURATIONit along with the positive relation between RCIit and CFOit is consistent with lower salary increases being given by struggling companies that have lower operating cash flows and have frozen the pension plan or reduced benefits in the past. The coefficient on the DEBTit-1*FAS158 interaction term is negative and significant (p<0.05). This suggests that after the adoption of SFAS-158 companies with debt contracting incentives were more likely to decrease the rate of compensation increase, which would decrease the PBO and increase the funded status. When the industry adjusted change in the initial healthcare trend rate (HTRit) is the dependent variable (columns (5) and (6)), the coefficient on UNREC_OPEBit is positive and significant for the entire sample period. Companies with a higher level of unrecorded liabilities (when UNREC_OPEBit is low) have a smaller increase or higher decrease in the assumed healthcare trend rate. In addition, the coefficient on the DEBTit-1*FAS158 interaction term is negative and significant (p<0.05). This suggests that after the adoption of SFAS-158 companies with debt contracting incentives were more likely to decrease the healthcare trend rate, which would decrease the PBO and increase the funded status. Taken as a whole, the results support hypothesis 1b. In the post-SFAS 158 period, companies with higher leverage made PBO decreasing changes to all disclosed retirement assumptions. Conversely, there is no support for hypothesis 2b and limited support for hypothesis 3. Finally, the positive relation between firm size and the change in the discount rate in the postSFAS 158 time period is opposite what is predicted by hypothesis 4.

5. Conclusion Beginning in 2006, SFAS 158 required the funded status of pension plans and

25

postretirement plans to be recognized on the balance sheet. In the time period leading up to SFAS 158, there were several years of stock market declines, which lead to large, realized losses on pension plan assets. In many cases, this resulted in underfunded pension plans with no corresponding liability recognized on the balance sheet, primarily due to accounting rules designed to smooth the effect of changes in interest rates and market stock returns. Interestingly, in the same year that SFAS 158 was first in effect, there was a sharp increase in the funded status of most retirement plans, giving rise to the question of whether this improvement was due to an overall economic recovery or whether managers opportunistically selected retirement assumptions to minimize the liability recognized on the balance sheet. After controlling for changes in market conditions, the change in the number of employees, the duration of the pension plan, the age of the firm, and the existence of freezes of retirement benefits, I find that companies with debt contracting incentives to manage their balance sheet had a larger improvement in their pension plan funded status after the adoption of SFAS 158. This finding is supported when I directly examine both the net change in the funded status due to actuarial assumptions as well as the industry adjusted changes in both the separately disclosed discount rate and rate of compensation increase assumptions. This paper contributes to the literature in several important ways. First, although there is extensive evidence on earnings management, there are very few studies investigating whether managers attempt to manage the balance sheet. Second, prior to SFAS 158, information was required to be disclosed about retirement plans, including the funded status. SFAS 158 mandated a switch from disclosure of this information to recognition on the balance sheet with no concurrent change in measurement. This study utilizes this unusual setting to provide evidence on the debate about recognition versus disclosure. Third, accounting based debt contract provisions might be affected by the recognition of a new retirement plan liability. Prior research in this area is inconclusive (Fields et al. 2001) and the adoption of SFAS 158 provides a unique opportunity to

26

examine a situation where only the balance sheet was affected and managers choices to mitigate the impact of this event were limited. One limitation of the study is that I find inconclusive evidence related to hypotheses about the magnitude of the unrecorded liability and size of the firm. More specifically, as expected, I find that there was a larger increase in both the pension plan and postretirement plan funded status for smaller companies and companies with larger unrecorded retirement obligations. However, I find no evidence that the change in the assumptions used to estimate the obligations is different for these companies. This might be because the assumptions that are not disclosed (such as the mortality rate and the expected years to retirement), and therefore that receive less public scrutiny are managed. However, this cannot be verified.

27

Endnotes 1. For example, Statement of Financial Accounting Standards No. 123R (SFAS 123R) amends Statement of Financial Accounting Standards 123 (SFAS 123) to require that employee compensation in the form of stock options be recognized as an expense on the income statement. Under SFAS 123, companies could chose to disclose the employee stock option expense instead of recognizing it. In addition to requiring recognition, SFAS 123R also clarified and expanded guidance on how the stock option expense is to be measured. At the same time, the Securities and Exchange Commission issued Staff Accounting Bulletin 107, which provided guidance on the measurement of the stock option expense. So, potential changes in measurement occurred simultaneously with the mandating of recognition, making it difficult to disentangle measurement from financial statement location. 2. Debt covenants are sometimes based on "frozen" generally accepted accounting principles (GAAP) or GAAP in existence at the time the bank contract was written (Beatty et al. 2002). However, many lending agreements are based on short to medium (less than three years) term loans, such as revolving credit agreements (Dichev and Skinner 2002), and even if the current debt contract was based on "frozen" GAAP, the provisions of SFAS 158 are likely to apply to upcoming contract renegotiations. In addition, the use of "frozen" GAAP might not apply to other debt contract provisions such as performance pricing provisions, or to how the company is perceived by debt rating agencies. 3. Prior research specifically related to retirement plans documents that the stock and bond markets consider both pension and postretirement benefit obligations as liabilities (Shaw 2008; Hann et al. 2007; Choi et al. 1997; Amir 1996; Amir 1993; Gopalakrishnan and Sugrue 1993; Barth 1991; Dhaliwal 1986). However, the stock market does not fully impound the

28

future valuation implications of PBO amounts disclosed in the footnotes but not recorded on the balance sheet (Picconi 2006). In addition, Davis-Friday et al. (1999) provide tentative evidence that the stock market puts less weight on postretirement benefit liabilities disclosed prior to the adoption of SFAS 106 than on the amounts subsequently recognized, possibly due to the disclosed liabilities having higher measurement error (Davis-Friday et al. 2004). 4. All financial variables are deflated by total assets. I also deflated by the projected benefit obligation for pension plan regressions and the accumulated postretirement benefit obligation for postretirement plan regressions and the results were consistent. 5. I also defined the change in the funded status as the industry adjusted change and the results were consistent. For ease of interpretation, the tables present the results not industry adjusted. 6. I did not include a variable for nearness to actual debt covenants or "covenant slack" for the following reasons. First, debt covenant information is rarely disclosed in annual reports and requiring this information would severely limit the sample size, and may even bias the sample towards companies disclosing debt covenant information. For example, Beatty and Weber (2006) obtain debt covenant data from the Loan Pricing Corporation database and reduce their sample of companies with goodwill write-offs from 553 companies to 176 companies when they require data on debt covenants. Second, debt contracts contain accounting-based provisions other than debt covenants, such as performance pricing provisions or dividend restrictions. Beatty and Weber (2003) find that companies are more likely to make incomeincreasing accounting method changes if their debt contract allows the changes to affect contract calculations, but only if the debt contract contains performance pricing provisions or dividend restrictions. Performance pricing provisions, which make the interest rate a function of the borrowers current credit rating or debt-related financial ratios, would not be captured by nearness to debt covenants but would be captured by leverage ratios. Third, many lending

29

agreements are based on short to medium (less than three years) term loans, such as revolving credit agreements (Dichev and Skinner 2002), and leverage ratios are a better measure of the ability to renegotiate loans or obtain new debt than closeness to current covenants. 7. Dichev and Skinner (2002) also find that leverage is negatively related to covenant slack for a sample of 30,000 private lending agreements in the Loan Pricing Corporation database; however, the correlation is economically modest implying that leverage is a noisy proxy for closeness to debt covenants. 8. I also define leverage as total liabilities divided by the market value of equity and the results were consistent. 9. Compustat began reporting information on postretirement plans in 2004 and even after 2004 data on postretirement plans reported by Compustat is limited. For example, Compustat collects no information on healthcare trend rate assumptions and the information on the amounts recorded on the balance sheet was sometimes missing. Therefore, most of the postretirement plan data used in this study were hand collected and the amounts reported by Compustat were checked against the retirement plan footnote. 10. Companies with retirement plans in foreign countries sometimes (but not always) report the discount rate and rate of compensation increase used to estimate the PBO of the foreign plans. As a sensitivity analysis, I averaged the foreign discount rate and rate of compensation increase with the domestic assumptions reported by Compustat and estimated models (3) and (4). The results were consistent. 11. Note that change in the funded status due to actuarial changes and amendments does not add up to the total change because there are other reasons for the funded status changing not reported in the tables (such as foreign currency translation).

30

REFERENCES Aboody, D. 1996. Recognition versus disclosure in the oil and gas industry. Journal of Accounting Research 34 (Supplement): 21-32. Aboody, D., M. Barth, and R. Kasznik. 2004. Companies voluntary recognition of stock-based compensation expense. Journal of Accounting Research 42 (2): 123-150. Amir, E. 1996. The effect of accounting aggregation on the value-relevance of financial disclosures: the case of SFAS No. 106. The Accounting Review 71 (4): 573-590. Amir, E. 1993. The market valuation of accounting information: the case of postretirement benefits other than pensions. The Accounting Review 68 (4): 703-724. Amir, E., and E. Gordon. 1996. Companies choice of estimation parameters: empirical evidence from SFAS No. 106. Journal of Accounting, Auditing, and Finance 11 (3): 427-448. Asquith, P., A. Beatty, and J. Weber. 2005. Performance pricing in bank debt contracts. Journal of Accounting and Economics 40 (1-3): 101-128. Asthana, S. 1999. Determinants of funding strategies and actuarial choices for defined-benefit pension plans. Contemporary Accounting Research 16 (1): 39-74. Barth, M. 1991. Relative measurement errors among alternative pension asset and liability measures. The Accounting Review 66 (3): 433-463. Beatty, A., K. Ramesh, and J. Weber. 2002. The importance of accounting changes in debt contracts: the cost of flexibility in covenant calculations. Journal of Accounting and Economics 33 (2): 205-227. Beatty, A., and J. Weber. 2006. Accounting discretion in fair value estimates: An examination of SFAS 142 goodwill impairments. Journal of Accounting Research 44 (2): 257-288. Beatty, A., and J. Weber. 2003. The effects of debt contracting on voluntary accounting method changes. The Accounting Review 78 (1): 119-142. Beaudoin, C., N. Chandar, and E. Werner. 2010. Are potential effects of SFAS 158 associated with firms decisions to freeze their defined benefit pension plans? Working paper, University of Vermont, Rider University, and Drexel University. Bens, D., and S. Monahan. 2008. Altering investment decisions to manage financial reporting outcomes: Assetbacked commercial paper conduits and FIN 46. Journal of Accounting Research 46 (5): 1017-1055. Bergstresser, D., M. Desai, and J. Rauh. 2006. Earnings manipulation, pension assumptions, and managerial investment decisions. The Quarterly Journal of Economics 121 (1): 157-195. Brown, L., and Y. Lee. 2007. The determinants and consequences of changes in executive option-based compensation around the issuance of SFAS 123R. Working paper, Georgia State University. Campbell, J., D. Dhaliwal, and W. Schwartz. 2010. Equity valuation effects of the Pension Protection Act of 2006. Contemporary Accounting Research, forthcoming. Carter, M., L. Lynch, and I. Tuna. 2007. The role of accounting in the design of CEO equity compensation. The Accounting Review 82 (2): 327-357. Chen, K., and M. Schoderbek. 2000. The 1993 tax rate increase and deferred tax adjustments: A test of functional fixation. Journal of Accounting Research 38 (1): 23-44.

31

Choi, B., D. Collins, and B. Johnson. 1997. Valuation implications of reliability differences: the case of nonpension postretirement obligations. The Accounting Review 72 (3): 351-383. Comprix, J., and K. Muller. 2006. Asymmetric treatment of reported pension expense and income amounts in CEO cash compensation calculations. Journal of Accounting and Economics 42 (3): 385-416. Davis-Friday, P., C. Liu, and H. Mittelstaedt. 2004. Recognition and disclosure reliability: evidence from SFAS No. 106. Contemporary Accounting Research 21 (2): 399-429. Davis-Friday, P., L. Folami, C. Liu, and H. Mittelstaedt. 1999. The value relevance of financial statement recognition vs. disclosure: evidence from SFAS No. 106. The Accounting Review 74 (4): 403-423. Dhaliwal, D. 1986. Measurement of financial leverage in the presence of unfunded pension obligations. The Accounting Review 61 (4): 651-661. Dichev, I., and D. Skinner. 2002. Large-sample evidence on the debt covenant hypothesis. Journal of Accounting Research 40 (4): 1091-1123. DSouza, J., J. Jacob, and K. Ramesh. 2001. The use of accounting flexibility to reduce labor renegotiation costs and manage earnings. Journal of Accounting and Economics 30: 187-208. Duke, J., and H. Hunt. 1990. An empirical examination of debt covenant restrictions and accounting-related debt proxies. Journal of Accounting and Economics 12 (1-3): 45-63. Eaton, T., and J. Nofsinger. 2004. The effect of financial constraints and political pressure on the management of public pension plans. Journal of Accounting and Public Policy 23 (3): 161-189. Fields, T., T. Lys, and L. Vincent. 2001. Empirical research on accounting choice. Journal of Accounting and Economics 31 (1-3): 255-307. Financial Accounting Standards Board (FASB). 2006. Statement of Financial Accounting Standards No. 158: Employers' Accounting for Defined Benefit Pension and Other Postretirement Plans. Norwalk, CT: FASB. Financial Accounting Standards Board (FASB). 2003. Statement of Financial Accounting Standards No. 132(R): Employers' Disclosures about Pensions and Other Postretirement Benefits. Norwalk, CT: FASB. Financial Accounting Standards Board (FASB). 1990. Statement of Financial Accounting Standards No. 106: Employers' Accounting for Postretirement Benefits Other Than Pensions. Norwalk, CT: FASB. Financial Accounting Standards Board (FASB). 1985. Statement of Financial Accounting Standards No. 87: Employers' Accounting for Pensions. Norwalk, CT: FASB. Godwin, J., S. Goldberg, and J. Duchac. 1996. An empirical analysis of factors associated with changes in pensionplan interest-rate assumptions. Journal of Accounting, Auditing, and Finance 11 (2): 305-322. Gopalakrishnan, V., and T. Sugrue. 1993. An empirical investigation of stock market valuation of corporate projected pension liabilities. Journal of Business, Finance & Accounting 20 (5): 711-724. Gramlich, J., W. Mayhew, and M. McAnally. 2001. Debt reclassification and capital market consequences. Journal of Business, Finance & Accounting 33 (7-8): 1189-1212. Hann, R., Y. Lu, and K. Subramanyam. 2007. Uniformity vs. flexibility: evidence from pricing of the pension obligation. The Accounting Review 82 (1): 107-137.

32

Healy, P., and K. Palepu. 1990. Effectiveness of accounting-based dividend covenants. Journal of Accounting and Economics 12 (1-3): 97-123. Imhoff, E., and J. Thomas. 1988. Economic consequences of accounting standards: the lease disclosure rule change. Journal of Accounting and Economics 10 (4): 277-310. Johnston, D. 2006. Managing stock option expense: The manipulation of option-pricing model assumptions. Contemporary Accounting Research 23 (2): 395-425. Langer, R., and B. Lev. 1993. The FASBs policy of extended adoption for new standards: an examination of FAS No. 87. The Accounting Review 68 (3): 515-533. Matsa, D. 2010. Capital structure as a strategic variable: evidence from collective bargaining. Journal of Finance 65 (3): 1197-1232. Mittelstaedt, H., W. Nichols, and P. Regier. 1995. SFAS No. 106 and benefit reductions in employer-sponsored retiree health care plans. The Accounting Review 70 (4): 535-556. OHara, M., and W. Shaw. 1990. Deposit insurance and wealth effects: the value of being ,,too big to fail. The Journal of Finance 45 (5): 1587-1600. Picconi, M. 2006. The perils of pensions: does pension accounting lead investors and analysts astray? The Accounting Review 81 (4): 925-955. Press, E., and J. Weintrop. 1990. Accounting-based constraints in public and private debt agreements. Journal of Accounting and Economics 12 (1-3): 65-95. Shaw, K. 2008. Revised pension rules and the cost of debt. Research in Accounting Regulation 20: 3-25. Schultz, E. 2004. Ford and GM get SEC request on pension accounting practices. The Wall Street Journal (October 20): A1. Scott, T. 1991. Pension disclosures under SFAS No. 87: theory and evidence. Contemporary Accounting Research 8 (1): 62-81. Solomon, D., and L. Hawkins. 2005. Pension inquiry shines spotlight on assumptions--Small changes in calculations at companies have a big effect on retiree liability and profit. The Wall Street Journal (November 9): C1. Sweeney, A. 1994. Debt-covenant violations and managers accounting responses. Journal of Accounting and Economics 17 (3): 281-308. Tung, S., and J. Weygandt. 1994. The determinants of timing in the adoption of new accounting standards: A study of SFAS No. 87, Employers accounting for pensions. Journal of Accounting, Auditing, and Finance 9 (2): 325-337. Walsh, M. 2008. After losses, pensions ask for a change. The New York Times (November 20). Watts, R., and J. Zimmerman. 1990. Positive accounting theory: a ten year perspective. The Accounting Review 65 (1): 131-156.

33

TABLE 1 Descriptive statistics and variable definitions Panel A: All sample companies with pension plans Mean Regression Variables: FS_PEN_TOTit 0.004 FS_PEN_ACTit -0.010 FS_PEN_AMDit -0.0002 DISCit -0.020 RCIit -0.043 MKTRETit 0.119 YIELDt 0.059 DEBTit-1 0.636 NWit 0.029 UNREC_PENit-1 -0.033 CFOit -0.002 ASSETS_PENit 0.126 SIZEit 7.538 EMPit 0.00003 DURATIONit 0.910 FIRMAGEit 35 FREEZE_PENit 0.098 Firm and Retirement Plan Information: Market value of equity Pension plan assets Projected benefit obligation Total retirement obligations as a % of total liabilities 7,484 2,042 2,298 48.1% Std. Dev. 0.019 0.014 0.002 0.169 0.268 0.058 0.001 0.199 0.075 0.038 0.020 0.053 1.806 0.001 0.060 16 0.306 16,062 7,428 7,833 35.3% 10th Percentile -0.015 -0.027 -0.001 -0.250 -0.375 0.052 0.057 0.382 -0.042 -0.077 -0.025 0.067 5.111 -0.001 0.829 10 0.000 166 37 47 17.1% Median 0.001 -0.007 0.000 0.000 0.000 0.101 0.058 0.635 0.030 -0.026 -0.001 0.116 7.624 0.000 0.917 42 0.000 2,048 365 448 37.0% 90th Percentile 0.026 0.004 0.00003 0.230 0.100 0.212 0.061 0.858 0.099 -0.003 0.021 0.206 9.763 0.001 0.983 55 0.000 17,377 4,280 4,861 93.4%

Panel B: All sample companies with other postretirement plans Mean Std. Dev. Regression Variables: FS_OPEB_TOTit 0.0003 0.008 FS_OPEB_ACTit -0.001 0.006 FS_OPEB_AMDit 0.001 0.005 HTRit -0.097 1.061 MEDACTit 0.001 0.002 UNREC_OPEBit-1 -0.009 0.015 ASSETS_OPEBit 0.032 0.065 FREEZE_OPEBit 0.022 0.148 Firm and Retirement Plan Information: Market value of equity 8,984 Postretirement plan assets 154 Accumulated postretirement 810 benefit obligation 18,954 946 4,602

10th Percentile -0.006 -0.007 0.000 -1.000 0.000 -0.026 0.000 0.000 260 0 7

Median 0.000 -0.001 0.000 0.000 0.000 -0.005 0.000 0.000 2,563 0 78

90th Percentile 0.006 0.003 0.003 1.000 0.002 0.002 0.123 0.000 20,756 221 1,242

34

TABLE 1 continued Descriptive statistics and variable definitions Shown are the descriptive statistics for the final sample of companies used in the regression analyses. For Panel A, the sample covers the period 2003 to 2006, and is comprised of 433 companies with the required data for all four years and that report a combined pension and other postretirement plan obligation greater than 10% of total liabilities. Panel B reports descriptive statistics for a sub-sample of 355 companies with other postretirement plans. Outliers are excluded by removing the top and bottom 1% of each variable. Variables are defined as follows: FS_PEN (OPEB)_TOTit = total change in the funded status of the defined benefit pension plan (other postretirement plan), deflated by total assets; where the funded status is defined as the fair value of plan assets less the projected benefit obligation (accumulated postretirement benefit obligation); FS_PEN (OPEB)_ACTit = change in the funded status of the defined benefit pension plan (other postretirement plan) due to changes in actuarial assumptions, deflated by total assets; hand-collected from the retirement plan footnote; FS_PEN (OPEB)_AMDit = change in the funded status of the defined benefit pension plan (other postretirement plan) due to plan amendments, deflated by total assets; hand-collected from the retirement plan footnote; DISCit = industry adjusted change in the discount rate used to estimate the projected benefit obligation; hand-collected from the retirement plan footnote; RCIit = industry adjusted change in the rate of compensation increase used to estimate the projected benefit obligation; hand-collected from the retirement plan footnote; HTRit = industry adjusted change in the initial healthcare cost trend rate used to estimate the accumulated postretirement benefit obligation; hand-collected from the retirement plan footnote; MKTRETit = market return (defined as the return on large company stocks weighted by the percentage of pension plan assets held in equity investments plus the yield on high quality corporate bonds weighted by the percentage of pension plan assets held in debt and other investments); YIELDt = yield on high quality corporate bonds (defined as the average of the yield on corporate bonds rated Aaa and Baa by Moodys as reported in the monthly Federal Reserve statistical release); MEDACTit = effect of the Medicare Prescription Drug, Improvement and Modernization Act of 2003 on the accumulated postretirement benefit obligation; DEBTit-1 = beginning of the year ratio of total liabilities to total assets; NWit = change in net worth, deflated by total assets; UNREC_PEN (OPEB)it-1 = beginning of the year unrecorded pension (other postretirement) asset or liability (defined as the pension (other postretirement) funded status less the amount recorded on the balance sheet), deflated by total assets; CFOit = average change in cash from operations over the past three years, deflated by total assets; ASSETS_PEN (OPEB)it = actual percentage return on pension (other postretirement) plan assets; SIZEit = the log of the market value of equity; EMPit = change in the number of employees, deflated by total assets; DURATIONit = duration of the pension plan (defined as the accumulated benefit obligation divided by the projected benefit obligation); FIRMAGEit = age of the firm FREEZE_PEN (OPEB)it = indicator variable equal to 1 if there was a hard, soft, or closed freeze to the pension plan (postretirement plan), and 0 otherwise.

35

TABLE 2 Funded status of defined benefit pension plans and other postretirement plans over time Panel A: Average across all sample companies with defined benefit pension plans (in millions) Pension Plan Change in PBO Year 2003 2004 2005 Average Pre-158 2006 PBO 2,170 2,371 2,557 2,366 2,637 Plan Assets 1,845 2,028 2,233 2,035 2,520 Funded Status -325 -343 -324 -331 -117

Change in Plan Assets Freezes and Settlements -3.26 -17.42 -3.18 -7.95

Total Change 93 -18 19 31 207

Service Cost 47 52 57 52 59

Interest Cost 125 129 132 129 139

Actuarial Change 128 125 125 126 -44

Amendments 12.44 1.63 0.20 4.76 -11.77

Actual Return 311 220 228 253 289

Contributions 121 84 91 99 81

-16.26

Panel B: Average across all sample companies with other postretirement pension plans (in millions) Other Postretirement Change in PBO Year 2003 2004 2005 Average Pre-158 2006 PBO 852 902 956 903 905 Plan Assets 131 172 196 166 210 Funded Status -721 -730 -760 -737 -695 Total Change -64 -9 -30 -34 65

Change in Plan Assets Freezes and Settlements 1.59 0.17 -0.16 0.53 7.76 Actual Return 18 15 17 17 22 Contributions 43 48 31 41 24

Service Cost 11 12 14 12 14

Interest Cost 50 50 51 50 51

Actuarial Change 87 55 36 59 -41

Amendments -3.38 -17.82 -17.75 -12.99 -67.74

Panel A reports the average each year across the final sample of 433 companies with the required data for all four years and that report a combined pension and other postretirement plan obligation greater than 10% of total liabilities. Panel B reports descriptive statistics for a sub-sample of 355 companies with other postretirement plans. The funded status is defined as the fair value of plan assets less the projected benefit obligation (PBO) for defined benefit pension plans and the accumulated postretirement benefit obligation (APBO) for other postretirement plans. The change in the funded status was hand collected from the companies retirement footnote and is broken down into changes to the PBO (service cost, interest cost, changes in actuarial assumptions, amendments to the retirement plan, and freezes, curtailments, and settlements) and changes to the plan assets (actual return on plan assets and contributions). Note that an increase to the PBO decreases the funded status.

, ,

Significantly different from the pre-SFAS 158 period at the 0.01, 0.05, 0.10 level, respectively (two-tailed test), based on a t-test of the difference in two means.

36

TABLE 3 Distribution of assumptions and market information over time Panel A: Discount rate Year 2003 2004 2005 2006 10th percentile 6.00% 5.65% 5.45% 5.65% 25th percentile 6.00% 5.75% 5.50% 5.75% Median 6.25% 5.80% 5.65% 5.86% 75th percentile 6.25% 6.00% 5.75% 6.00% 90th percentile 6.30% 6.00% 5.87% 6.00%

Panel B: Rate of compensation increase Year 2003 2004 2005 2006 10th percentile 3.25% 3.06% 3.20% 3.26% 25th percentile 3.60% 3.50% 3.50% 3.60% Median 4.00% 4.00% 4.00% 4.00% 75th percentile 4.50% 4.50% 4.42% 4.50% 90th percentile 5.00% 4.83% 4.75% 4.93%

Panel C: Initial health care cost trend rate Year 2003 2004 2005 2006 10th percentile 7.50% 7.94% 7.60% 7.25% 25th percentile 8.75% 9.00% 9.00% 8.50% Median 10.00% 10.00% 9.50% 9.00% 75th percentile 11.00% 10.00% 10.00% 10.00% 90th percentile 11.50% 11.00% 11.00% 10.50%

Panel D: Market Information Large Company Year Stock Return 2003 2004 2005 2006 28.70% 10.87% 4.91% 15.80%

Bond Yield 6.11% 5.81% 5.85% 5.77%

Percentage Equity Investments 64% 65% 65% 65%

Panels A, B, and C report the distribution over time of commonly disclosed retirement plan assumptions. The discount rate and rate of compensation increase is hand collected for 433 companies with defined benefit pension plans and the initial and ultimate health care cost trend rates is hand collected for a sub-sample of 355 companies with other postretirement plans. Panel D reports information on market returns and the average composition of pension plan assets. The large company stock return was obtained from Ibbotsons Stocks, Bills, Bonds, and Inflation. The bond yield is the yield on high quality corporate bonds defined as the average of the yield on corporate bonds rated Aaa and Baa by Moodys as reported in the monthly Federal Reserve statistical release. The percentage equity investments is the percentage of pension plan assets held in equity investments averaged across all companies.

37

TABLE 4 Distribution of changes in assumptions Panel A: Mean changes in assumptions for companies with low, medium, and high levels of debt Change in rate of compensation Change in initial healthcare Change in discount rate increase cost trend rate 2003-2005 2006 2003-2005 2006 2003-2005 2006 Low Medium High -0.34 -0.34 -0.35 0.18 0.20 0.21 -0.07 -0.08 -0.06 0.08 0.03 0.04 -0.21 -0.10 -0.17 -0.23 -0.34 -0.41

Panel B: Mean changes in assumptions for companies with low, medium, and high levels of unrecorded retirement assets or liabilities Change in rate of compensation Change in initial healthcare Change in discount rate increase cost trend rate 2003-2005 2006 2003-2005 2006 2003-2005 2006 Low Medium High -0.34 -0.34 -0.34 0.18 0.20 0.20 -0.05 -0.04 -0.12 0.04 0.02 0.09 -0.18 -0.15 -0.14 -0.36 -0.38 -0.24

Panel C: Mean changes in assumptions for companies with a low, medium, and high actual return on retirement plan assets Change in rate of compensation Change in discount rate increase 2003-2005 2006 2003-2005 2006 Low Medium High -0.35 -0.35 -0.34 0.22 0.20 0.17 -0.08 -0.08 -0.06 0.08 0.02 0.04

Panel D: Mean changes in assumptions for companies with small, medium, and large market value of equity Change in rate of compensation Change in initial healthcare Change in discount rate increase cost trend rate 2003-2005 2006 2003-2005 2006 2003-2005 2006 Small Medium Large -0.34 -0.35 -0.34 0.18 0.20 0.21 -0.07 -0.11 -0.03 0.05 0.07 0.02 -0.17 -0.14 -0.16 -0.29 -0.35 -0.34

This table reports the mean change in commonly disclosed retirement plan assumptions for the pre-SFAS 158 period (2003-2005) and the post-SFAS 158 period (2006). The discount rate and rate of compensation increase is hand collected for 433 companies with defined benefit pension plans and the initial and ultimate health care cost trend rates is hand collected for a sub-sample of 355 companies with other postretirement plans. Panel A reports the mean changes in assumptions for companies with low, medium, and high levels of debt (defined as the ratio of total liabilities to total assets at the beginning of the year). Panel B reports the mean changes in assumptions for companies with low, medium, and high levels of unrecorded retirement assets or liabilities (defined as the beginning of the year retirement plan funded status less the amount recorded on the balance sheet). Companies in the low category have the highest unrecorded liabilities and in the high category have the highest unrecorded assets. Panel C reports the mean changes in assumptions for companies with low, medium, and high levels of actual return on retirement plan assets. Panel D reports the mean changes in assumptions for companies with a small, medium, and large market value of equity.

38

TABLE 5 Changes in the defined benefit pension plan funded status Total change (FS_PEN_TOTit) Pred Sign (1) (2) Intercept -0.067 -0.067 (-6.24) (-6.47) MKTRETit 0.065 0.071 (1.80) (2.09) DEBTit-1 -0.001 -0.005 (-0.21) (-1.95) NWit 0.014 0.017 (2.38) (2.63) UNREC_PENit-1 -0.020 -0.089 (-1.62) (-7.87) CFOit -0.013 -0.045 (-0.62) (-1.85) SIZEit 0.001 0.001 (2.89) (4.43) EMPit -1.099 -1.146 (-3.08) (-3.37) DURATIONit 0.053 0.053 (7.22) (7.55) FIRMAGEit 0.00004 0.00004 (1.57) (1.52) FREEZE_PENit 0.006 0.005 (4.42) (3.67) FAS158 + 0.015 (2.66) DEBTit-1*FAS158 + 0.018 (4.00) NWit*FAS158 -0.002 (-0.13) UNREC_PENit-1 -0.259 *FAS158 (-10.81) CFOit*FAS158 ? 0.024 (0.51) SIZEit*FAS158 -0.002 (-4.58) 2 Adj. R 0.28 0.35 # observations 1,515 1,515

Actuarial change (FS_PEN_ACTit) (3) (4) -0.001 -0.0001 (-0.09) (-0.02) -0.007 -0.002 (-0.30) (-0.09) -0.005 -0.008 (-3.25) (-4.46) 0.007 0.013 (1.82) (2.76) 0.090 0.132 (11.59) (15.10) 0.005 -0.001 (0.30) (-0.03) 0.0004 0.001 (2.02) (2.61) 0.124 0.064 (0.51) (0.27) -0.007 -0.006 (-1.32) (-1.29) -0.00003 -0.00003 (-1.67) (-1.87) 0.002 0.001 (1.59) (0.85) 0.012 (2.92) 0.011 (3.53) -0.013 (-1.38) -0.160 (-9.56) -0.042 (-1.28) -0.001 (-2.60) 0.40 1,507

Amendments (FS_PEN_AMDit) (5) (6) -0.002 -0.001 (-1.34) (-1.16) 0.003 0.003 (0.68) (0.87) 0.0002 -0.00001 (0.99) (-0.04) 0.0001 0.0002 (0.22) (0.28) -0.001 -0.003 (-0.60) (-2.71) 0.001 -0.0001 (0.51) (-0.05) 0.00002 -0.000002 (0.64) (-0.05) -0.003 -0.006 (-0.07) (-0.15) 0.0002 0.0002 (0.21) (0.28) 0.000001 0.000001 (0.16) (0.13) 0.001 0.001 (3.42) (3.20) -0.001 (-1.03) 0.001 (1.94) 0.001 (0.47) -0.009 (-3.46) 0.002 (0.36) 0.0001 (1.13) 0.02 1,505 0.03 1,505

0.35 1,507

Shown are the results from a regression of the change in the pension plan funded status on the market return (MKTRETit), beginning of the year ratio of total liabilities to total assets (DEBTit-1), change in net worth (NWit), beginning of the year unrecorded pension asset or liability (UNREC_PENit-1), average change in cash from operations over the past three years (CFOit), the log of the market value of equity (SIZEit), change in the number of employees (EMPit), duration of the pension plan (DURATIONit), age of the firm (FIRMAGEit), indicator variable that is 1 if there was a plan freeze (FREEZE_PENit), and industry and year dummy variables (untabulated). All financial variables are deflated by total assets. FAS158 is a dummy variable that is 0 before SFAS No. 158 and 1 after. The sample covers the period 2003 to 2006, and is comprised of 433 companies with the required data for all four years and that report a combined pension and other postretirement plan obligation greater than 10% of total liabilities. Outliers are excluded by removing the top and bottom 1% of each variable.

, ,

Significant at the 0.01, 0.05, 0.10 level, respectively, (two-tailed test). T-statistics in parentheses.

39

TABLE 6 Changes in the other postretirement plan funded status Total change (FS_OPEB_TOTit) Pred Sign (1) (2) Intercept -0.003 -0.005 (-0.73) (-1.04) MKTRETit -0.006 -0.0004 (-0.30) (-0.02) MEDACTit 0.276 0.356 (2.73) (3.52) DEBTit-1 0.004 0.004 (3.15) (2.95) NWit -0.004 -0.007 (-1.13) (-1.97) UNREC_OPEBit-1 0.011 -0.043 (0.57) (-2.71) CFOit -0.010 -0.006 (-0.74) (-0.43) SIZEit -0.0001 -0.00002 (-0.85) (-0.12) EMPit -0.707 -0.744 (-2.84) (-3.02) FIRMAGEit 0.00001 0.00001 (0.72) (0.62) FREEZE_OPEBit 0.003 0.002 (1.79) (1.62) FAS158 0.006 + (1.55) DEBTit-1*FAS158 -0.0004 + (-0.16) NWit*FAS158 0.023 (2.79) UNREC_OPEBit-1 -0.195 *FAS158 (-5.71) CFOit*FAS158 ? -0.013 (-0.48) SIZEit*FAS158 -0.0003 (-0.82) Adj. R2 0.09 0.11 # observations 1,246 1,246

Actuarial change (FS_OPEB_ACTit) (3) (4) 0.001 -0.001 (0.31) (-0.26) -0.015 -0.010 (-1.05) (-0.69) 0.110 0.157 (2.35) (3.38) -0.002 -0.002 (-1.76) (-2.55) 0.003 0.004 (1.46) (1.57) 0.051 0.098 (4.75) (8.14) -0.005 -0.005 (-0.64) (-0.49) -0.0001 0.00004 (-0.78) (0.32) -0.039 -0.095 (-0.23) (-0.56) 0.00001 0.00001 (1.13) (1.00) 0.001 0.001 (1.38) (1.20) 0.008 (3.11) -0.002 (-1.10) 0.0003 (0.05) -0.185 (-7.86) -0.008 (-0.40) -0.0004 (-1.95) 0.18 1,243

0.14 1,243

Amendments (FS_OPEB_AMDit) (5) (6) -0.005 -0.005 (-1.76) (-1.69) 0.017 0.017 (1.45) (1.49) 0.048 0.044 (0.94) (0.86) 0.003 0.003 (4.68) (4.02) -0.003 -0.004 (-1.51) (-1.88) -0.045 -0.048 (-5.15) (-4.63) -0.009 -0.008 (-1.28) (-1.00) -0.00004 -0.0001 (-0.46) (-0.66) -0.186 -0.182 (-1.34) (-1.30) -0.00003 -0.00003 (-3.26) (-3.21) 0.003 0.003 (3.00) (3.01) 0.0002 (0.01) 0.0003 (0.17) 0.006 (1.32) 0.006 (0.32) -0.003 (-0.17) 0.0001 (0.65) 0.07 0.07 1,240 1,240

Shown are the results from a regression of the change in the other postretirement plan funded status on the market return (MKTRETit), impact of the Medicare Act (MEDACTit), the log of beginning of the year ratio of total liabilities to total assets (DEBTit-1), change in net worth (NWit), beginning of the year unrecorded other postretirement asset or liability (UNREC_OPEBit-1), average change in cash from operations over the past three years (CFOit), the market value of equity (SIZEit),change in the number of employees (EMPit), age of the firm (FIRMAGEit), indicator variable that is 1 if there was a plan freeze (FREEZE_OPEBit), and industry and year dummy variables (untabulated). All financial variables are deflated by total assets. FAS158 is a dummy variable that is 0 before SFAS No. 158 and 1 after. The sample covers the period 2003 to 2006, and is comprised of 355 companies with the required data for all four years and that report a combined pension and other postretirement plan obligation greater than 10% of total liabilities. Outliers are excluded by removing the top and bottom 1% of each variable.

, ,

Significant at the 0.01, 0.05, 0.10 level, respectively, (two-tailed test). T-statistics in parentheses.

40

TABLE 7 Choice of retirement plan assumptions Discount Rate (DISCit) (1) (2) Intercept YIELDt DEBTit-1 NWit UNREC _PEN/OPEBit-1 CFOit ASSETS _PEN/OPEBit SIZEit EMPit DURATIONit FIRMAGEit FREEZE _PEN/OPEBit FAS158 DEBTit-1*FAS158 NWit*FAS158 UNREC_PEN/ OPEBit-1*FAS158 CFOit*FAS158 ASSETS_PEN/ OPEBit*FAS158 SIZEt*FAS158 Adj. R2 # observations 0.05 1,503 -1.171 (-2.71) 18.522 (2.67) -0.013 (-0.55) 0.063 (1.04) 0.188 (1.65) -0.022 (-0.10) -0.023 (-0.18) 0.001 (0.19) 1.627 (0.45) 0.058 (0.75) 0.0003 (1.13) 0.027 (1.81) -1.134 (-2.66) 17.905 (2.59) -0.044 (-1.66) 0.104 (1.51) 0.135 (1.03) 0.125 (0.49) 0.236 (1.72) -0.005 (-1.70) 0.995 (0.28) 0.068 (0.88) 0.0004 (1.41) 0.031 (2.13) 0.013 (0.19) 0.105 (2.18) -0.165 (-1.16) 0.168 (0.65) -0.316 (-0.62) -1.712 (-4.52) 0.021 (3.78) 0.07 1,503 0.04 1,392 Rate of compensation increase (RCIit) (3) (4) 0.065 (0.48) 0.040 (0.29) Healthcare trend rate (HTRit) (5) (6) 0.059 (0.16) -0.084 (-0.21)

-0.016 (-0.40) 0.105 (1.02) -0.192 (-1.01) 0.748 (2.03) 0.307 (1.44) -0.002 (-0.40) -7.872 (-1.16) -0.287 (-2.25) 0.001 (1.72) -0.019 (-0.74)

0.033 (0.75) 0.046 (0.39) -0.222 (-1.00) 0.946 (2.20) 0.409 (1.78) -0.005 (-1.05) -8.037 (-1.19) -0.283 (-2.22) 0.001 (1.81) -0.014 (-0.54) 0.267 (2.42) -0.184 (-2.22) 0.219 (0.92) 0.090 (0.21) -0.538 (-0.64) -0.600 (-0.98) 0.014 (1.46) 0.05 1,392

0.223 (1.32) 0.376 (0.89) 5.818 (2.95) -0.226 (-0.13) 0.721 (1.35) -0.021 (-1.00) 25.925 (0.75)

0.398 (2.04) 0.652 (1.40) 5.611 (2.53) 0.065 (0.03) 0.593 (0.93) -0.021 (-0.91) 26.168 (0.75)

0.001 (0.58) -0.347 (-1.74)

0.001 (0.52) -0.350 (-1.76) 0.438 (0.92) -0.797 (-2.09) -1.984 (-1.75) 2.739 (0.57) -1.970 (-0.50) 0.513 (0.47) -0.0003 (-0.01)

0.01 1,220

0.01 1,220

41

TABLE 7 continued

Shown are the results from a regression of the change in industry adjusted assumptions on the yield on high quality bonds (YIELDt), beginning of the year ratio of total liabilities to total assets (DEBTit-1), change in net worth (NWit), beginning of the year unrecorded retirement plan asset or liability (UNRECit-1), average change in cash from operations over the past three years (CFOit), percentage return on retirement plan assets (ASSETSit), log of the market value of equity (SIZEit), change in the number of employees (EMPit), duration of the pension plan (DURATIONit), age of the firm (FIRMAGEit), indicator variable that is 1 if there was a plan freeze (FREEZEit), and industry and year dummy variables (untabulated). All financial variables are deflated by total assets. FAS158 is a dummy variable that is 0 before SFAS No. 158 and 1 after. Columns (1), (2), (3), and (4) report the results for a sample of 433 companies with the required data for the period 2003 to 2006, and that report a combined pension and other postretirement plan obligation greater than 10% of total liabilities. Outliers are excluded by removing the top and bottom 1% of each variable. The variables UNRECit-1, ASSETSit, and FREEZEit are based on the pension plan (_PEN). Columns (5) and (6) report the results for a subsample of 355 companies with other postretirement plans and the variables UNRECit-1, ASSETSit, and FREEZEit are based on the postretirement plan (_OPEB).

, ,

Significant at the 0.01, 0.05, 0.10 level, respectively, (two-tailed test). T-statistics in parentheses.

42

Information

44 pages

Find more like this

Report File (DMCA)

Our content is added by our users. We aim to remove reported files within 1 working day. Please use this link to notify us:

Report this file as copyright or inappropriate

1003635


You might also be interested in

BETA
Microsoft Word - SEM cover template for web.doc
Microsoft Word - ChevronValuation.doc
Microsoft Word - v2mod11d.doc