Pension conversion, termination, and wealth transfers.
This article explores the motivation to change their defined benefit pension plan by either terminating the plan and replacing it with a defined contribution plan or converting it to a cash balance plan. Using Form 5500 data as well as firm financial data, we find firms wishing to change their defined benefit plans are motivated by potential wealth transfer and tax implications. Firms terminating pension plans tend to have lower potential wealth transfers and lower taxes than firms converting to a cash balance plan, indicating a desire to modify the implicit contract instead of terminating the plan.
Traditional defined benefit plans sponsored by an employer provided a means to attract and retain a skilled workforce, offer lifetime income security, and provide for an effective compensation package. However, many defined benefit plan sponsors have determined the cost and risk of offering defined benefit plans burden the sponsor's operations and profitability. Sponsors face uncertain costs resulting from funding an unknown and estimated liability based on future salaries, years of service, and the average life expectancy of the workforce. In addition, sponsors also bear investment risk of the plan assets, which can increase the volatility of annual funding requirements. Finally, changes in regulatory environment also increase the cost and risk to the plan sponsor. The Employee Retirement Income Security Act (ERISA) increased protections to beneficiaries' pensions, but also increased regulatory, maintenance, and reporting costs, including the costs associated with restrictions placed on the firm's ability to design an optimal plan benefiting both the firm and employees (Clark and McDermed, 1990). McGill et al. (2005) find the cost of administrating a defined benefit plan has increased 7.43 percent annually between 1981 and 1996, but increased only 4.42 percent for defined contribution plans. Furthermore, between 1977 and 1985, Clark and McDermed (1990) document a decline of over 10 percent in pension plan participants covered by a defined benefit plan, demonstrating ERISA marked the beginning of a steady decline in the number of participants covered by defined benefit plans. Although one motivation for ending a traditional defined benefit plan is the increased costs related to a defined benefit plan, reducing financial distress or responding to the needs of the contemporary mobile workforce (i.e., attract better qualified employees) are other factors in pension plan design and implementation.
We model a firm's decision regarding each of its traditional defined benefit pension plans as a two-step process. First, a firm must decide whether to maintain the existing defined benefit plan or alternatively replace it with another plan. Second, if the firm decides to replace the plan, there are two common choices: (1) the firm may replace the plan with a defined contribution (DC) plan, or (2) the firm may convert to a cash-balance (or similar) plan. The purpose of this article is to determine the factors in this two-step decision leading to termination or to conversion. To do this we use a matched-firm cross-sectional analysis as well as firm- and plan-level data.
Our research adds to the existing literature of plan terminations and the relatively new literature on hybrid pension plans. Niehaus and Yu (2005) employ financial statement data and a change in the tax regime to test a tax motivation for choosing between a cash balance conversion and a defined contribution plan. We extend the study of Niehaus and Yu (2005) by analyzing individual plan-level, as opposed to aggregate firm-level data, as well as implementing a cross-sectional methodology. This methodology tests the cross-sectional variation in tax-paying status and funding status to determine if tax avoidance is a motivation for the method of defined benefit plan change. In addition, the method also allows simultaneously tests of alternative hypotheses of wealth expropriation and financial distress. The advantage to using plan-level data is twofold. First, firms may sponsor many DB plans, but choose to terminate or convert only one or two, so using firm-level data does not precisely measure the impact for these types of changes for firms with multiple DB plans if not all plans are changed. Second, these data, and not financial accounting data, are used to determine regulatory funding levels, as well as any pension funds reverting to the firm upon termination, which is essential in measuring tax implications as well as wealth transfer. We find taxes are an important motivation in both the decision to change the defined benefit plans as well as the decision between cash balance and defined contribution. Increase in a firm's tax burden increases the probability a firm chooses a cash balance conversion over a defined contribution plan. Furthermore, firms with large plans and potential wealth transfers are also more likely to choose a cash balance conversion. The results are broadly consistent with previous studies.
The rest of the article is as follows. The next section of the article reviews the relevant literature on plan terminations and cash balance conversions of defined benefit plans. Motivations for pension plan changes are given special attention. The "Hypothesis and Sample Selection" section discusses hypotheses related to ending a defined benefit plan and sample selection. Empirical methods and data requirements are discussed in the "Empirical Methods and Data" section. Empirical results and discussion are presented in the "Empirical Results" section. A final section summarizes the results and offers conclusions from the study.
Most research in this area focuses on the termination of a defined benefit plan, with only a few recent studies including a cash balance conversion piece. Early studies such as Alderson and Chen (1986), Moore and Pruitt (1990), and Mittelstaedt and Reiger (1990) find positive stock price reactions when firms terminate DB plans. This implies a wealth gain to shareholders from employees in the form of excess pension assets reverting to the firm once the plan is terminated or a decrease in future cash flows from the firm to employees or both. This wealth transfer extends to bondholders of the firm (Hsieh and Ferris, 1994).
If assets revert to the firm upon termination of the DB plan, it creates a taxable event and a rather expensive source of cash. As a result, only firms likely to pay lower taxes should terminate. Peterson (1992) and Haw, Ruland, and Hamdallah (1988) find support for this hypothesis, but Thomas (1989) does not find taxes as a motivating factor for terminating pensions. However, Thomas does find tax avoidance strategies do add value when terminating a DB plan.
In the presence of taxes, firms may still terminate a DB plan as a source of cash if they are financially constrained. Stone (1987), VanDerhei (1987), Thomas (1989), and Alderson and VanDerhei (1992) all find evidence consistent with financial constraints (access to external capital markets) is an important factor in the decision to terminate a DB plan. Furthermore, Rauh (2006) finds the funding requirement of DB plans poses a real constraint on capital expenditures of the firm and reduces capital investment. In another test of financial constraints, Datta, Iskandar-Datta, and Zychowicz (1995) find a wealth transfer from bondholders to stockholders as a result of DB plan termination, consistent with the financially weak firms terminating DB plans.
Instead of a financial constraint, DB plans may be a source of financing for acquisitions. A firm with an overfunded plan becomes a target of an acquisition. Upon completion of the takeover, the target's DB plan is terminated and excess assets are used to fund (pay down debt, of fund the acquirer's underfunded DB plan) the acquisition. Pontiff, Shleifer, and Weisbach (1990) and Cather, Cooperman, and Wolfe (1991) find evidence for acquisition motivated terminations, but Mitchell and Mulherin (1989) find little support for this theory.
Besides the immediate wealth transfer resulting from reversion of excess DB assets to the firm, another wealth transfer resulting from the termination of the implicit employment contract is also researched. A traditional view of DB plans stems from the implicit contract theory. The implicit employment contract is created when the employee accepts lower current compensation for a deferred compensation package following retirement (Ippolito, 1985). (1) The firm fulfills this contract by maintaining and funding the pension fund. Terminating the DB plan breaks this implicit contract. Both Petersen (1992) and Mittelstaedt and Regier (1993) model the implicit contract between the firm and employees created by a DB plan. Using plan-level data, Mittelstaedt and Regier (1993) follow Ippolito's (1986) methodology and estimate the implicit and explicit liabilities for DB plans, and then measure the potential and actual wealth transfer resulting from plan termination. Under implicit contract theory, the accrued (explicit) liability is the amount employees have earned while the additional expected future benefits (implicit liability) resulting from of continued employment, salary increases, and cost of living adjustments together create the total economic liability. They find their measure of actual wealth transfer, the difference between economic and explicit liabilities (the estimated implicit liabilities) explains more of the abnormal returns than just the asset reversion amount. Petersen (1992) also shows larger pension plans and those most overfunded have a greater probability of termination, after controlling for tax considerations and firm financial strength.
An alternative to termination began in the mid 1980s with the introduction of cash balance plans by the Bank of America. Following this example, firms began converting traditional DB plans to CB or other hybrid plans. Under a cash balance plan, the benefit is determined by the amount contributed and a stated rate of return. Upon retirement, the employee receives the accumulated "cash balance" of the plan, which can be annuitized at the discretion of the employee. The lump-sum benefit more closely resembles a defined contribution than a traditional defined benefit where the employee receives an annuity usually based upon final average salary, number of years service, and a multiplier. In addition, a plan conversion does not terminate the original plan, only modifies the benefit formula. As a result, conversions do not create a taxable event and do not require a particular funding status at the time of conversion.
Conversions to cash balance plans have been contentious and subject of age-biased lawsuits as demonstrated by IBM's high-profile case. In 2003, a court ruled the cash balance conversion was a form of age discrimination, effectively stopping many cash balance conversions. An appellate court reversed the lower court's decision and ruled in favor of IBM in 2006. The Supreme Court refused to hear the appeal, finally resolving the case in 2007.
Research by Clark and Schieber (2004a, 2004b) and Coronado and Copeland (2004) find labor market considerations (especially a more mobile workforce) as a primary motivation for CB conversions, but D'Souza, Jacob, and Louge (2004) find less profitable firms, service industry firms, and those with higher than average pension costs tend to convert to a CB plan. However, the research by Niehaus and Yu (2005) is most relevant to our study. They sample firms converting to a CB plan with those terminating and replacing a DB plan with a defined contribution plan. They test and find support for tax avoidance as a primary motivation for choosing a cash balance conversion by comparing a change in the tax law during the sample period that affected reversion of pension assets. The change in the tax law created an excise tax of 50 percent of pension assets reverting to the sponsor. The tax on asset reversion would decrease to 20 percent if the sponsor increased benefits by 20 percent prior to termination or transferred 25 percent of excess assets to a qualified replacement plan. In effect, tax change ensured the plan sponsor will only gain just over half of the excess assets (Allen, 2003). (2) To test the implicit contract theory, they use poorly performing firms (less profitable) as a proxy for the potential wealth loss by employees. They find that poorly performing firms are more likely to terminate a DB plan than to convert to a cash balance, concluding cash balance conversions are not consistent with implicit contract theory.
Our study differs from the previous research on cash balance plans in several respects. First, we employ a methodology that matches firms transitioning from a DB plan with those retaining a DB plan. The method allows us to capture a cross-section of funded status, tax status, and plan size while controlling for industry and firm size effects. In addition, the sample period does not include any major regulatory changes or tax policies related to DB plans. Second, we use more precise measures of implicit contract relationships by using plan-level data from annual plan filings in Form 5500 (similar to Ippolito, 1986; Mittelstaedt and Regier, 1993), as opposed to only firm-level data from financial statements. This is important because a firm may choose to change only one of several defined benefit plans it sponsors. It is also important to use plan-level data because it is these regulatory data that determine if the plan is funded and if excess assets exist. In addition, data from regulatory filings are less subject to discretionary assumptions found in reported financial statement information. Finally, it best estimates the wealth transfers, which result from ending a DB plan. This is in contrast to Niehaus and Yu (2005) who test the change in regulation and tax treatment on the cash balance/defined contribution decision.
HYPOTHESIS AND SAMPLE SELECTION
Based upon the previous literature, a firm has the choice to continue with its current DB plan or replace it with another type of retirement plan. The choice to end the current DB plan is determined by the sponsors need to attract qualified employees, mitigate financial constraints, or outright wealth expropriation. Tax implications further influence the decision to change the plan. Any change in the current defined benefit plan represents a change or elimination of the implicit employment contract, leading to a wealth transfer to the firm if no offsetting compensation or other benefit changes are added. Once the decision to change the current defined benefit plan is made, the firm pursues one of two options. (3) The first option terminates the DB plan, replacing it with a DC plan. The alternative converts the define benefit plan to a hybrid plan such as a CB plan.
Hypotheses Related to Changing a Traditional Defined Benefit Plan
Financial constraints, tax avoidance, the potential wealth transfer, as well as labor market factors are not mutually exclusive hypotheses for ending a DB plan. Under this set of hypotheses, the CB conversion option or DB plan termination (replaced with a DC plan) is treated as a similar decision. From an employee perspective, the plans are similar, but not identical, in terms of benefits accrued and portability. Industries relying on a more mobile workforce, typically more technology and service oriented, will tend toward changing the traditional DB plan to attract qualified personnel (Coronado and Copeland, 2004). The financial constraint hypothesis indicates firms with higher debt ratios, lower cash flows, and less fixed assets are more likely to end traditional DB plans. Firms can end a traditional DB plan to use excess pension assets to fund operations if the firm has difficulty raising external capital. The tax avoidance hypothesis would show firms that have excess pension assets along with lower tax rates and higher net operating losses will be more likely to end DB plans since the tax cost of reverting assets would be lower. Finally, support for the wealth transfer hypothesis would indicate a positive relationship between the wealth transfer or excess funding of the plan and the decision to end a DB plan.
Hypotheses Favoring Cash Balance Conversion or Termination
Given the decision to end the traditional DB plan, the firm decides whether to convert the plan to a CB plan, or terminate the plan and replace it with a DC plan. With either decision, the firm changes the implicit contract. When a DB plan is part of employee compensation, the loss or change to the plan entails a change in compensation and possibly a wealth transfer, particularly when the plan is "backloaded" as most DB plans are. A termination of a DB plan immediately voids the deferred compensation expected and creates a wealth transfer (or expropriation) in the absence of other changes in compensation and benefits. Cash balance plans are still defined benefit plans and also change the deferred compensation expected. However, plans assets remain in the plan and are property of the participants and beneficiaries. The plan sponsor funds the plan, promises a rate of return on plan assets to the beneficiary, and is responsible for managing plan assets. In addition, cash balance plans require the plan sponsor to maintain funding levels and make additional contributions when plan assets fall short of the present value of pension liabilities and all pension fund assets remain with the plan after conversion (no asset reversion to the firm). However, cash balance plans do not provide lifetime income security offered by a traditional defined benefit plan. In return for this potential (not actual) wealth expropriation, employees gain some degree of portability, both in career and retirement. By comparison, a DC plan requires none of the above. In fact, while employers contribute to an employee's defined contribution account, the majority of DC plans require at least partial, if not majority of, contributions from the beneficiary (i.e., matched contributions). Furthermore, employers can easily suspend contributions for many types of defined contribution plans, such as profit sharing and 401(k) plans, and are not required to make a set contribution. After contributions, there is no promised rate of return on invested assets, leaving the employee to bear the investment risk. Unless there are increases in salary or other benefits, a plan termination with a DC replacement is similar to a wealth transfer from employees to the firm, while a CB conversion can be viewed to some extent as an implicit contract modification. As such, firms ending a traditional DB plan to expropriate wealth will choose to terminate the DB plan.
From a funding and tax perspective the choice between termination and conversion is complex. To voluntarily terminate (i.e., not part of a bankruptcy filing) a DB plan and replace it with a DC plan first requires the plan to be fully funded at termination, requiring a contribution to an underfunded DB plans prior to termination. For deficits in a large plan, this requirement is in many cases prohibitive, but much less so for small plans or funding deficits. However, CB conversions do not require funding at conversion since conversion is considered a modification to the defined benefit plan. Firms facing a financial constraint are less likely to terminate a severely underfunded DB plan because of the explicit funding requirement, but termination is more likely for a slightly underfunded plan.
If the plan is overfunded, excess plan assets revert to the corporation upon termination less an excise tax of 20 percent if the sponsor either increases benefits in the current plan or transfers at least 25 percent of excess assets to a new, qualified replacement plan. (4) In addition, the sponsor may still face taxes based on gains and income from pension assets. For a CB conversion, pension assets remain with the plan and the conversion does not create a taxable event. As with all overfunded DB plans, the firm can use this excess funding in terms of a slow withdrawal (making reduced or no further contributions to the plan), reducing the funding status of the plan to full funding. The unequal treatment of excess assets for tax purposes introduces the importance of tax strategies. Operating losses and low tax rates increase the likelihood and attractiveness of terminating an overfunded plan as the firm offsets gains from plan investments with net operating losses, mitigating the taxes paid. (5) A financially constrained firm without tax saving strategies can still choose to terminate the plan and use excess assets, but this is a costly form of funding. For firms that are not financially constrained and do not have tax saving tools, CB conversion of overfunded plans is more likely than termination. The hypotheses suggest significantly overfunded or underfunded plans tend to convert to CB plans, but plans near full funding are more likely to terminate, ceteris paribus.
To test the hypotheses, we first determine a sample of firms who terminated a DB plan and replaced it with a qualified DC plan, as well as a sample of firms from converting a traditional DB plan to a CB plan. From the Pension Benefit Guaranty Corporation (PBGC), we obtain the list of firms from 1997 through 2003 voluntarily terminating plans and filing Form 500 and associated forms. These include all plans, both large and small employer plans (less than 100 participants). As with all IRS filings, plans are identified by the firm's employer identification number (EIN). Each plan files/terminates separately, even if sponsored by the same firm. There were over 6,300 plans terminated during this sample period. However, the vast majority of these plans are "small" plans (about 5,600) with less than 100 participants. (6) Until 1999, these small plans did not file a full report annually. Of the remaining 730 plans, there are 168 plans sponsored by 144 firms that could be identified as publicly traded companies based on EIN matches with Compustat. (7) Of these 144 firms, plans were removed from the sample if the termination was due to an acquisition. This screen yielded 62 unique firms voluntarily terminating 68 defined benefit plans. In addition, the sample was screened for data availability on Compustat as well as plan annual reports filed through Form 5500. (8) The final termination sample resulted in 43 firms with complete data. Of these firms in the termination sample, 18 were overfunded and had assets subject to an excise tax. Nine firms did not return 25 percent of reverting assets to a new plan, but it is unclear if these firms increased current plan benefits prior to termination to lower the excise tax.
Because converting to a CB plan is considered to be a modification of a DB plan, no special notice or public filings are required. As a result, identifying firms undergoing a cash balance conversion becomes more difficult. We initially identify 93 publicly traded firms converting 103 plans for the time period 1996-2003. We begin with a sample from a list in Pensions & Investments published in 1999 (Anand, 1999). We then searched annual reports to verify the year of conversion. Next we obtained data from the pension filing Form 5500 for years 1999-2003 that list description codes for defined benefit plans and then identify new instances of CB plans. Part of our sample period overlaps with Niehaus and Yu (2005). They find 76 plans converting from 1996 to 2000 (over half their sample), while we are able to identify 90 firms. Of these identified plans, data are available from both Compustat and Form 5500 for 80 firms. The resulting sample is 122 (9) firms that either terminated or converted their sponsored DB plan during the sample period.
The final part of our sample consists of firms with DB plans with no changes in their plan during the sample period. We select a comparison firm sample to analyze the decision of whether to change the DB plan (by termination or conversion to CB plan) or to continue the plan in its current form. From the Form 5500 database, we delete any plan that is frozen during our sample period. From the remaining plans of publicly traded firms, we select matching firms based on firm size and industry. Matched firms are in the same three- or two- digit SIC code as the sample firm and are approximately the same size as sample firms, within 20 percent of total assets of the firm. The matching techniques produce at least one firm match for 111 of the 122 firms in the sample with enough data to be used in the estimation techniques described later. Since pension plan data from Form 5500 are reported by individual plans (not by firm), firms may have multiple plans and entries. These plan- level data are computed as a weighted average of the plan based on the plan's liabilities for each firm so pension data are the average plan for the firm.
We use three different types of control samples from this matched set. The first technique uses the closest firm on average total assets prior to pension change and SIC code in the estimation techniques described later. The second technique uses all firms (up to four matches) in the estimation techniques. (10) By using all matched firms we demonstrate the results are not sensitive to the particular matched firms that were selected. However, since not all sample firms have an equal number of matches, a third technique constructs an equally weighted portfolio of matched firms (to the same sample firm). The matched portfolio firm averages all data used in the estimation techniques, such as plan size, firm size, potential wealth transfer, funded level, and so on. This creates a one-to-one match to each sample firm.
EMPIRICAL METHODS AND DATA
The choice to change or terminate the current DB plan the firm sponsors depends upon several factors, including cost, risk to the firm, ability to attract and retain employees, and the financial condition of the firm. Given a firm has decided to end its DB plan in the current form, it chooses between terminating the current plan and replacing it with a DC plan or converting the plan to a CB plan.
The Decision to End the Traditional DB Plan
The two-step decision process (first, the decision to change; second, the choice of cash balance versus termination) leads to several possible choice models, including a nested logit, conditional logit, and multinomial logit models. To determine which model is appropriate, we first conduct the Hausman test for the Independence of Irrelevant Alternatives (IIA) assumption. The test did not reject the IIA assumption (probability of 0.604); therefore, a simple logit or multinomial logit models are appropriate. We create the variable, Change, equal to 1 if firms change their DB plan and equal to 0 if firms continue their defined benefit pension plans during the sample period. We estimate the following model:
P(Change) = [[beta].sub.0] + [[beta].sub.1]Tax*Over + [[beta].sub.2]NOL*Over + [[beta].sub.3]OperatingCF + [[beta].sub.4]FixedAsset + [[beta].sub.5]IntangAsset + [[beta].sub.6]DebtRatio + [[beta].sub.7]PlanSize + [[beta].sub.8]RelativePWT + [[beta].sub.9]Funded + [[beta].sub.10][FundedRatio.sup.2] + [epsilon], (1)
where Tax*Over is equal to 1 if the firm paid federal tax in the previous year and was overfunded, 0 otherwise, and NOL*Over is a dummy variable indicating if the firm has a net operating loss carryforward and was overfunded, and 0 otherwise. These two variables relate to the tax hypotheses. If taxes are an important constraint in the termination decision, then coefficients for Tax*Over should be negative and NOL*Over should be positive. OperatingCF is the average cash flow to assets for the firm over the previous 3 years and DebtRatio is the firm's debt-to-asset ratio in the year prior to the pension change. These two variables relate to the financial constraint hypothesis. If financial constraints are relevant in the firm's decision, estimated coefficients for DebtRatio should be positive and OperatingCF should be negative.
FixedAsset and IntangAsset are the firm's fixed assets to total assets and intangible assets to total assets, respectively. These variables control for the firm's operating leverage as well as industry effects. In addition, the level of intangible assets proxies for the firm's industry and technology and could indicate different levels of employee mobility. The data for these variables are collected from Compustat and are data from the firms' financial statements.
The final set of variables is obtained from the pension annual reports through filings of Form 5500, which are reported annually for individual plans. The control variable of PlanSize, the log of pension liabilities, is used to account for the size of the defined benefit plan. RelativePWT, Funded, and [FundedRatio.sup.2] are included to test the importance of wealth transfer and funded status to the decision to change the defined benefit plan. Funded is calculated as the plan's pension assets divided by pension liabilities minus 1. A fully funded plan (100 percent funded) will have a value of zero, whereas an underfunded (overfunded) plan's value will be negative (positive). In addition, we include the square of the funding ratio to control for the degree of funding difference from full funding.
The PWT variable is the potential wealth transfer and estimates the value of the implicit contract and wealth expropriation hypothesis (estimated in a process described later). Since PWT and the size of the plan are related, we scale the PWT by the amount of the pension liabilities to create a relative PWT measure. A positive relationship between RelativePWT and the decision to change or terminate the DB plan indicates breaking the implicit contract and transfer of wealth is an important factor. RelativePWT measures the possible additional dollars wealth transfer to the sponsor's obligations. The PWT is approximately equal to the difference between the economic liabilities (the liabilities of the plan if allowed to continue, including future salary and service accrual) and explicit liabilities, the amount employees have currently earned. The difference is a loss to employees of pension benefits resulting in a wealth transfer to the firm (and shareholders). Ippolito (1986) calculates the potential wealth transfer as:
PWT = ([EC.sup.Act] - [EX.sup.Act]) + ([EC.sup.Ret] - [EX.sup.Ret]), (2)
where EC and EX represent the economic and explicit liabilities, respectively, and ACT and RET denote the liabilities for active employees and retired employees. (11) To adjust the reported liabilities on Form 5500 (annual pension plan report) to economic and explicit liabilities, estimates of the appropriate discount rate from the reported information in the annual report are obtained from the following equation:
[MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII], (3)
where L, C, P, and N are the reported on the annual report as the total pension liability, number of active participants, average annual pension, and number of retired participants, respectively. [I.sup.Act] and [I.sup.Ret] are the assumed interest rates for active and retired employees, respectively. The estimated [[beta].sub.2] and [[beta].sub.4] are used to calculate the discount factors for active and retired employees and are 0.0528 and 0.0108, respectively. These estimates are lower than those obtained by Mittelstaedt and Regier (1993) but are consistent with those findings given the prevailing interest rate environments. Once these coefficients are estimated, the economic liabilities for active and retired employees are calculated as:
[MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII], (4)
[MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII]. (5)
To convert reported liabilities to explicit liabilities:
[MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII] (6)
[MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII], (7)
where r is the real interest rate, AAA is the average interest rate on a AAA bond, and i is the inflation rate calculated as the difference between the real rate and the AAA rate. The resulting estimates of explicit and implicit liabilities are used to calculate the PWT in Equation (2) for each pension plan.
The above simple logit model (Equation (1)) assumes converting to a cash balance plan and terminating the plan and replacing it with a DC plan are essentially the same decision, and factors affecting that decision are the same. However, the two methods differ with respect to several important factors including tax consequences, access to surplus assets, the breach of the implicit contract, or the firm's need to attract employees in a mobile labor market.
Determining the Method of Change
Since a DB plan termination and a cash balance conversion have different funding requirements, as well as different tax and cash flow implications, we now test the determinants of the firms' choice of methods when ending the traditional plan. We again employ a logit model to test competing hypotheses. The model uses many of the same explanatory variables used in the logit model above. The model to explain the method of ending the DB plan takes the form:
P(DC) = [[beta].sub.0] + [[beta].sub.1]OperatingCF + [[beta].sub.2]FixedAsset + [[beta].sub.3]IntangAsset + [[beta].sub.4]DebtRatio + [[beta].sub.5]FirmSize + [[beta].sub.6]PWT + [[beta].sub.7]RelativePWT + [[beta].sub.9]Funded + [[beta].sub.9]Overfunded, + [[beta].sub.10]NOL*Over + [[beta].sub.11]Tax*Over + [[beta].sub.12][FundedRatio.sup.2] + [epsilon], (8)
where DC is equal to 1 for defined contribution and 0 for a cash balance conversion. The tax, financial constraint, funding, and operating structure variables are the same as in Equation (1). The additional variable Overfunded is a dummy variable indicating if the plan is overfunded. This variable allows for alternative model specifications to test the degree funding level impacts the decision. Also, FirmSize is the log of firm assets and is used as a control variable for the model.
Since plans must be fully funded when voluntarily terminated, we expect the funded status of the plan to be an important determinant in the decision to terminate versus convert. Underfunded plans are less likely to terminate but more likely to convert, with severely underfunded plans least likely to terminate. Overfunded plans are more likely to terminate, given this requirement. However, substantially overfunded plans may convert (as opposed to terminate) to avoid the taxes on reverting assets. Because of this conflict between tax implications and plan funding requirements at termination, we expect a nonlinear or asymmetric relationship with the termination decision and funding level. We include the squared value of the funded ratio, [Funded Ratio.sup.2] and an indicator variable if the plan is overfunded in the estimation of the logit model.
Mitigating the effects of funding levels on the termination/conversion decision are tax consequences, with overfunded plans facing significant tax implications. We specify an interaction variable of Tax*Over that takes the value of 1 if the plan is overfunded and the firm is currently paying tax. We expect the coefficient to be negative and significant if taxes are a major factor in determining the method of changing the DB plan. We also include information about the firm's current tax status. If the firm has net operating losses it is carrying forward and has excess pension assets (overfunded), then it is more likely to terminate than to convert, but if the firm is currently paying federal tax and has excess pension assets, termination of an overfunded pension plan will likely involve paying substantial taxes.
The choice of whether to terminate or convert to a CB plan is also subject to financial constraints. Firms facing financial constraints (proxied by low operating cash flows and high debt ratio) will be more likely to terminate an overfunded plan, even in the event of high tax rates. However, underfunded plans would likely not be terminated by plan sponsors in financial distress as the plans do not provide capital to the firm.
Previous studies raised issues of implicit contracts and the changing labor market as motivations for conversions to CB plans as opposed to terminations. Because of an increase in labor mobility, decline of manufacturing jobs, and increased competition in those industries, a defined benefit plan is not as appealing to workers as in the past. Even if the firm is financially healthy, it may want to change the defined benefit plan to attract or retain valuable employees. If the firm and employees value the implicit contract, the firm is less likely to break the implicit contract outright (terminate the defined benefit plan) and more likely to convert to a cash balance plan (modify the contract) and the estimated coefficient for PWT should be negative. We estimate both the size of the PWT (logPWT) and the relative size of the PWT to plan liabilities.
Given the discussion earlier, we expect the DebtRatio, NOL*Over, and FixedAsset to be positively associated with termination versus cash balance conversion. Firms with high fixed assets generally have higher operating costs and would be more likely to terminate a plan to reduce fixed expenses. We also expect firms with higher Operating CF and [FundedRatio.sup.2] to be less likely to terminate and more likely to make a cash balance conversion. The higher the funding ratio [FundedRatio.sup.2], the less likely the firm will terminate due to taxes if it is overfunded and because of the contribution required to fully fund the plan if it is underfunded.
The question of the impact of the funded ratio depends upon which hypothesis dominates. Only fully funded plans can voluntarily terminate. If the decision is based solely on terminating at the lowest cost (no large contributions to fully fund the plan prior to termination), then we would expect a positive relationship. However, if the firm is trying to avoid paying taxes when the plan is overfunded as well as avoid large contributions to fully fund the plan, then we would expect a negative relationship with the funded ratio.
The analysis of the decision to change the defined benefit plan uses a control sample matched on industry and size. Summary statistics of firms either converting or terminating defined benefit plans along with the matching firms (multiple matches) are presented in Table 1. As the table indicates, substantial differences between the sample and the matching firms exist. In the matching procedure, we identify up to four matching firms to each firm in the sample; however, not all firms have four matches. Although matched on size, the matching sample is on average smaller than the sample. This is primarily due to more multiple matches on the largest firms in the sample and fewer matches on the smaller firms. The closest one-to-one matched firms exhibit similar average asset amounts. In addition to firm assets, the pension liabilities and potential wealth transfers are on average larger for sample firms than for matched firms, but median values indicate a close and consistent relationship between the two sets. However, when the PWT is measured in relation to plan liabilities, both mean and median values indicate firms changing the defined benefit plan have higher ratios, relative PWT, than the comparison sample. (12) A larger relative PWT ratio is an indication of the greater implicit liability (yet to be earned by beneficiaries) relative to accrued liabilities and indicates the sample firms face relatively large future liabilities from the pension plan and have chosen to reduce those expected liabilities.
For firm characteristics, the sample and matching firms demonstrate similar values for fixed and intangible assets, as well as operating cash flow ratios, indicating a consistent match on industry type (manufacturing or service and technology) as well as relative performance of firms. However, there are differences related to taxes (net operating losses and federal taxes paid) as well as financial leverage. Sample firms appear to be less likely to owe federal tax and have higher financial leverage ratio, indicating these firms generally have more risk and experienced recent or current losses.
With regard to funding, both sample and matching firms tend to be overfunded when measured according to the Department of Labor and PBGC. The average funding ratio for sample firms is 0.19, and matching firms have an average value of 0.26. In addition, sample firms tend to have values closest to full funding while the matched sample experiences wide deviations as demonstrated by the higher absolute funding ratio and square of the funding ratio. Overall, these summary statistics indicate sample firms are a bit larger, have higher leverage ratios, are less likely to owe federal tax, and have funding ratios closer to full funding than do control sample firms. However, the matched firms are very similar to the firms in the sample in many operational and pension related aspects.
We test, in a univariate setting, the differences in plan and firm characteristics between sample firms choosing to convert to a cash balance plan and those terminating the defined benefit plan and replacing it with a defined contribution plan. Table 2 presents the results from these tests. The pension data presented in Table 2 are at the plan level and compare the plans that were terminated with those that converted to a cash balance plan. In general, firms choosing to convert plans typically have plans with larger pension liabilities and potential wealth transfers, but are also larger firms in general. Although either set of firms typically have the same amount of leverage, firms choosing to convert as opposed to terminate the pension plan are more likely to pay federal tax, and have a greater percentage of fixed assets, fewer intangible assets, and higher operating cash flows.
With respect to pension funding levels, firms choosing to convert have higher funding ratios and are typically overfunded but have lower squared funding ratios. The univariate results indicate a firm's tax status, its ability to spread the fixed regulatory and administrative cost of DB plan (as proxied by firm size), and the potential pension wealth transfer from changing the benefit formula are related to a firm's decision to switch to a DC-type plan. In addition, those firms with lower operating cash flow tend to favor terminations (and immediate access to reverting plan assets), but these plans typically do not offer a large cash flow (smaller liabilities and assets), and the firms do not have other indications of financial distress (higher leverage ratios). For the funding ratios the results indicate that large overfunded plans are more likely to convert; however, significantly overfunded or underfunded plans (as measured by the squared funding variable) are more likely to terminate and switch to a define contribution plan.
Although the univariate tests give us some insights into the motivations and implications of the decision to change and method of changing the defined benefit plan, we formally test these hypotheses in logit a model presented in Equations (1) and (8). A simple logit and multinomial logit model tests the above hypotheses with firms choosing to change their defined benefit plans (either through termination or a cash balance conversion) or remain the same. We present both the estimated coefficients and the marginal effects from the estimated models in Table 3.
The first logit model uses a one-to-one match to test the decision to end the traditional defined benefit plan. We find a match for 111 of our sample firms (73 conversions and 38 terminations). The first model specification includes all sample firms with the corresponding matched firms. In general, the results confirm the findings in the univariate tests. There is strong support for the tax motivation hypothesis indicated by the positive coefficient and marginal effects of NOL and negative values (in multinomial model) for federal tax. Firms with overfunded plans and not owing federal tax or have net operating losses (reduce taxable income) will have a greater probability of changing the defined benefit plan. The marginal effects indicate firms with a net operating loss and an overfunded plan are approximately 30 percent more likely to change. In addition, there is a direct relationship between the level of fixed assets (operating leverage) and probability of changing the defined benefit plan. Finally, potential wealth transfer is a motivating factor for changing the defined benefit plan with greater RelativePWT providing a motivation for changing the plan. The marginal effects of the RelativePWT indicate that every percentage point increase in the size of the wealth transfer as a percentage of the liabilities, there is a threefold increase in the likelihood to change the defined benefit plan, thereby eliminating large future explicit liabilities.
The effects found in the combined model are consistent when the logit model is applied to each type of pension change when estimated separately. The marginal effects for the sample of cash balance conversions are similar to those reported for the combined model, while the termination/defined contribution sample demonstrate differences, especially with respect to plan size as seen in the multinomial logit estimation. However, given the small mean and median plan size reported in Table 1, the magnitude of the marginal effects is amplified. The last two columns of the table report the results for the multinomial estimate. As with the traditional logit estimations, the multinomial estimations confirm the strong findings related to net operating loss, RelativePWT, but also find a strong relationship with plan size. The size of the plan, which is related to the dollar amount of the PWT, is directly related to a cash balance conversion and negatively related to plan termination. This indicates large pension plans are more likely to convert to a cash balance plan than terminate the defined benefit plan altogether. We test the equivalence of the estimated coefficients between the CB conversions and terminations. We find the estimated coefficients on PlanSize and interaction of tax and overfunding are significantly larger in the CB equation than in the DC equation, indicating larger plans and firms likely paying taxes choose to convert traditional plans rather than terminate. The empirical results do not support a financial constraint motivation for changing a DB plan. These results are robust to the method of matched sample construction as Tables A1 and A2 indicate in the appendix. With each matching method, plan size and federal tax coefficients are significantly different for the DC and CB samples.
Table 4 presents the choice between cash balance conversion and termination of the plan and replacing it with a defined contribution given the decision to end a traditional defined benefit plan. Consistent with the univariate results in Table 2, we find taxes a major factor in the decision to terminate rather than convert to a cash balance plan. The role the funding level plays in this decision enters the estimation in several ways, the level of funding, the squared level of funding (to account for nonlinearity), interaction of funding with taxes, and net operating losses. The decision to terminate the plan versus convert to a cash balance plan is not affected solely at the funding level of the plan, but is a complex interaction of funding level with other variables. The tax implications of terminating an overfunded plan are significant. Firms paying federal taxes with overfunded DB plans are less likely to terminate and more likely to convert to a CB plan, avoiding taxes on asset reversions. However, the effect of funding level is nonlinear as demonstrated by the significant and positive coefficients for the square of the funding ratio. (13) Plans in the extremes of funding levels tend to terminate rather than convert, contrary to our hypothesized expectations. We can reject the hypothesis at the 0.05 level that all the funding variables are simultaneously zero. We can also reject the combination of the variables sum to zero at the 0.10 level. Although the funding level and overfunded indicator variables are not significant on their own, overfunded plans are less likely to terminate in the presence of taxes.
In addition to taxes, the other major factor is the size of the potential wealth transfer, with larger PWT firms most likely to convert rather than terminate. The negative sign for Log(PWT) confirms the hypotheses of firms with large future liabilities from the plan are converting to cash balance plans, not terminating the plan altogether. This is a characteristic of large pension plans. This finding is consistent with the difficulty in terminating large plans, but also the goal of modification of the implicit contract to attract employees with a desire for mobility, but it is inconsistent with outright wealth expropriation since the plan assets stay with the plan and does not immediately revert to the firm. There is no evidence supporting the financial constraint hypothesis as OperatingCF and DebtRatio are not significant.
SUMMARY AND CONCLUSIONS
Ending a traditional defined benefit plan is a two-choice process, with the first choice related to the decision to move away from a traditional pension arrangement and the second choice on what the new benefit system will be. In the past, the only alternative was to terminate the defined benefit plan and replace it with a defined contribution plan. However, the advent and growth of cash balance plans creates an alternative. Although many aspects of the new plan are similar, a cash balance plan requires the company to make contributions and maintain plan funding as well as manage plan assets. Defined contribution plans require contributions but shift the majority of the funding and investment requirements to the employee. In light of these differences, we find evidence of a tax motivation as part of the decision in both decision steps. We also find the relative wealth transfer is a major decision to end a traditional defined benefit plan, but the size of the plan and wealth transfer impacts the decision to conversion or termination. Plans with larger potential wealth transfers tend to favor cash balance conversion. We find no evidence to support financial constraints as a major determinate in either decision.
TABLE A1 Logit Estimations of Plan Changes to Multiple Firm Matched Sample Logit Estimation and Marginal Effects All Changes Estimated Marginal Coefficient Effects Intercept -4.805 *** (1.401) Tax*Over -0.276 -0.051 (0.300) (0.056) NOL*Over 1.204 *** 0.267 *** (0.366) (0.088) OperatingCF 0.422 0.079 (1.607) (0.299) FixedAssets 0.766 0.143 (0.535) (0.100) Intang.Assets 0.751 0.140 (0.989) (0.184) DebtRatio 0.830 0.154 (0.544) (0.101) P1anSize 0.060 0.011 (0.064) (0.012) RelativePWT 7.978 *** 1.484 *** (1.627) (0.298) FundedRatio 0.112 0.021 (0.291) (0.054) [FundedRatio.sup.2] -0.027 -0.005 (0.056) (0.010) N 422 Pseudo [R.sup.2] 93 Cash Balance Conversion Estimated Marginal Coefficient Effects Intercept -9.031 *** (2.139) Tax*Over 0.066 0.012 (0.374) (0.068) NOL*Over 1.131 ** 0.248 ** (0.481) (0.116) OperatingCF 1.671 0.307 (2.505) (0.460) FixedAssets 0.381 0.070 (0.675) (0.124) Intang.Assets 0.090 0.017 (1.363) (0.250) DebtRatio 0.594 0.109 (0.753) (0.138) P1anSize 0.277 *** 0.051 *** (0.094) (0.017) RelativePWT 9.264 *** 1.700 *** (2.298) (0.403) FundedRatio -0.406 -0.075 (0.568) (0.104) [FundedRatio.sup.2] 0.123 0.023 (0.465) (0.085) N 275 Pseudo [R.sup.2] 0.097 Termination Define Cont. Estimated Marginal Coefficient Effects Intercept 5.544 * (2.960) Tax*Over -1.374 ** -0.188 ** (0.657) (0.080) NOL*Over 1.386 * 0.276 * (0.722) (0.166) OperatingCF 1.066 0.164 (2.381) (0.367) FixedAssets 0.617 0.095 (1.146) (0.178) Intang.Assets 1.111 0.171 (1.699) (0.263) DebtRatio 0.678 0.105 (0.888) (0.137) P1anSize -0.527 *** -0.081 *** (0.163) (0.024) RelativePWT 6.733 ** 1.039 ** (2.761) (0.440) FundedRatio 0.615 0.095 (0.446) (0.068) [FundedRatio.sup.2] -0.060 -0.009 (0.054) (0.008) N 147 Pseudo [R.sup.2] 0.27 Multinomial Logit Estimation CB Estimated DC Estimated Coefficient Coefficient Intercept -10.742 *** 5.902 ** (1.853) (2.441) Tax*Over 0.071 * -1.781 *** (0.361) (0.599) NOL*Over 1.035 ** 1.689 *** (0.424) (0.604) OperatingCF 1.543 0.785 (2.221) (2.276) FixedAssets 0.576 0.402 (0.626) (1.000) Intang.Assets -0.081 2.046 (1.269) (1.555) DebtRatio 0.709 0.827 (0.668) (0.742) P1anSize 0.352 *** (a) -0.615 *** (0.084) (0.137) RelativePWT 8.080 *** 7.998 *** (1.758) (1.918) FundedRatio 0.256 0.621 (0.487) (0.444) [FundedRatio.sup.2] -0.401 -0.056 (0.414) (0.047) N 422 Pseudo [R.sup.2] 0.209 Notes: Estimates of the logit model (0 if firm does not change DB plan, 1 if firm changes DB plan to a cash balance plan or terminate and adopt a DC plan) and multinomial logit model. Variable definitions are the same as in Table 3. Standard errors are in parentheses. *, **, *** denote significance at the 10%, 5%, and 1% levels, respectively. (a,b) denote a difference in estimated coefficients between the multinomial logit model at the 1% level. TABLE A2 Logit Estimations of Plan Changes to Portfolio Matched Sample Logit Estimation and Marginal Effects All Changes Estimated Marginal Coefficient Effects Intercept -6.403 *** (2.016) Tax *Over -1.067 ** -0.262 ** (0.497) (0.122) NOL *Over 1.295 ** 0.318 ** (0.587) (0.144) OperatingCF 1.567 0.385 (3.254) (0.800) FixeclAssets 3.533 *** 0.869 *** (0.990) (0.241) Intang.Assets 2.385 0.586 (1.829) (0.449) DebtRatio 0.549 0.135 (1.069) (0.263) PIanSize 0.049 0.012 (0.094) (0.023) RelativePWT 21.277 *** 5.207 *** (3.019) (0.723) FundedRatio 0.482 0.119 (0.516) (0.127) [FundedRatio.sup.2] -0.125 -0.031 (0.170) (0.042) N 222 Pseudo [R.sup.2] 0.331 Logit Estimation and Marginal Effects Cash Balance Conversion Estimated Marginal Coefficient Effects Intercept -9.849 *** (2.909) Tax *Over -0.683 -0.171 (0.591) (0.148) NOL *Over 0.865 0.216 (0.774) (0.193) OperatingCF 4.995 1.248 (4.929) (1.232) FixeclAssets 2.584 ** 0.646 ** (1.132) (0.283) Intang.Assets 0.414 0.103 (2.323) (0.580) DebtRatio 0.712 0.178 (1.623) (0.406) PIanSize 0.247 * 0.062 * (0.135) (0.034) RelativePWT 19.550 *** 4.885 *** (3.500) (0.873) FundedRatio -0.878 -0.219 (1.271) (0.318) [FundedRatio.sup.2] 0.872 0.218 (0.784) (0.196) N 146 Pseudo [R.sup.2] 0.278 Logit Estimation and Marginal Effects Termination Define Cont. Estimated Marginal Coefficient Effects Intercept 6.516 (6.390) Tax *Over -3.105 ** -0.586 * (1.475) (0.304) NOL *Over 2.127 0.401 (1.745) (0.360) OperatingCF 4.410 0.832 (6.461) (1.265) FixeclAssets 4.572 0.862 (3.607) (0.594) Intang.Assets 5.638 1.063 (5.169) (0.906) DebtRatio -0.724 -0.137 (1.640) (0.313) PIanSize -0.775 ** -0.146 * (0.381) (0.077) RelativePWT 30.297 *** 5.714 *** (9.343) (1.512) FundedRatio 1.079 0.204 (0.918) (0.180) [FundedRatio.sup.2] -0.248 -0.047 (0.259) (0.050) N 76 Pseudo [R.sup.2] 0.621 Multmomial Logit Estimation CB Estimated DC Estimated Coefficient Coefficient Intercept -11.413 *** 5.327 (2.431) (3.436) Tax *Over -0.7726 -2.813 *** (0.540) (0.916) NOL *Over 0.875 1.729 ** (0.629) (0.829) OperatingCF 4.774 2.642 (3.887) (4.796) FixeclAssets 3.006 *** 3.401 ** (1.026) (1.592) Intang.Assets -0.119 (b) 5.770 ** (2.074) (2.686) DebtRatio 0.123 0.290 (1.083) (1.242) PIanSize 0.308 *** (c) -0.670 *** (0.113) (0.192) RelativePWT 21.130 *** 20.988 *** (3.118) (3.331) FundedRatio 0.574 0.863 (0.628) (0.717) [FundedRatio.sup.2] -0.266 -0.152 (0.342) (0.204) N 222 Pseudo [R.sup.2] 0.406 Notes: Estimates of the logit model (0 if firm does not change DB plan, 1 if firm changes DB plan to a cash balance plan or terminate and adopt a DC plan) and multinomial logit model. Variable definitions are the same as in Table 3. Standard errors are in parentheses. *, **, *** denote significance at the 10%, 5%, and 1% levels, respectively. (a,b,c) denote a difference in estimated coefficients between the multinomial logit model at the 10%, 5%, and 1% levels, respectively.
Alderson, M. J., and K. C. Chen, 1986, Excess Asset Reversions and Shareholder Wealth, Journal of Finance, 41: 225-241.
Alderson, M. J., and J. L. VanDerhei, 1992, Additional Evidence on the Reaction of Shareholders to the Reversion of Surplus Pension Assets, Journal of Risk and Insurance, 59: 262-274.
Allen, E. T., 2003, Pension Planning : Pension, Profit Sharing, and Other Deferred Compensation Plans (Boston, MA: McGraw-Hill/Irwin).
Anand, V., 1999, Other Firms Build on Trend that BankAmerica Initiated, Pensions & Investments, 27: 22-23.
Cather, D. A., E. S. Cooperman, and G. A Wolfe, 1991, Excess Pension Asset Reversions and Corporate Acquisition Activity, Journal of Business Research, 23: 337-348.
Clark, R. L., and A. A. McDermed, 1990, The Choice of Pension Plans in a Changing Regulatory Environment (Washington, DC: The AEI Press).
Clark, R. L., and S. J. Schieber, 2004a, Adopting Cash Balance Pension Plans: Implications and Issues, Journal of Pension Economics and Finance, 3: 271-295.
Clark, R. L., and S. J. Schieber, 2004b, An Empirical Analysis of the Transition to Hybrid Pension Plans in the United States, in: W. Gale, J. Shoven, and M. Warshawsky, eds., Private Pensions and Public Policies (Washington, DC: The Brookings Institution), pp. 11-42.
Coronado, J. L., and P. C. Copeland, 2004, Cash Balance Pension Plan Conversions and the New Economy, Journal of Pension Economics and Finance, 3: 297-314.
Datta, S., M. E. Iskandar-Datta, and E. J. Zychowicz, 1995, Pension Plan Terminations, Excess Asset Reversions and Securityholder Wealth, Journal of Banking and Finance, 19: 245-259.
D'Souza, J., J. Jacob, and B. Logue, 2004, Why Do Firms Convert to Cash Balance Pension Plans? An Empirical Investigation, Working Paper, Cornell University.
Haw, I., W. Ruland, and A. Hamdallah, 1988, Investor Evaluation of Overfunded Pension Plan Terminations, Journal of Financial Research, 11: 81-93.
Hsieh, S. J., and K. R. Ferris, 1994, An Investigation of the Market Effects of Overfunded Pension Plan Termination, Journal of Accounting, Auditing, and Finance, 9: 61-90.
Ippolito, R., 1985, The Labor Contract and True Economic Pension Liabilities, American Economic Review, 75: 1031-1043.
Ippolito, R., 1986, Pensions, Economics and Public Policy Homewood, IL: Dow Jones-Irwin).
McGill, D. M., K. N. Brown, J. J. Haley, and S. J. Schieber, 2005, Fundamentals of Private Pensions Plans, 8th edition (Oxford, New York: Oxford University Press).
Mitchell, M. L., and J. H. Mulherin, 1989, The Stock Price Response to Pension Plan Terminations and the Relation of Terminations With Corporate Takeovers, Financial Management, 18: 41-56.
Mittelstaedt, F. H., and P. R. Regier, 1990, Further Evidence of Excess Asset Reversions and Shareholder Wealth, Journal of Risk and Insurance, 57: 471-486.
Mittelstaedt, F. H., and P. R. Regier, 1993, The Market Response to Pension Plan Terminations, Accounting Review, 68: 1-27.
Niehaus, G., and T. Yu, 2005, Cash-Balance Plan Conversions: Evidence on Excise Taxes and Implicit Contracts, Journal of Risk and Insurance, 72: 321-352.
Petersen, M. A., 1992, Pension Reversions and Worker-Stockholder Wealth Transfers, Quarterly Journal of Economics, 107: 1033-1056.
Pontiff, J., A. Shleifer, and M. S. Weisbach, 1990, Reversions of Excess Pension Assets after Takeovers, Rand Journal of Economics, 21: 600-613.
Rauh, J. D., 2006, Investment and Financing Constraints: Evidence from the Funding of Corporate Pension Plans, Journal of Finance, 61: 33-58.
Stone, M., 1987, A Financing Explanation for Overfunded Pension Plan Terminations, Journal of Accounting Research, 25: 317-326.
Thomas, J. K., 1989, Why Do Firms Terminate Their Overfunded Pension Plans? Journal of Accounting and Economics, 11: 361-398.
VanDerhei, J. L., 1987, The Effect of Voluntary Termination of Overfunded Pension Plans on Shareholder Wealth, Journal of Risk and Insurance, 54: 131-156.
(1) The implicit contract represents the view that the employment contract of an employee is long term in nature, and the compensation the employee receives early in employment is less than the marginal value produced, but compensation (including pension income) later in the career and retirement is more than the marginal product produced.
(2) Plan sponsors will pay 25 percent of excess assets to replacement plan and then a 20 percent excise tax for a total of 45 percent, or a 50 percent excise tax (IRS Ruling 22.214.171.124.9).
(3) When terminating, a firm can choose to replace with several types of defined contribution plans (including 401k) or choose to not offer a retirement plan benefit. For the sample of firms terminating, about half choose not to offer any plan, about a third choose to offer a 401k plan, with the remaining offering a profit sharing or other plan. If the sample is restricted to large plans (over 100 participants), 63 percent offer a 401k plan, 20 percent do not offer a plan, with the remaining offering other types of retirement plans.
(4) If the terminated DB plan is not replaced with a qualified DC plan, the excise tax rate increases to 50 percent. Please see earlier discussion.
(5) This is not the same as the excise tax discussed earlier, but a separate tax on investment gains withdrawn from the plan.
(6) In fact, there were over 3,200 plans with 10 or fewer participants.
(7) The PBGC and other federal agencies report pension plan information by the EIN. EINs are assigned to each business, including both subsidiaries and parent firms. Compustat also reports the EIN for the parent corporation. We match the plan EIN to the one reported in Compustat. However, there may be plans reported under a subsidiary's EIN that do not match to the one reported in Compustat, even though it is part of a publicly traded firm.
(8) Form 5500 is filed annually for each pension plan with the Department of Labor and IRS. It contains detailed information on number of participants and status, plan assets, and plan liabilities.
(9) One firm changed two different plans at different time periods. Since this represented two distinct decisions, we have 123 unique change decisions for 122 unique firms.
(10) Some firms have less than four matched sample firms, but the average is three matches per firm.
(11) Computing the potential wealth transfer requires an estimation of the explicit and economic liabilities resulting from the pension plan for both active and retired employees using a nonlinear relationship. See either Ippolito (1996) or Mittelstaedt and Regier (1993) for a more in-depth description of estimation procedures.
(12) Current liabilities on Form 5500 are the accrued liabilities earned at the moment of filing and the amount required to be funded under IRS/PBGC regulation. This amount is not the same as the projected benefit obligation (PBO) reported in the financial statements.
(13) The interaction of taxes and funding levels squared is not significant in our model specifications.
Joel T. Harper is an Associate Professor of Finance and Williams Companies Professor of Business, Oklahoma State University. Stephen D. Treanor is an Assistant Professor of Finance, California State University. Joel T. Harper can be contacted via e-mail: firstname.lastname@example.org. The authors would like to thank Frank de Jong and participants at the 2007 Southern Finance Association meeting, NetSPAR symposium, University of Arkansas, and Oklahoma State University for valuable comments.
TABLE 1 Descriptive Statistics Pension Change (N = 112) Standard Mean Median Deviation Size and wealth transfer Firm Assets 19,516.2 3,988.9 50,346.5 Pension Liabilities 603.5 78.9 1,166.1 Potential Wealth Transfer 116.8 19.7 210.9 (PWT) PWT to Pen. Liabilities 25.9% 24.0% 18.6% Tax and operating variables NOL 0.259 0 0.440 Federal Tax 0.598 1 0.492 Fixed Assets (pct.) 32.1% 28.7% 24.7% Intangible Asset (pct.) 8.2% 2.1% 12.5% Debt Ratio 72.0% 70.2% 26.0% Operating Cash (pct.) 7.3%, 7.8% 7.4% Pension funding Funding Level 0.187 0.162 0.478 Abs (funding level) 0.345 0.228 0.379 Squared Funding Level 0.261 0.052 0.629 Overfunded 0.759 1.000 0.430 Matched Sample (N = 310) Standard Mean Median Deviation Size and wealth transfer Firm Assets 12,560.3 2,509.6 29,978.9 Pension Liabilities 491.7 108.1 1,149.6 Potential Wealth Transfer 71.8 19.1 123.7 (PWT) PWT to Pen. Liabilities 19.5% 19.3% 8.6% Tax and operating variables NOL 0.148 0 0.356 Federal Tax 0.674 1 0.469 Fixed Assets (pct.) 33.1% 29.8% 24.5% Intangible Asset (pct.) 8.0% 1.4% 12.3% Debt Ratio 66.6% 66.5% 22.9% Operating Cash (pct.) 8.0% 8.3% 8.0% Pension funding Funding Level 0.257 0.168 0.954 Abs (funding level) 0.400 0.223 0.903 Squared Funding Level 0.973 0.0499 12.306 Overfunded 0.739 1.000 0.440 Notes: Descriptive statistics for sample firms either terminating or converting a defined benefit plan with at least one (and up to four) firm matches. Matched sample firms are based on firm size and industry and include several firms matched to a single sample firm. Firm Assets is total assets for the firm and is a proxy for size. Pension Liabilities is liabilities reported on Form 5500 and reflects the current liability owed. PWT is the log of the firm's potential wealth transfer calculated using Ippolito's methodology. PWT to Pension Liabilities reflects the wealth transfer (implicit liabilities) to the current liability owed. Federal Tax is a dummy variable indicating the firm paid federal tax. NOL is a dummy variable indicating a net operating loss carry forward. Avg. OCF is the average firm's averaging cash flow to assets, and Fixed Asset is the firm's fixed assets as a percentage of total assets, and Intangible Asset is the percentage of intangible assets to firm's total assets. Debt Ratio is a proxy for financial leverage. Funded Ratio is the average of the actuarial plan assets divided by liabilities minus 1. Overfunded is an indicator variable taking the value of 1 if the plan is overfunded. TABLE 2 Univariate Tests of Cash Balance Converted and Termination Plans Converted to Cash Balance (N = 80) Standard Mean Median Deviation Size and wealth transfer Firm Assets 27,353.1 8,712.7 58,473.4 Pension Liabilities 1,677.6 468.4 4,448.9 Pot. Wealth Trans. (PWT) 249.8 83.4 527.4 PWT to Pen. Liabilities 22.4% 21.4% 7.5% Tax and operating variables NOL 0.213 0 0.412 Federal tax 0.688 1 0.466 Fixed Assets (pct.) 35.4% 32.3% 24.5% Intangible Asset (pct.) 6.3% 0.8% 12.0% Debt Ratio 70.7% 68.1% 17.5% Operating Cash (pct.) 8.7% 8.3% 5.5% Pension funding Funding Level 0.244 0.224 0.354 Squared Funding Level 0.183 0.059 0.342 Overfunded 0.838 1.000 0.371 Changed to Defined Contribution (N = 42) Diff. in Standard Means Test Mean Median Deviation Statistic Size and wealth transfer Firm Assets 7,179.5 1,010.9 17,838.5 2.204 ** Pension Liabilities 32.9 9.2 63.3 2.420 ** Pot. Wealth Trans. (PWT) 8.5 2.0 16.4 2.993 *** PWT to Pen. Liabilities 31.2% 29.8% 28.0% -2.643 *** Tax and operating variables NOL 0.372 0 0.489 -1.918 * Federal tax 0.419 0 0.500 2.975 ** Fixed Assets (pct.) 25.9% 20.4% 22.8% 2.099 ** Intangible Asset (pct.) 11.9% 7.2% 12.7% -2.404 ** Debt Ratio 75.5% 71.9% 37.1% -0.986 Operating Cash (pct.) 4.7% 4.9% 9.1% 3.040 *** Pension funding Funding Level 0.006 0.026 0.650 2.637 *** Squared Funding Level 0.412 0.050 0.905 -2.016 ** Overfunded 0.535 1.000 0.505 3.789 *** *, **, *** denote significance at the 10%, 5%, and 1% levels, respectively. TABLE 3 Logit Estimations of Plan Changes to Single-Firm Matched Sample Logit Estimation and Marginal Effects All Changes Estimated Marginal Coefficient Effects Intercept -4.117 ** (1.717) Tax*Over -0.646 -0.160 (0.399) (0.097) NOL*Over 1.508 *** 0.326 *** (0.533) (0.089) OperatingCF 1.037 0.258 (2.624) (0.654) FixeclAssets 1.814 ** 0.452 ** (0.740) (0.184) IntangAssets 1.450 0.361 (1.301) (0.324) DebtRatio 1.184 0.295 (0.951) (0.237) P1anSize 0.004 0.001 (0.078) (0.019) RelativePWT 12.189 *** 3.037 *** (2.342) (0.581) FundedRatio 0.571 0.142 (0.416) (0.104) [FundedRatio.sup.2] -0.317 -0.079 (0.244) (0.061) N 222 Pseudo [R.sup.2] 0.167 Cash Balance Conversion Estimated Marginal Coefficient Effects Intercept -5.915 ** (2.531) Tax*Over -0.424 -0.106 (0.489) (0.121) NOL*Over 1.609 ** 0.352 *** (0.742) (0.122) OperatingCF 2.129 0.523 (3.888) (0.972) FixeclAssets 1.900 ** 0.475 ** (0.958) (0.239) IntangAssets 0.723 0.181 (1.671) (0.418) DebtRatio 2.166 0.541 (1.362) (0.341) P1anSize 0.086 0.021 (0.109) (0.027) RelativePWT 9.534 *** 2.383 *** (2.918) (0.729) FundedRatio 0.615 0.154 (0.619) (0.155) [FundedRatio.sup.2] -0.768 -0.192 (0.575) (0.144) N 146 Pseudo [R.sup.2] 0.122 Termination/ Define Cont. Estimated Marginal Coefficient Effects Intercept 3.496 (4.367) Tax*Over -1.467 -0.348 (1.011) (0.225) NOL*Over 1.928 0.350 ** (1.245) (0.156) OperatingCF 1.105 0.262 (5.101) (1.210) FixeclAssets 2.039 0.483 (1.883) (0.442) IntangAssets 2.442 0.579 (2.596) (0.612) DebtRatio -0.015 -0.004 (1.409) (0.334) P1anSize -0.488 * -0.116 * (0.254) (0.061) RelativePWT 18.683 *** 4.430 *** (5.636) (1.230) FundedRatio 0.834 0.198 (0.752) (0.176) [FundedRatio.sup.2] -0.273 -0.065 (0.380) (0.089) N 76 Pseudo [R.sup.2] 0.431 Multinomial Estimation Logit DC Estimated CB Estimated Coefficient Coefficient Intercept -8.755 *** 4.768 * (2.096) (2.806) Tax*Over -0.356 * -1.834 ** (0.445) (0.735) NOL*Over 1.259 ** 1.914 ** (0.580) (0.750) OperatingCF 2.618 0.568 (3.159) (3.945) FixeclAssets 1.607 ** 1.348 (0.796) (1.268) IntangAssets 0.228 3.208 * (1.506) (1.906) DebtRatio 0.987 1.139 (1.017) (1.220) P1anSize 0.239 *** -0.568 *** (0.095) (0.150) RelativePWT 11.712 *** 12.484 (2.458) (2.788) FundedRatio 0.699 0.927 (0.534) (0.585) [FundedRatio.sup.2] -0.710 -0.209 (0.442) (0.295) N 222 Pseudo [R.sup.2] 0.277 Notes: Estimates of the logit model (0 if firm does not change DB plan, 1 if firm changes DB plan to a cash balance plan or terminate and adopt a DC plan) and multinomial logit model. Tax * Over is a dummy variable indicating the firm paid federal tax and the plan is overfunded, NOL * Over is a dummy variable indicating a net operating loss carryforward and the plan is overfunded. OperatingCF is the average firm's averaging cash flow to assets and FixeclAsset is the firm's fixed assets as a percentage of total assets, and IntangibleAsset is the percentage of intangible assets to firm's total assets. DebtRatio is a proxy for financial leverage. PWT is the log of the firm's potential wealth transfer calculated using Ippolito's methodology. Funded Ratio is the average actuarial plan assets divided by liabilities minus one. PIanSize is the log of the pension liabilities. Standard errors are in parentheses. *, **, *** denote significance at the 10%, 5%, and 1% levels, respectively. (a,b,c) denote a difference in estimated coefficients between the multinomial logit model at the 10%, 5%, and 1% levels, respectively. TABLE 4 Logit Estimates of Plan Terminations Versus Cash Balance Conversions Estimated Marginal Estimated Coefficient Effects Coefficient Intercept 2.635 2.770 (2.346) (2.341) OperatingCF -1.163 -0.155 -1.264 (6.237) (0.837) (6.260) FixedAssets 1.559 0.208 1.580 (1.592) (0.214) (1.613) Intang.Assets 2.763 0.369 2.757 (2.786) (0.392) (2.785) DebtRatio 0.620 0.083 0.605 (1.282) (0.174) (1.290) FirmSize -0.218 -0.029 -0.225 (0.183) (0.025) (0.181) Log(PWT) -0.103 *** -0.014 *** -0.104 *** (0.025) (0.003) (0.025) RelativePWT 1.755 0.234 1.773 (2.476) (0.340) (2.516) FundedRatio -0.244 -0.032 (0.720) (0.098) Overfunded -0.070 (0.812) Tax *Over -2.837 *** -0.363 *** -2.897 *** (1.023) (0.113) (1.090) NOL *Over 0.613 0.093 0.608 (0.920) (0.158) (0.935) [FundedRatio.sup.2] 1.050 ** 0.140 * 0.971 ** (0.526) (0.078) (0.477) N 123 123 Pseudo [R.sup.2] 0.583 0.583 Marginal Estimated Marginal Effects Coefficient Effects Intercept 2.475 (2.445) OperatingCF -0.167 -1.074 -0.143 (0.833) (6.216) (0.831) FixedAssets 0.209 1.624 0.216 (0.214) (1.631) (0.218) Intang.Assets 0.364 2.758 0.367 (0.388) (2.769) (0.388) DebtRatio 0.080 0.595 0.079 (0.174) (1.280) (0.173) FirmSize -0.030 -0.213 -0.028 (0.024) (0.184) (0.025) Log(PWT) -0.014 *** -0.104 *** -0.014 (0.003) (0.025) (0.003) RelativePWT 0.234 1.721 0.229 (0.343) (2.495) (0.341) FundedRatio -0.399 -0.053 (1.017) (0.137) Overfunded -0.009 0.251 0.032 (0.110) (1.154) (0.141) Tax*Over -0.368 *** -2.920 *** -0.373 *** (0.120) (1.095) (0.121) NOL*Over 0.092 0.573 0.086 (0.160) (0.938) (0.159) [FundedRatio.sup.2] 0.128 * 1.106 * 0.147 * (0.067) (0.586) (0.084) N 123 Pseudo [R.sup.2] 0.584 Notes: Dependent variable is 1 if the firm terminates plan and replaces it with a defined contribution plan, and 0 if the firm converts to a cash balance plan. Variables are the same as described in Table 3. Size is proxied by the log of firm's assets. RelativePWT is the estimated potential wealth transfer divided by the pension liabilities. FundedRatio is the current funded ratio at the time of termination or conversion, I FundedRatio is the absolute value of the funded ratio, and Overfunded is an indicator variable taking the value of 1 if the plan is overfunded. Standard errors are in parentheses. *, **, *** denote significance at the 10%, 5%,, and 1% levels, respectively.
|Printer friendly Cite/link Email Feedback|
|Author:||Harper, Joel T.; Treanor, Stephen D.|
|Publication:||Journal of Risk and Insurance|
|Date:||Mar 1, 2014|
|Previous Article:||On the use of information in oligopolistic insurance markets.|
|Next Article:||Optimum hurricane futures hedge in a warming environment: a risk-return jump-diffusion approach.|