Why do firms offer risky defined-benefit pension plans?

INTRODUCTION

In the aggregate, private defined-benefit (DB) pension plans in the United States are underfunded by a considerable margin. The Pension Benefit Guaranty Corporation (PBGC), the federal insurer of such plans, puts the aggregate shortfall of assets from liabilities among insured single-employer plans at about $350 billion as of September 30, 2006 (PBGC 2006, p. 8). In addition, the characteristics of the assets that these plans hold are very different from the characteristics of their liabilities. Coronado and Liang (2006) find that the typical pension trust holds 60 percent to 70 percent of its value in equities. In contrast, by at least one definition, liability is fixed and known with certainty in nominal terms. (1) These two factors--the substantial shortfall of assets from liabilities and the mismatch between assets and liabilities--imply that private DB plans are a risky proposition for both workers and firms. (2)

As Bodie (1990) and others have pointed out, even risky firms could greatly reduce the risk of their pension promises by contributing a sufficient level of resources to their pension trust funds and by investing those resources in fixed-income securities designed to deliver their payoffs just as pension obligations are coming due. This strategy would immunize the pension fund from market fluctuations because stock returns would be irrelevant and interest-rate changes would affect pension assets (through bond values) and liabilities (through the present value of future obligations) at the same time and by the same amounts. (3) Despite this possibility, almost all firms choose to make their pension promises risky. (4) This leads to the question motivating this paper: Is it optimal for firms to introduce more than the minimum amount of risk into their pension promises? In particular, can firms really gain value by making their pension promises risky rather than free of risk? In this paper, we develop a model of pension financing in which the total compensation offered to workers must clear the labor market, an approach that formalizes some of the ideas suggested by Bulow (1982) and Bodie (1988, 1990, inter alia). The firms in our model maximize their value by making their pension promises free of risk.

Many readers will find this conclusion to be highly counterintuitive. How could a firm improve its financial position by offering a riskless pension promise rather than a risky one? Are risky promises not discounted at higher rates of return than riskless ones, ensuring that the present-discounted value (PDV) of a riskless promise exceeds the PDV of a risky one? By making the pension promise riskless, will firms, therefore, not be taking on additional cost?

The answer is yes, firms that eliminate the risk from previously risky pension promises will take on additional pension cost. But minimizing pension cost alone is not the appropriate objective for firms to pursue: Firms should aim to minimize the market value of total compensation cost, not pension cost in isolation (holding the real activity of the firm constant). We show that if workers understand the implications of pension risk, they will demand greater compensation for riskier pension promises than for safer ones, all else being equal. Thus, while riskier pension promises may reduce pension cost, they do not reduce the total compensation cost of the firms in our model.

One way to see the economic intuition for this result is to think of workers as disadvantaged bondholders of the firm--"bondholders" because they hold promises of future payments just as ordinary bondholders do, but "disadvantaged" because, unlike ordinary bondholders, workers are assumed to be unable to diversify away the company-specific risk to which they are exposed through risky pension promises. In return for the promise of a risky DB pension, workers offer their employers a lower "price" (in the form of wage concessions) than the employers could obtain from bond-market investors in return for the same promised cash flows. Thus, we generate a violation of the standard Modigliani-Miller result that purely financial operations do not affect the value of the firm: Firms that offer a risky pension plan are committing a form of financial inefficiency because they are obtaining part of their financing on worse-than-market terms. The more they avail themselves of this source of financing, the more they reduce their own value. Taking the risk out of the pension promise eliminates the financial inefficiency because in that case--and that case only--workers are willing to provide financing on terms that are as good as the ones that the firm could obtain from financial markets.

This paper is only a first step in making the argument because, for the sake of transparency and tractability, we suppress a number of important features of the pension landscape, some of which would reinforce the argument in favor of making the pension promise riskless, and some of which would weaken it. One of these factors is the PBGC, which offers insurance against downside risk, and--in return--charges premiums far below the economically fair level. Another factor that we suppress is the tax code, and its associated influence on portfolio allocation. A third factor is the empirical regularity, noted by Bodie (1990), that workers seem to hold a call option on part of the surplus in pension trust funds. The way for firms to minimize the market value of this call option is to eliminate the upside potential associated with the trust fund; this, in turn is accomplished by matching the characteristics of the assets in the trust fund to the characteristics of the plan's liabilities. So, again, this consideration strengthens the argument for making the pension promise free of risk.

In work in progress, we are extending the model to capture at least some of the factors omitted here, and studying the implications of allowing the firm to invest part of the pension trust in equities, the second of the two mechanisms available to firms for increasing the risk of the pension promise.

This paper proceeds as follows. In the next section, we briefly review the literature that has considered whether firms should offer a risky pension promise or one that is free of risk. Then we present a bare-bones model with a consumer that lives for two periods, earning cash wages when young and a pension when old. We find that, in this model, the firm's optimal strategy is to fund the pension promise fully. We also begin the process of exploring the robustness of this result to variations in model specification and calibration. We close by revisiting the question of why no real-world firm follows the prescription of our model.

RELATED LITERATURE

Early work on pension funding recognized that pension obligations represent contingent liabilities for shareholders and contingent assets for employees, where the contingency is the solvency of the firm. Sharpe (1976), Black (1980), and Treynor (1977) demonstrate that this structure implies that a firm's pension obligation can be analyzed in an options-pricing framework: Firms effectively own an option that entitles them to put the funding gap to their employees if and when the firms become insolvent. (5) The value of the pension obligation in an economic balance sheet therefore consists of promised benefits (a liability) and a put option on the difference between promised benefits and the value of the trust fund (an asset).

Using standard options theory, Treynor (1977) shows that increasing the risk of the underlying assets increases the value of the put option and therefore--holding other forms of compensation constant--the value of shareholder equity. As a result, again ignoring any changes in other components of compensation, the put option provides an explanation for why corporations might prefer that pension fund managers invest in risky assets. Sharpe (1976) demonstrates that if workers have access to perfect capital markets and if there is no firm-specific risk, workers can offset the pension funding decisions of the firm so long as the firm is guaranteed to survive. In this case, the funding decision is irrelevant: A firm can either fully fund its pension, or it can underfund it and pay employees an additional amount equal to the put value of the shortfall.

Harrison and Sharpe (1983) analyze pension asset allocation and funding decisions in the presence of potentially mispriced pension insurance. Following Tepper (1981), they allow tax considerations to affect the investment of pension assets. Abstracting from mispriced pension insurance, they confirm Tepper's result that firms will optimally invest pension assets (which are tax-exempt) entirely in the higher-taxed assets (bonds). They demonstrate, however, that presence of mispriced pension insurance can overturn this result. In particular, if equities are a more effective vehicle for taking advantage of naive insurance pricing, the optimal pension strategy may involve a combination of equities and debt.