THE PENSION FORUM - SOA · booklet of The Pension Forum for personal, internal, classroom or other...

FORUMTHE PENSION

published by the Pension Section of the Society of Actuaries

April 2005Volume 16, No. 2

PFN0504.qxd 5/3/05 8:40 AM Page a

The Pension Forum is published on an ad-hoc basis by the Pension Section of the Society ofActuaries. The procedure for submitting an article appears on page 63.

The Pension Forum is sent without charge to all members of the Pension Section.

All correspondence should be addressed to: Arthur J. AssantesEditor, Pension Section News / The Pension ForumThe Pension SectionSociety of Actuaries475 N. Martingale Road, Suite 600Schaumburg, IL 60173

Expressions of Opinion

The Society of Actuaries assumes no responsibility for statements made or opinions expressed inthe stories. Expressions of opinion are those of the writers and, unless expressly stated to the con-trary, are not the opinion or position of the Society of Actuaries or the Pension Section.

Comments on any of the papers in this Forum are welcomed. Please submit them to Art Assantes,editor. They will be published in a future issue of the Pension Section News.

Printed in the United States of America

Copyright 2005 © Society of Actuaries

All rights reserved by the Society of Actuaries. Permission is granted to make brief excerpts for apublished review. Permission is also granted to make limited numbers of copies of items in thisbooklet of The Pension Forum for personal, internal, classroom or other instructional use, on con-dition that the forgoing copyright notice is used so as to give reasonable notice of the Society’scopyright. This consent for free limited copying without prior consent of the Society does notextend to making copies for general distribution, for advertising or promotional purposes, forinclusion in new collective works or for resale.

PFN0504.qxd 5/3/05 8:40 AM Page c

THE PENSION

FORUMVolume 16, Number 2 April 2005

Table of Contents PAGE

Pension Deficits: An Unnecessary Evil 1by Lawrence N. Bader

Comments on “Pension Deficits: An Unnecessary Evil” 10by Eric Klieber

A Critique of “Pension Deficits: An Unnecessary Evil” 11by Dimitry Mindlin

Author’s Response to Mr. Klieber’s and Mr. Mindlin’s Comments 14by Lawrence N. Bader

Fixing the Pension Plan Funding Rules 19by Edward E.Burrows

Reaffirming Pension Actuarial Science 30by Dimitry Mindlin

A Critique of “Reaffirming Pension Actuarial Science” 43by Lawrence N. Bader

Comments on “Reaffirming Pension Actuarial Science” 47by Tony Day

Comments on “Reaffirming Pension Actuarial Science”by Robert McCrory 49

Author’s Response to Mr. Bader’s Commentsby Dimitry Mindlin 50

Author’s Response to Mr. Day’s Commentsby Dimitry Mindlin 58

Author’s Response to Mr. McCrory’s Commentsby Dimitry Mindlin 62

PFN0504.qxd 5/3/05 8:40 AM Page e

THE PENSION FORUM

1

Pension Deficits: An Unnecessary Evilby Lawrence N. Bader, F.S.A.

© Copyright 2004, CFA Institute. Reproduced and published from Financial Analysts Journal with permission from the CFA. All rights reserved.

Falling equity markets and interest rates have devastated pension plans worldwide during the past several years. The S&P 500 Index companies enjoyed a collective pension surplus of $250 billion in 1999. Even after the 2003 market rally, they face a deficit of $168 billion in 2003 (Bianco, Deng, and Suri 2004). These circumstances spotlight the weakness of currentfunding and investment practices for corporatedefined benefit pension plans. This article presents a case for securing all accrued benefitsthrough full funding.1

“Full funding” is commonly understood to meanthat assets are sufficient to cover liabilities meas-ured at an arbitrary discount rate, with no consid-eration of how the assets are invested. Here, I use“full funding” to signify a much stronger fundingcondition, one in which an immunizing bondportfolio secures all benefits to which employeeswould be entitled upon service termination. Thecombination of sufficient assets and an immuniza-tion strategy eliminates dependence on the credit-worthiness of the pension sponsor. Furthermore,the sponsor commits not to undermine that secu-rity by changes in investment or funding policy,by plan amendments that are not immediatelyfunded, or by plan mergers or spin-offs.

I discuss pension funding initially in the absenceof governmental guarantees because most coun-tries lack guarantees and because this approachyields insights that are useful in evaluating guar-antee programs.

Preregulatory Environment

The setting for this discussion is a transparentfinancial system in which plan sponsors,investors, creditors, and employees fully under-stand the value and risk of pension plans. In thistransparent system,

• capital providers understand that a dollar owedto a pensioner and a dollar owed to a creditorhave the same (tax-adjusted) effects on corpo-rate value and

• employees understand the risks of both under-funding and asset/liability mismatches. They cor-rectly value their pensions and are able to makerational trade-offs between pensions and salary.

These assumptions are heroic. But we cannot basean optimal pension system on the behavior ofstakeholders who view pension plans onlythrough a veil of ignorance.

The simple preregulatory environment has notaxes, no regulation, and no governmental guar-antee of pension promises. Later in the discus-sion, I introduce these factors.

A Simple Pension Promise. Suppose an employee’s compensation for a year includes a salaryand a promise of a $20,000 lump sum payable in 25years. The lump sum is vested and payable whetheror not the employee is alive at the due date.2 Thispension promise is economically equivalent to theemployer’s issuing its own nontransferable bond tothe employee as part of his pay package.

1 This article draws substantially on the thinking of Sharpe (1976), Black (1980), and Tepper (1981).2 I assume full vesting throughout this article. Unvested benefits—a small percentage of the liability of most plans—raise

several issues beyond the scope of the discussion. Also, the article considers only hedgeable, bondlike accrued pensions, noteconomically uncertain projected pensions. Projected pensions are not a true corporate liability (Bader 2003b).

PFN0504.qxd 5/3/05 8:40 AM Page 1

THE PENSION FORUM

2

Suppose this nontransferable bond is fully collat-eralized by a portfolio of matching risk-freebonds. In this case, the employer’s bond itself isrisk free and would be valued at riskless rates bythe market and the employee. But suppose thecollateral is too small or too risky and there is adanger that the company might default. In thiscase, the employee discounts the bond for itsdefault risk.

Nondiversifiable Risk. If the plan sponsor issuedsuch a bond publicly, investors would treat it likeany other similarly risky bond in their diversifiedportfolios. But for employees, the risk of theemployer’s bond is different from that of othercompanies’ bonds. The employer bond adds tothe large employer-specific risk that the employ-ees already bear through their employment, andthe employees cannot diversify or hedge this riskin any practical manner.3

If a company were to sell its own risky bonds toits own employees, therefore, the company wouldbe selling to unwilling buyers. Unlike theinvestors who determine market prices, employ-ees cannot diversify the company-specific risk towhich they are already overexposed, so theywould not pay the full market price. Nor would itbe rational for them to give up enough salary tocover the full market value of the risky pension.

A company might still, despite this inefficiency,wish to provide pension plans. Such plans mighthelp manage retirement patterns and assureretirees a decent standard of living. Also, societyencourages pension plans through tax subsidies,which can close the gap between company costand employee valuation of their pensions. But cancompanies improve the value of pensions toemployees without commensurate cost?

Full Funding of Accrued Benefits. Companiescan accomplish such an improvement by securingpension promises through full funding. As noted,any employer-specific risk in a pension fund makesthe pension inefficient because its cost to theemployer is greater than its value to employees. Fullfunding eliminates the risk that can arise from pen-sion assets that are either too small or too risky.

If the risk is from pension assets that are toosmall, the company should borrow in the capitalmarkets from willing lenders to “refinance” itsinefficient “debt” to its employees. The companyis better off borrowing from investors who candiversify than from employees who cannot.

If the risk arises from aggressive investing, thecompany can shift to an immunizing bond port-folio. Exchanging one class of marketable assetsfor another creates no first-order change in share-holder value, but the company gains by raisingthe value that employees attach to their pensionsand, therefore, the salary that they will sacrificefor those pensions.

Tax Arbitrage. Companies can also gain from full funding by saving taxes for their share-holders. Like a number of other countries, theUnited States taxes bonds at a higher rate thanequities and gives favorable tax treatment to pension funding. Under these conditions, Black(1980) and Tepper (1981) showed that it is tax effi-cient to fully fund pension plans, invest the pen-sion fund in bonds, and shift equity risk to theshareholders’ own portfolios or elsewhere in the company.

The arguments about employee risk and tax arbitrage do not demean equity investment. They merely redirect the equity investment awayfrom pension plans so that it will not subject-

3 Although a short position in the company’s debt offers a theoretical (and very approximate) hedge for the pension promise, such a strategy would be costly or impossible for rank-and-file employees and would be frowned on or forbidden for management-level employees.

PFN0504.qxd 5/3/05 8:40 AM Page 2

THE PENSION FORUM

3

shareholders to unnecessary taxes and employeesto nondiversifiable dependence on their employ-ers’ creditworthiness.

A Note on Immunization. The argument so faris that eliminating market risk is more valuable toemployees than costly to sponsors. This argumentweakens, however, for the final increment of riskreduction—that is, replacing the highest-qualitycorporate bond portfolio with U.S. Treasuries. Inthis replacement, sponsors pay for the stateincome tax exemptions and high liquidity ofTreasuries. These qualities are unimportant topension funds and may make reducing pensionrisk to “absolute zero” overly expensive.

Unfortunately, no riskless securities exist that do not have these costly—but in this context,useless—properties. Therefore, this potentialfinal improvement in pension security may notjustify the cost of squeezing out the last bit ofdefault risk.

The shortcomings of Treasury immunization donot, however, make corporate bonds the correctmeasurement standard. Only government bondsoffer a risk-free, objective, and hedgeable stan-dard.4 But in practical situations, an imperfectimmunization—one that relies on bonds that arevery high quality but not riskless—may offer theoptimal balance of cost and security. The sponsorof an imperfectly immunized plan should main-tain sufficient assets to meet a Treasury-basedstandard at all times by slight overfunding inanticipation of possible losses.

Funding under a Guarantee System

Now consider how Pension Benefit GuarantyCorporation guarantees change the desirability of

funding.5 The PBGC is financed by premiumspaid by plan sponsors to insure each other’s pen-sion plans. Thus, we may refer to the PBGC asthe “OPSGC”—the “other plan sponsors’ (OPS)guaranty corporation”—to remind us that thecost of one sponsor’s pension plan failure is borneby other plan sponsors, not by some outsideparty. The law provides no taxpayer money, soeconomically, the other plan sponsors are theguarantors and the PBGC is only an administra-tor and collection agency.

The PBGC guarantees most, although not all,corporate defined-benefit pensions. These guar-antees undercut the major advantage of fundingin the unregulated system described previously. APBGC-guaranteed pension is secure with or with-out company funding, and employees with suchguaranteed pensions have no company-specificrisk to worry about.

By fully funding a pension on which it mighthave defaulted and forced the PBGC to pay, thecompany transfers value to the PBGC withoutbenefit to its own employees. In the absence oflegal funding requirements, each sponsor’s narrowinterest is thus to fund as little as possible. At thesame time, each sponsor wants all other plans tobe well funded so that it will not have to pay fortheir failures. In game theory terms, this situationis a “prisoner’s dilemma.”

As the guarantee system shifts risk from employ-ees to the OPS, legislation becomes necessary toprevent each sponsor’s pursuit of selfinterest fromproducing the worst result for all sponsors. Acompulsory guarantee system, if combined withpermissive funding and investment standards, canenable weak companies to drag down and preyupon strong ones. So, beneath the veneer of an

4 I have argued elsewhere (Bader 2003b) that the valuation of corporate plan sponsors’ pension obligations, like the valuation of their debt, should reflect credit risk (after factoring in the security provided by any pension assets). The current article, however, addresses optimal funding policy, which should aspire to eliminate, rather than reflect, risk.

5 Although I refer to the PBGC specifically, this analysis also applies to other governmental guarantee systems, such as those in Ontario (Canada), Germany, and the proposed U.K. Pension Protection Fund.

PFN0504.qxd 5/3/05 8:40 AM Page 3

THE PENSION FORUM

4

insurance operation, the PBGC serves primarilyto extract capital from successful companies topay the obligations of unsuccessful ones.

For example, suppose a failing company cannotpay competitive salaries. It may be able to solvethat problem by promising outsized pensions andfunding them inadequately. The guarantees givethe full value of the pensions to the employees,and the company gets to use in its business themoney that should go toward employee compen-sation. In this sense, the OPS involuntarily pro-vides a loan guarantee to the failing company andthe company gets full value for its pension prom-ise from its employees, value that it could not getfrom its employees or from the capital markets fora similar promise without the guarantee.

Two broad legislative solutions are available:

1. The government can require full funding,thereby preventing plan sponsors from taking risks that are borne by others.

2. The government can charge each plan sponsor a premium that accurately reflectsthe risks that the sponsor imposes on the system.6

The second solution is appealing because of thefreedom it gives sponsors to manage their plans.But charging true risk-based premiums would putthe PBGC in a uniquely difficult position amongthe government regulators of financial intermedi-aries. Think how closely we regulate banks, insur-ance companies, and brokerage firms. Thesefinancial intermediaries must have assets thatcover their liabilities and maintain a reasonablematch in risks between assets and liabilities. Ifsimilar standards were applied to pension plans,the PBGC could limit its regulatory focus to the

plans themselves. But suppose pension plans werenot held to the standards governing other finan-cial intermediaries, so they remained dependenton their sponsors’ financial health. Then, thePBGC would have to extend its regulatory reachto evaluating and monitoring the operations ofevery sponsor of an underfunded plan. This rolewould be daunting for a government agencywhose mission is simply to insure pensions.

A final and critical problem with permissive fund-ing and investment rules is that the risks borne bythe PBGC are not diversified. The vast majorityof sponsors are taking the same risk—betting onequities instead of hedging their pension liabilitieswith bonds. A severe and prolonged decline instock prices can thus trigger an assessment spiralamong plan sponsors and, eventually, a taxpayerbailout of the PBGC.

So, mandatory full funding, not risk-based premi-ums, is the only practical prevention for the dis-eases that can afflict a guarantee system.7 A work-able, equitable, and financially sound guaranteesystem would have the following characteristics:

• The guarantee agency would function mainly as a monitor and enforcer rather thanas a claims-paying insurer.

• The failures that it covered would be raremisfortunes rather than inevitable outcomesof widespread risky practices.

• Pension plans would be fully funded withrespect to the benefits that would be dueupon plan termination.

• Plans would remain fully funded at all times,without the need for extended periods or fullmarket cycles to correct deficiencies.

• Plans would not take on new liabilities with-out sufficient assets to cover them.

6 See Bodie and Merton (1992). Currently, PBGC premiums are modestly risk related; they include a charge of 0.9 percent of theunfunded liability. The premiums are not equitably risk based because they do not reflect the investment policy or strength of thesponsor.

7 Bodie (1996) discussed this problem in similar terms, but he suggested another possible solution: replacing the PBGC with private-sector guarantees that rely on the risk management products developed since the PBGC was founded.

PFN0504.qxd 5/3/05 8:40 AM Page 4

THE PENSION FORUM

5

Questions and Objections

I have argued here that nonguaranteed pensionsshould be voluntarily fully funded in a transpar-ent but unregulated pension system and that asound government guarantee system must man-date full funding. In this section, I consider somequestions and objections concerning full funding.

1. I suggested that companies with underfundedplans should borrow money to fund theirdeficits. But companies may object that debt is alimited resource. Alternative uses for borrowedfunds must compete with each other, and com-panies should have far better uses for debt thanbuying bonds for their pension funds.

Borrowing to fund a pension deficit does not use scarce capital; it simply refinances or restruc-tures liabilities. Pension deficits affect corporatevalue in the same way that debt does. By borrowing and funding, the company replacesinefficient and expensive pension debt with conventional debt. The restructuring leaves itsnet liabilities unchanged and its borrowingcapacity undiminished.

A company eager to borrow for an attractive cap-ital investment would gain, not lose, by first refi-nancing inefficient or expensive debt. The debtmay be an old loan that can be replaced at a lowerinterest rate, or it may be a pension deficit—which is highly inefficient, not only because ofthe employee or PBGC risk, but also because thecompany is deferring the tax deduction availablefor paying off the pension debt and forgoing theuse of the pension tax shelter on the earnings ofthat payoff.

Either type of refinancing reduces the company’safter-tax debt cost and strengthens its financialposition. So, these types of borrowing do not com-pete with borrowing to fund capital investment.

The downside of borrowing to fund a pensiondeficit is that it increases the likelihood that the

pension will be paid and raises the liabilityvalue—effects that are similar to those from vol-untarily collateralizing a risky debenture. If thepensions are not guaranteed, the employees arebearing the risk and the cost of eliminating therisk has to be recovered from the employeesthrough salary concessions (or from tax savings).If the pensions are guaranteed by the PBGC—that is, other plan sponsors—the cost of that riskshould properly be borne by the company, eitherby full funding (preferably) or through full risk-based premiums.

2. Doesn’t funding pension plans harm the econo-my by depriving plan sponsors of capital theycould use in their businesses?

Companies would, of course, like to divert toother business uses the portion of their compen-sation costs that should go into their pensionplans. Troubled plan sponsors are especially fondof this argument, which would save them thebother of competing for capital in the public mar-kets. But of course, money contributed to a pen-sion fund does not go down a rat hole; pensionfund investments recirculate it into the capitalmarkets to efficient users of capital.

ERISA’s intent is to limit plan sponsors’ ability touse their pension funds in their businesses.Permissive funding standards, however, create amassive loophole. ERISA generally restrictsdefined-benefit plans to investing no more than10 percent of the plan assets in the sponsor’s secu-rities. But that restriction applies only to theassets actually invested; it ignores the implicitemployer bond that covers the shortfall of thoseassets relative to full funding. By ignoring thisemployer bond, ERISA enables sponsors to turnhundreds of billions of dollars of pension capitalto their own uses.

3. If full funding is so attractive, why doesn’teverybody do it voluntarily?

PFN0504.qxd 5/3/05 8:40 AM Page 5

THE PENSION FORUM

6

Part of the answer to this question lies in the guar-antees provided by the PBGC, which largelyeliminate the employee pension risk that providesthe main incentive for full funding. The broaderreason that we do not see full funding, however, isthat pension finance is not currently transparent.Even for nonguaranteed pensions, employeesseem to be generally unaware of their pensionrisk. Not only employees but capital providersalso commonly fail to understand pensionfinance. When pension funds invest in equities,current accounting rules permit the sponsors toanticipate the risk premiums in their reportedearnings and to conceal the risk by smoothing outthe effect of market fluctuations. Financial econ-omists commonly assume that investors lookthrough the reported earnings to the underlyingeconomic reality. Companies, however, do notappear to share that assumption about investorsophistication, and recent empirical research sup-ports the company view with regard to pensionaccounting (Coronado and Sharpe 2003). Thus,companies have been able to deal with pensionrisk through sponsor-friendly accounting rulesrather than genuine asset/liability management.

4. Why not fund pension liabilities with equitiesor other risky assets that have higher expectedreturns than bonds?

By funding with risky assets (risky beyond themodest level suggested in the section “A Note onImmunization”), a company fails to eliminate theplan’s dependence on the company’s credit. Thatcompany-specific risk is inefficiently borne eitherby employees (for nonguaranteed pensions) or bythe PBGC.

Furthermore, investing the pension fund in riskyassets leaves the plan leveraged rather thandefeased. In the transparent financial worldtoward which we are moving, pension risk wouldraise the company’s cost of capital. By absorbingsome of the company’s risk-taking capacity, pen-sion fund equity risk would come at the expense

of other risks that the company could take with-out introducing inefficiencies into employee com-pensation and tax management.

Corporate investing in marketed equities deliversno value to shareholders: The shareholders canmake those investments for themselves. But thosepension fund equity investments may crowd outthe investments in the core business that canuniquely deliver value to shareholders.

In addition, funding with equities gives up the taxgain available with bonds (Tepper).

5. Isn’t funding with immunizing bonds more expensive than funding with equityinvestment?

Yes, under the standard actuarial or accountingmodel. No, in terms of shareholder value.Although the expected contributions over the lifeof immunized plans are higher, there is a compen-satory drop in the company’s risk, so shareholdervalue is unaffected. The only “loss” to the compa-ny comes from the transfer of value to employeesor the PBGC by better collateralization of thepensions (see the answer to Question 1), and thecompany can recover any value transferred toemployees through salary concessions that recog-nize the greater pension value. Overall, sharehold-ers gain from substituting bonds for stock in thepension plan because of tax efficiencies and othersecondorder effects (Bader 2003a).

6. Full funding would generate considerabledemand for high-quality, long-duration bonds.This demand would disrupt the U.S. capitalmarkets and cause the interest rates on suchbonds to drop to levels that pension sponsorswould find unattractive. In most countries, theinadequate supply of such bonds would makelarge-scale immunization impossible.

Since 1980, the sleep of pension plan sponsorshas been untroubled by the Tepper–Black

PFN0504.qxd 5/3/05 8:40 AM Page 6

THE PENSION FORUM

7

critique of their errors. To worry that sponsorswill all awaken one morning in a headlong rushto implement the Tepper–Black advice seemsrather alarmist.

In free markets, new demand for long-durationbonds should, over time, call forth an adequatesupply. As companies immunize their long-duration pension liabilities, they will acquirecapacity to issue long-term debt without net dam-age to their balance sheets. (They will simply besubstituting one long-term liability for another.)And if long-term market debt carries low interestrates, companies will choose to issue such debt inpreference to using other capital sources, such asprivate credit, short-term debt, or equity financ-ing.

7. Even granting that secure pensions serve the company’s or the PBGC’s interests, why fund beyond the amount needed to purchaseannuities?

The actual purchase of an annuity contract wouldprovide adequate security. But simply funding toa level that is believed to be adequate for an annu-ity purchase would not.

The private annuity market for pension plan ter-minations is small, and its pricing is opaque.Pension plans cannot hedge their funding levelson an annuity purchase basis, so they cannotassure that adequacy today means adequacytomorrow. Also, insurance companies combinetheir gross interest rate with conservative demo-graphic assumptions and loadings for profit andexpenses. Therefore, annuity purchase rates areunlikely to be significantly (if at all) below liabil-ities that combine Treasury rates with the demo-graphic assumptions used for funding the plans.

8. Why would companies establish defined-benefitplans with such funding strictures? Defined con-tribution plans can give employees similar bene-fits (through investment in a Treasury portfolio)and other options they might prefer (such asequity investments).

In the United States, this is a trillion-dollar question, to which the answer is not at all clear:Can the virtues of defined-benefit plans out-weigh the clarity, relative administrative simplic-ity, and employee choice offered by defined-contribution plans?

A defined-benefit plan cannot provide the samebenefits as a defined-contribution plan morecheaply if the risks to the shareholders are correct-ly reflected. But neither is it a more expensivevehicle. It is simply a different vehicle—one inwhich the company may provide value to theemployee by absorbing certain demographicrisks.8 It is also a more efficient human resourcetool. Unlike defined-contribution plans, defined-benefit plans can provide guaranteed incomeamounts targeted to achieve various humanresource objectives, such as encouraging early,normal, or late retirement. The target levels canbe met through good times and bad, so humanresource planners need not worry that a marketplunge will discourage retirements just when thecompany most desires voluntary departures.Defined-benefit plans also lend themselves morereadily than defined-contribution plans to “win-dow programs” that might be needed to copewith temporary conditions.

Employees who want equity exposure can obtainit with assets other than their pensions.(Companies might assist with supplementaldefined-contribution plans.) For employees whohave no other financial assets, it may be just as

8 Defined-benefit plans have the apparent advantage of paying lifetime pensions, which free employees from the danger ofoutliving their retirement plans. This advantage is diluted, however, because these plans also commonly offer lump-sumoptions, which are heavily used. Also, defined contributions can, and often do, offer annuity purchase options.

PFN0504.qxd 5/3/05 8:40 AM Page 7

THE PENSION FORUM

8

well that their savings take the form of fixed andsecure pensions.

Transition

Transition from the current permissiveness to a full funding standard even over an extendedtime would be painful to some major businessesand their employees. An important first step,however, would be to stop the bleeding—by preventing plan sponsors from taking on newunfunded liabilities. Specifically, a plan should bepermitted to accrue additional benefits, by planamendment or by continuing accrual of creditsunder existing provisions, only if

• the sponsor fully funds those new accruals or• existing plan assets are sufficient to maintain

full funding.9

How can such a draconian provision be justified?If a company cannot afford currently to pay itsemployees’ salaries, other companies are notrequired to chip in. The same standard shouldapply to a company that provides part of itsemployees’ pay in the form of pensions. If thecompany cannot afford to pay for those pensionscurrently, it should not be able to impose on other companies the cost of guaranteeing thosepensions. Although dumping pension liabilitieson the PBGC is fast becoming a major corporatepastime, encouraging the weak to prey on thestrong is neither a fair nor an efficient way to runan economy.

Conclusion

The idea that underfunding pension plans is away for companies to borrow inexpensively fromtheir employees is a myth. It may be true for com-panies with weak credit, but only if someoneelse—someone other than the company—is bear-

ing the pension risk without full compensation.For nonguaranteed pensions, the someone elsemust be employees who do not recognize the riskthey are bearing. For guaranteed pensions, thesomeone else must be a guarantor who does notcharge enough for the risk.

Without a guarantee, informed employees woulddeeply discount an underfunded pension promisefrom a weak company. They would discount it,first, for the normal default risk and, second, forthe employer-specific nature of that risk. So, theywould charge for the borrowing by requiringmuch larger salaries than if the pension were fullyfunded. Thus, the employees’ inability to diversi-fy firm-specific risk makes them a poor financingsource for their employers.

If the pensions are guaranteed, the cost of thepension fund “borrowing” depends on the premi-ums charged by the guarantee agency. If the pre-miums are accurately risk based, they effectivelyimpose a market interest rate on the borrower.

In this article, we began with considering aneconomy without governmental guarantees forpension funding. We found that transparencyshould lead to voluntary full funding.Otherwise, employers and employees wouldhave inefficient compensation contracts thatexposed employees to risk that they could notdiversify. We then introduced a guarantee pro-gram and found that it reversed the main incen-tive for full funding. We noted that insufficientfunding, however, enables weak or irresponsibleplan sponsors to dip into the pockets of othersponsors—and perhaps of taxpayers. So, thegovernment that includes a guarantee programmust require plan sponsors to fund their plans;that is, it must compel behavior that would

9 This condition would often make the introduction of plan amendments (or new plans) that provide significant “past service bene-fits” impractical. Although intended as an incentive for employees to render future service, these benefits are credited to employeesimmediately, creating substantial current liabilities. Gold (2003) suggested an alternative plan design that would credit the benefitincreases only over employees’ future service, which would improve both the incentive effects and the economics.

PFN0504.qxd 5/3/05 8:40 AM Page 8

THE PENSION FORUM

9

occur naturally in an unregulated, transparentpension system.

In short, pension risk is inefficiently borne byemployees or governmental guarantors. Full fund-ing eliminates the pension risk. With or withoutguarantees, full funding is the optimal conditionfor all stakeholders in the pension system.

I thank Bruce Cadenhead, Jeremy Gold, TomLowman, Wendy McFee, Bob North, and PeggyWarner for their comments and suggestions.

References

Bader, Lawrence N. 2003a. “The Case againstStock in Corporate Pension Funds.” PensionSection News (February): www.soa.org/library/sectionnews/pension/PSN0203.pdf.

________2003b. “Treatment of Pension Plans ina Corporate Valuation.” Financial AnalystsJournal, vol. 59, no. 3 (May/June): 19–24.

Bianco, David, Zhen Deng, and Pinto Suri.2004. “Equities Ahead but Pensions StillBehind.” UBS Securities (April).

Black, F. 1980. “The Tax Consequences of Long-Run Pension Policy.” Financial Analysts Journal,vol. 36, no. 4 (July/August): 21–28.

Bodie, Zvi. 1996. “What the Pension BenefitGuaranty Corporation Can Learn from theFederal Savings and Loan InsuranceCorporation.” Journal of Financial ServicesResearch, vol. 10, no. 1 (March): 83–100.

Bodie, Zvi, and Robert C. Merton. 1992. “On the Management of Financial Guarantees.”Financial Management, vol. 21, no. 4 (Winter): 87–109.

Coronado, Julia Lynn, and Steven A. Sharpe.2003. “Did Pension Plan Accounting Contributeto a Stock Market Bubble?” In Brookings Paperson Economic Activity. Edited by William C.Brainard and George L. Perry. Washington, DC:The Brookings Institution.

Gold, Jeremy. 2003. “Periodic Cost of EmployeeBenefits.” Pension Actuarial Practice in Light ofFinancial Economics Symposium. Vancouver,Canada: Vancouver Society of Actuaries (24 June).

Sharpe, W.F. 1976. “Corporate Pension FundingPolicy.” Journal of Financial Economics, vol. 3, no.2 (June):183–93.

Tepper, Irwin. 1981. “Taxation and CorporatePension Policy.” Journal of Finance, vol. 34, no. 1(March): 1–13.

PFN0504.qxd 5/3/05 8:40 AM Page 9

THE PENSION FORUM

10

Mr. Bader’s paper includes a number of short-comings that may very well be curable by moredetailed treatment. I look forward to a more com-prehensive paper from the author on this topic.

A Simple Pension Promise

Assume two vested participants in a defined ben-efit plan are identically situated and have identi-cal work histories. The sponsor has always madethe same contribution to the plan for each ofthem. They both have the same “pension prom-ise.” One terminates employment in the nextyear, and the other works until normal retirementage. The latter participant realizes more from hispension promise than the former, possibly a lotmore depending on what point in their careers wemake our observation. The bottom line is that thepension promise is not simple. Arguments thatmake perfect sense in the context of the simplifiedarrangement the author uses in his example maybreak down entirely when the complexities ofactual plans are considered.

Use of a Bond Model

While the pension promise has many characteris-tics of a bond, it is not a bond and has many non-bond-like characteristics, such as the following: (1)the payment amounts are contingent rather thancontractually set; (2) the term is longer than anycurrently marketed bond even without taking intoaccount future participants; (3) there is no balloonpayment date when all or a significant portion ofliability becomes due; (4) liabilities behave differ-ently from bonds in certain economic conditions,such as rapid inflation. The author makes an enor-mous and unjustified leap by applying a bondmodel under these circumstances.

Tax Arbitrage

Investment advice is not a service most actuariesprovide. Indeed, few actuaries have the professional

qualifications to provide such advice. Althoughthere are exceptions, most of those who do pro-vide investment advice to plan sponsors have con-sistently recommended that a substantial portion,usually more than half, of a plan’s assets be invest-ed in equities. While I have no doubt that themathematics behind the author’s tax arbitrageargument is impeccable, the fact that those withexpertise in this area consistently flout theauthor’s advice is good evidence that other con-siderations compete with and, in many cases,override the tax arbitrage argument.

Alternatives to Full Funding

The author states that the only alternative to fullfunding is that “the government can charge eachplan sponsor a premium that accurately reflectsthe risks that the sponsor imposes on the system.”Another alternative is that such premiums can becharged by a private self-insurance pool or bycommercial insurers. Several recent papers haveexplored this possibility. I would be curious toknow how the author views this alternative.

Equity Risk

In his “final and critical problem with permissivefunding and investment rules,” the author raisesthe specter of “a severe and prolonged decline instock prices.” Historically, even taking intoaccount the worst financial cataclysms, equitiesalways have outperformed bonds when measuredover a sufficiently long time period. Therefore,the author must be talking about a scenario worsethan any in history. Such a circumstance wouldlikely be accompanied by large-scale defaults onbonds, even among previously highly rated com-panies, so full funding with bonds would not be asure remedy. Full funding with Treasury bondswould solve this problem, but this alternativewould be costly if all plans decided to cover alltheir liabilities with Treasuries.

Comments on “Pension Deficits: An Unnecessary Evil”

by Eric Klieber, F.S.A., M.A.A.A., E.A.

PFN0504.qxd 5/3/05 8:40 AM Page 10

THE PENSION FORUM

11

Perspective Is Complete NowIt’s always a pleasure to read, comment, agree, ordisagree with Mr. Bader’s papers. Once again, Mr.Bader gives us a great example of his thought-pro-voking writings. My goal here is to challenge cer-tain arguments utilized in this piece to advanceMr. Bader’s views on the subject of asset alloca-tion for pension plans. A comprehensive treat-ment of the issues raised in Mr. Bader’s paper, aswell as alternative views on the subject, are out-side of the scope of this short and somewhatinformal piece.

“Pension Deficits: An Unnecessary Evil” is a nat-ural conclusion of a series of ideas presented inBader (2001), Bader and Gold (2003, also knownas Reinventing), Bader (2003a), and Bader(2003b). Publication of Reinventing became a sig-nificant event in actuarial circles and beyond: thepaper jump-started a stimulating discussion aboutfoundations of the pension actuarial science andits relationship with financial economics. I’vealways thought that the line of thinking present-ed in Reinventing is remarkably incomplete.While the authors have insisted that the only eco-nomically legitimate pension liability is, for allintents and purposes, a bond portfolio, they havestopped short of calling for a similar mandate onthe asset side … until now.

Mr. Bader clearly believes that the only asset classappropriate for pension funds is high-qualityfixed-income securities. After all, if you believethat the liability is a bond portfolio, which assetbehaves like a bond portfolio? It’s a no-brainer. Inseveral publications, Mr. Bader and like-mindedeconomists essentially have argued that it wouldbe nice if pension plans invested in bonds only.But as long as those niceties remain voluntary,they have the annoying tendency of failing to

materialize. The decision makers for both corpo-rate and public pension plans have been stub-bornly investing in a variety of financial instru-ments that, unlike bonds, make no promises topay the investors back, offering uncertain chancesof value appreciation instead.

Should we allow a bunch of possibly confusedand potentially ill-intentioned people—the ones who make the asset allocation decisions forpension plans—to get in the way of a good eco-nomic theory? Not according to Mr. Bader. Hestates that “mandatory full-funding … is the only practical prevention for the diseases that can afflict a guarantee system” and “a soundgovernment guarantee system must mandate fullfunding.” Liabilities must be a bond portfolio;assets must be a bond portfolio as well. End ofstory.

System without Guarantee

As a starting point, Mr. Bader discusses pensionfunding under an assumption of no governmentregulations and guarantees. I make the sameassumption in this section.

Mr. Bader starts off with two “heroic” assump-tions. His first assumption states “a dollar owedto a pensioner” is very similar to “a dollar owedto a creditor.” This assumption is debatable atbest: there are significant differences betweenthose “dollars.” His second assumption basicallyproclaims that we must view employees (and, Imay add, all stakeholders) as fully informedabout financial health of the pension plan. Iagree with this one—no argument should bebased on ignorance. There is one more assump-tion in the paper, although Mr. Bader doesn’tmention it. The third assumption is that thematching bond portfolio exists for every pension

A Critique of “Pension Deficits: An Unnecessary Evil”

by Dimitry Mindlin*, A.S.A., M.A.A.A., Ph.D.

PFN0504.qxd 5/3/05 8:40 AM Page 11

THE PENSION FORUM

12

plan out there. I believe that this is a dubiousassumption: the matching bond portfolio rarelyexists even if we were willing to accept an inexactbut “reasonably close” match. But this subject isone for some other time.

For more details regarding the first and thirdassumptions, see my “Reaffirming PensionActuarial Science,” Mr. Bader’s comments to thatpaper, and this author’s response to his comments in this issue of The Pension Forum.

Mr. Bader’s goal is to justify the “full funding” bymeans of investing in “an immunizing bond port-folio.” The section entitled “PreregulatoryEnvironment” contains two lines of arguments infavor of the “full funding”: (1) presence of thecompany-specific risk and (2) tax arbitrage.

As far as the first line of arguments is concerned,I agree with Mr. Bader that the company-specificrisk is a factor. Unfortunately the paper is silentabout the magnitude of this factor. I also agreethat the pension’s “cost to the employer is greaterthan its value to employees.” However, this ineffi-ciency is not unique to pensions. In almost everyform of compensation, cost to the employer isgreater than its value to the employees. A salary,for example, is terribly inefficient in that respect.In the presence of taxes on both the employer andemployee sides, it’s not unusual for an employeeto get less than $0.60 for every $1 the employerspends on her salary. Other employee benefits canbe very inefficient as well. An employer-providedhealth insurance policy could be needlessly com-prehensive for some employees; it could also behopelessly insufficient or even useless for others.

If Mr. Bader’s goal is to maximize the value to theemployees for every dollar of labor cost, he mustdemonstrate that the cost of elimination of thecompany-specific risk in the pension plan is themost efficient way to do so. The paper does noth-ing of the sort. It is true that the full fundingeliminates the inefficiency present in the pension

plan, but the cost of elimination of this inefficien-cy can be prohibitive. The choice of the pensionplan as the best place to spend the labor-cost dol-lars appears to be unsubstantiated in the paper.

As far as the tax arbitrage arguments are con-cerned, several important questions are in order.Is the role of the long-term strategic asset alloca-tion to follow every twist and turn of ever-chang-ing fiscal policies? If equities were taxed at a high-er rate than bonds, would Mr. Bader and like-minded economists advocate 100% equity port-folios? Is it possible that in the process of takingadvantage of the tax arbitrage we may eliminatesome other essential advantages that pensionplans enjoy over other market participants (e.g.,the pension plans’ long-term nature)? As long asthese questions remain unanswered, as they are inthis paper, the tax arbitrage arguments are lessthan convincing.

System with Guarantee

A plan sponsor established its pension plan withthe intention to make the plan a valuable part ofthe compensation and fund it properly. Mr. Badersuggests that the very presence of governmentguarantees fundamentally changes the intentionsof the plan sponsor. According to Mr. Bader,“each sponsor’s narrow interest is thus to fund aslittle as possible.” That’s a very strong claim, andI would like to see much more concrete evidenceof that phenomenon than presented in the paper.I find the alleged desire of plan sponsors tounleash the government agency’s awesome powerto take over a good chunk of the company’s assetsrather exaggerated.

The paper also contains a glaring inconsistency: “afailing company ... may be able to solve that prob-lem by promising outsized pensions and fundingthem inadequately.” No, it may not. According tothe second assumption, the employees “under-stand the risks of both under-funding andasset/liability mismatches.” They “are able to make

PFN0504.qxd 5/3/05 8:40 AM Page 12

THE PENSION FORUM

13

rational trade-offs between pensions and salary”—they will not be fooled by that scheme.

The paper has a very interesting corollary, perhapsunintended by the author. If we assume for amoment that Mr. Bader’s arguments are flawless,then the belief that pension plans should be allowedto take advantage of investment opportunities out-side of high-quality bonds and the belief in the use-fulness of government guarantees are irreconcilable!Indeed, if one believes in the government guaran-tees, then, according to Mr. Bader, one must believethat pension plans must invest exclusively in high-quality bonds. I don’t think the founding fathers ofthe PBGC anticipated that conclusion.

The usefulness of government guarantees is an openquestion. But that’s not the most important ques-tion now—the government guarantees do exist, thePBGC does have real powers. The real question ishow much power the PBGC should have. Somemay argue that the PBGC needs more authority toimprove its accounting statements. It may also betempting, as Mr. Bader is proposing, to take awaythe loosely regulated multitude of investmentoptions currently available to the plan sponsors.These solutions strike me as premature and, quitepossibly, counterproductive. The danger here is thatthe plan sponsors may view not the pension deficits,but the pension plans themselves, as an unnecessaryevil. They may simply get out of the business ofproviding secure retirement for their employees.Then the question of the usefulness of governmentguarantees becomes moot: there will be nobody toregulate and nothing to guarantee. But these issuesrequire more thorough treatment than is appropri-ate for this piece.

Conclusion

Reinventing (Bader and Gold 2003, p. 10) containsan example of a pension plan that, under certain

conditions, will never have a 100% funding ratio:“Under these conditions, the funding ratio will sta-bilize at just 70%, forever (italics added). Is thisresult professionally defensible?” I’m supposed to beashamed of myself for not being scared stiff, but Ifind myself rather happy for the plan sponsor andparticipants. Imagine that—people work, earn theirpensions, retire and live happily ever after—and theprocess continues forever! The price to pay for thisresult is the possibility of living with accountingstatements that are not good enough for some. Butif we force the sponsor into the perfect accountingsystem, the sponsor may find the “forever” part tooburdensome, or expensive, or both. Looking at thecurrent conditions in the pension industry, is it pos-sible that our choice is between “perfect account-ing” and “forever”? I’d take “forever,” thank youvery much.

References

Bader, L,. 2001. “Pension Forecasts, Part 2: TheModel Has No Clothes.” Pension Section News no.46 (June): 14–15. http://library.www.soa.org/library-pdf/PSN0406.pdf.

________“The Case against Stocks in CorporatePension Funds.” Pension Section News no. 51,(February): 17–19. http://library.soa.org/library/sectionnews/pension/PSN0302.pdf.

________2003b. “Treatment of Pension Plans ina Corporate Valuation.” Financial AnalystsJournal, vol. 59 (May/June).

Bader, L., and J. Gold. 2003. “ReinventingPension Actuarial Science.” The Pension Forum,15(1): 1–13. http://library.soa.org/library/sectionnews/pension/PFN0301.pdf.

* Dimitry Mindlin is a vice-president at Wilshire Associates Inc. Opinions presented in this paper are his own and do notrepresent positions of Wilshire Associates, Inc.

PFN0504.qxd 5/3/05 8:40 AM Page 13

THE PENSION FORUM

14

1. Response to Eric Klieber

Mr. Klieber speaks for many actuaries in his concise summary of objections to the financialeconomics view of pension plans. Every point heraises deserves more extended discussion than isappropriate here, but I will sketch out a responseto each.

1.1 A Simple Pension Promise

Mr. Klieber distinguishes between the pensionpromises to two vested participants who haveidentical accrued pensions today, but will ulti-mately receive very different pensions.

From a corporate finance perspective, only anaccrued benefit qualifies as an economic liabilityof the sponsor and an asset of the participant.Regardless of the attribution provided by the ben-efit formula and current accounting rules, anyincremental pensions are earned only by futureservice. Like future salaries, which are also earnedonly by future service, these incremental pensionsshould not be prefunded or thought of as currentliabilities. So I reserve the term “pension promis-es” for accrued pensions only. More detail appearsin Bader (2003).

1.2 Use of a Bond Model

Mr. Klieber enumerates several characteristics of pensions that he believes are “non-bond-like,” making a bond model unsuitable for understanding pensions.

Of course, there are many differences betweenpensions and bonds. But both represent well-defined cash flows that must be paid on pain ofbankruptcy. When the sponsor writes a check,why does it matter whether it is paid to a pension-

er or a creditor? If a portfolio of bonds has the same cash flow as a pension obligation—that is, the two cash flows are identical in timingand amount (or probability distribution ofamount)—why should they have different valuesto payer or payee?

For any pension obligation, can we identify abond portfolio with very similar cash flows? Let’sconsider Mr. Klieber’s list of how pensions differfrom bonds.

1. Pension payments are contingent rather than con-tractually set. Actuarial valuations of pensionseffectively reduce the probability distribution ofpayments to a single fixed-payment stream (equalto the expected payments). A bond portfolio thatmatches that stream has the same value as thepension liability, regardless of how the paymentsare defined. (By the way, many bonds do havecontingent payments, in the form of options.Bond analysts and traders price these options withsophisticated techniques that reflect the full prob-ability distribution rather than the average out-come.)

2. The term is longer than any currently marketedbond. In fact, some $7 billion of 100-year invest-ment-grade bonds is currently outstanding(November 2004), spread among 17 issuers suchas BellSouth, Citigroup, Coca-Cola, Disney,Ford, and IBM.

3. There is no balloon payment date. No singlebond will match a pension payment stream. But aportfolio of bonds, possibly including zero-coupon bonds, can be constructed to match orapproximate most pension streams.

Author’s Response to Mr. Klieber’s and Mr. Mindlin’s Comments

by Lawrence N. Bader, F.S.A.

PFN0504.qxd 5/3/05 8:40 AM Page 14

THE PENSION FORUM

15

4. Pension liabilities behave differently from bondsin certain economic conditions, such as rapid infla-tion. Per (1) above, pension liabilities compriseonly accrued benefits, which are generally unaf-fected by inflation.

In any event, we need not produce a perfect bondmatch for every pension stream. The bond marketis sufficiently deep and varied to permit excellentapproximations for any cash flows, whether ornot perfectly matching trading instruments exist.As I remark in my critique of Mr. Mindlin’s arti-cle in this issue of The Pension Forum, many dif-ferences exist within the bond market, andbetween bonds and pension obligations. Weaccommodate such differences by judgment andarithmetic, not by completely different valuationtechniques. A serious challenge to the bondmodel of accrued pensions must not merely pointto these differences. It must show why the differ-ences affect value to the payer or payee in suchfundamental ways that they require entirely dif-ferent valuation models.

1.3 Tax Arbitrage

Mr. Klieber observes that expert practitioners in asset allocation consistently flout my claim that all-bond portfolios are optimal for pensionfunds, thereby showing that “other considera-tions . . . override the tax arbitrage argument.”He is correct. The “other considerations” relatemainly to transparency, which I assume in myarticle, but which is certainly lacking in today’s pension world.

The transparency assumption is intended not toportray current reality, but rather to illuminatehow pension plans would be financed if all prin-cipals fully understood their exposure. Currentpension accounting rules block this understand-ing by, among other flaws, converting anticipatedequity risk premiums into current corporate earn-ings. Shifting pension fund assets from stocks tobonds may immediately lower executives’ earn-ings-related bonuses and disappoint investors

who fail to look through the financial reports tothe underlying economic reality. Major advancesin transparency seem inescapable and shouldencourage a shift toward bonds.

1.4 Alternatives to Full Funding

Mr. Klieber observes that the only alternative tofull funding that I recognize is a government-runinsurance system with true risk-related charges.He suggests another alternative: a private systemwith premiums set by the insurance market.

Mr. Klieber has an excellent point, and I agreethat a private insurance system may be viable.Such a system would impose the true risk-basedpremiums that are politically impractical for agovernmental agency. But I believe that such pre-miums would get us to about the same place asmy recommendation for mandatory full fundingand immunization. True risk-related premiumswould provide powerful inducements toward vol-untary full funding and immunization.

Here’s an oversimplified example. Suppose that asponsor has an unfunded plan. An insurer of theunfunded liability would charge a premium thatcovers the risk of the sponsor’s default on thisdebt-like obligation. The appropriate amountwould be the default premium embedded in thesponsor’s borrowing rate on debt similar to thepension liability—that is, the spread of his bor-rowing rate over comparable Treasuries.

Instead, the sponsor can fully fund the plan byborrowing in the capital markets and buying animmunizing Treasury portfolio. This transactionwould eliminate his risk-related insurance premi-um. His net cost now would be the excess of hisborrowing rate over the earnings of the pensionfund, that is, his spread over the immunizingTreasuries.

So, before considering taxes, the sponsor wouldbe just as well off borrowing the money and fund-ing the pension plan. After taxes, he would be

PFN0504.qxd 5/3/05 8:40 AM Page 15

better off because the interest on the borrowing isdeductible, while the interest on the pension fundis tax-free.

We can similarly show that it would be more eco-nomical for a sponsor to eliminate pension fundinvestment risk by a capital market transactionthan to take the risk and pay an insurer for pro-tection against it (again, neutral on a pre-tax basisand preferable after taxes). The general point hereis that we buy insurance when the cost of elimi-nating a risk is greater than the cost of insuring it.But if the risk can be eliminated through a stan-dard capital market transaction, no insurance car-rier, governmental or private, should be able tocover that risk at a lower cost than the capitalmarkets. Therefore, accurate (or excessive) risk-based premiums would strongly encourage thefull funding that I regard as the optimal conditionfor pension plans. The insurance system wouldthen be relieved from insuring systemic risks andwould mainly handle accidents such as demo-graphic losses.

1.5 Equity Risk

Mr. Klieber suggests that I am overstating thesafety of bonds. He claims that a market crash sodeep and prolonged that stocks would underper-form bonds over a very extended period would beaccompanied by large-scale defaults on bonds,which would not protect pension funds.

Over the past two decades Japan presents anexample of extended equity underperformanceduring which bonds were a safe haven. Closer tohome, a look at the Great Depression decade1930–1939 offers some insight. During this peri-od the average annual return in the U.S. was -0.1% on equities, 6.9% on long-term corporatebonds, and 4.9% on long-term governmentbonds (Ibbotson and Sinquefield 1982). If suchreturns were to prevail during a future (or cur-rent) decade, failures would be widespread amongcorporate pension plans that rely on large equityallocations to produce high-single-digit returns.

Any plans that rely only on earning Treasury rateswill be fine if they are wholly invested inTreasuries or corporates. Incidentally, the need forthe U.S. Treasury to fund trillions of dollars ofprojected federal deficits over the next decadesuggests that a migration of pension funds toTreasury bonds could be part of a solution ratherthan a problem for the capital markets.

In any event, the argument in my article does notdepend on bonds’ being perfectly safe, just signif-icantly safer than equities during periods whenplan sponsors are failing.

2. Response to Dimitry Mindlin

Mr. Mindlin’s contribution to this ongoingdebate is written in his customary engaging style.I will confine my response to substantive mattersnot addressed in our exchange of pleasantries overhis own article in this publication or in myresponse to Mr. Klieber above.

2.1 Inefficiency of Pensions as EmployeeCompensation

Mr. Mindlin states that whatever inefficiency Iidentify in the financing of pensions is dwarfed bythe inefficiency of salaries. He claims, “It’s notunusual for an employee to get less than $0.60 forevery $1 the employer spends on her salary.”

This 60% efficiency figure is surprising, consider-ing that salaries are deductible to the corporateemployer and that most payroll taxes confer somebenefit on the employee. Whatever the correctpercentage, pensions should be more efficientthan salary because of their favored treatmentwith respect to income and payroll taxes. And thepossible inefficiency of salaries or other forms ofcompensation does not excuse a failure to makepensions as efficient as possible.

Mr. Mindlin states that the costs of making pensions efficient “can be prohibitive.” But fund-ing unfunded benefits gives a clear tax gain, not

THE PENSION FORUM

16

PFN0504.qxd 5/3/05 8:40 AM Page 16

THE PENSION FORUM

17

a cost. Assets can be immediately and cheaplyreallocated by using futures. The sponsor mayadjust the underlying assets over time, by direct-ing contributions and investment income intobonds while depleting the equity allocationthrough attrition.

2.2 Tax Arbitrage

Mr. Mindlin refers to the twists and turns of“ever-changing fiscal policies” and asks, “If equi-ties were taxed at a higher rate than bonds, wouldMr. Bader . . . advocate 100% equity portfolios?”

Such a twist would be both unprecedented andimprobable. If it did occur, it would create a con-flict between pension safety and tax considera-tions, and the asset allocation decision would bedifficult. But as both safety and tax factors nowargue for bonds, the decision is simpler.

Also in this section, Mr. Mindlin refers to the“essential advantages that pension plans enjoyover other market participants (e.g., the pensionplans’ long-term nature).” In his own article in this Pension Forum, he refers to “the sponsor’s. . . risk aversion characteristics [that] may allowthe sponsor to take some risk and enjoy (orregret) the results.” In both instances he ignoresPrinciple 5 (Bader and Gold 2003): “Risks areborne and rewards are earned by individuals, notby institutions.”

Neither corporations that sponsor plans nor theplans themselves have investment horizons, riskpreferences, or risk capacities independent oftheir human stakeholders. They are insensatefinancial entities created to enable shareholders totransact with other parties. I do not feel theirpain, and neither do they. Actuaries should con-fine their pension plan concerns to the real peoplewith stakes in these plans—the shareholders whopay for them, the participants who receive bene-fits from them, and indirectly the members of thesociety inhabited by these shareholders and par-ticipants. It is these people who have the horizons

and risk tolerances by which we must measure theutility of pension plans.

2.3 Employees’ Understanding of the Risks of Underfunding andAsset/Liability Mismatch

In his section headed “System with Guarantee,”Mr. Mindlin discerns a “glaring inconsistency” inmy statement, “a failing company may . . . [prom-ise] outsized pensions and [fund] them inade-quately.” “No, it may not,” he objects: under mytransparency assumption, employees will see therisks and “will not be fooled by that scheme.”

On the contrary, informed employees will be wellsatisfied to receive pensions guaranteed by thegovernment, however poorly the company mayfund them. The employees are aware that thecompany’s scheme puts the insurance programrather than themselves at risk. The insuranceenables them to happily collect more compensa-tion than their employer can afford.

2.4 Usefulness of Government PensionGuarantees

In the same section, Mr. Mindlin finds “a veryinteresting corollary, perhaps unintended by theauthor,” that belief in permitting investmentsother than high-quality bonds and belief in gov-ernment guarantees are irreconcilable.

This “perhaps unintended corollary” is the entirepoint of my paper: with or without governmentguarantees, full funding (i.e., adequate assetsinvested in an immunizing portfolio) is the cor-rect standard for the pension system.

3. Conclusion

I thank both commenters for helping to frameand debate these issues that are so crucial to ourprofession. In particular, Mr. Klieber’s com-ments on a private insurance system direct ourattention to a line of thought that can be pro-ductive intellectually and perhaps practically as

PFN0504.qxd 5/3/05 8:40 AM Page 17

THE PENSION FORUM

18

well, given the staggering problems now facedby the PBGC.

References

Bader, Lawrence N. 2003. “Treatment of PensionPlans in a Corporate Valuation.” FinancialAnalysts Journal 59 (May/June 2003).

Bader, Lawrence N., and Jeremy Gold. 2003.“Reinventing Pension Actuarial Science.” ThePension Forum, 15 (1): 1–13. http://library.soa.org/library/sectionnews/pension/PFN0301.pdf

Ibbotson, Roger G., and Rex A. Sinquefield.1982. Stocks, Bonds, Bills and Inflation: The Pastand the Future, 1982 Edition. Financial AnalystsResearch Foundation.

PFN0504.qxd 5/3/05 8:40 AM Page 18

THE PENSION FORUM

19

Recent events suggest that it is time to reexamineexisting pension plan funding rules and considerchanges. This is a discussion of how the currentrules came into being, their shortcomings, andpossible replacing rules.

Responsible pension funding began long beforeERISA. For many years, most plan sponsors hadworked closely with their actuaries in developingrational funding programs. Generally there werethree objectives: smoothing year-by-year contri-butions, avoiding surprise contribution require-ments, and making steady progress toward target-ed funding levels. Often these targets involvedfully funded accrued liabilities.

One objective rarely discussed was to protect work-ers in event of business failure. Sponsors had inmind the success and growth of their businesses,not planning for failure. In contrast, the principalobjective of ERISA funding rules was the protec-tion of workers in event of business failure. Manyobservers felt this was the only legitimate objective.

Despite this difference in objectives, the originalERISA funding rules mimicked, very closely, thefunding techniques that responsible employershad followed voluntarily in the years beforeERISA. So the brave new world of ERISA wasborn, and we all sat back to see how the newfunding rules would work.

It turns out that sometimes they worked—andsometimes they didn’t. When they didn’t work,the reasons for failure quickly became so obviousthat many planners were chagrined they hadn’tanticipated the failures.

The original rules were designed to reach targetfunding levels gradually and relatively painlessly

over a long period. So far, so good. So long as thesponsor remained healthy over this period ofgradual buildup, employees would be fully pro-tected—without any help from the PensionBenefit Guaranty Corporation (PBGC).

However, the gradual buildup applied to the planas it existed at the time ERISA was enacted.Consider an enhanced benefit added later byamendment. The enhancement was treated as abrand-new plan—with a new gradual buildupperiod. Now, consider the collectively bargainedplan whose benefits were not pay-related. To keepthese plans up-to-date, it was necessary to bargainfor additional benefits with every contract. Oftenthis meant adding a new piece of benefit everythree years. Sometimes it was less than three years.Under the original funding rules, every new piecestarted its buildup from ground zero as if it werea brand new plan. So the typical bargained planwould make funding progress for three years,receive a setback, move ahead for three moreyears, receive another setback, and so on. It wastantamount to running in place.

In any dynamic economy there will be a certainpercentage of business failures. When sponsors ofthese perpetually amended plans went under, theyinevitably left behind unfunded promises. ThePBGC found that its job of picking up the pieceswas becoming progressively more burdensome.

So in 1987 we the people, acting throughCongress, rewrote the rules. This occurred barelymore than a decade after the brave new world hadbegun. We moved our focus away from orderlyfunding designed to reach long-term targets.Instead, we focused on what might happen to aplan if its sponsor failed tomorrow. But, instead of

Fixing the Pension Plan Funding Rules

by Edward E. Burrows, E.A., M.A.A.A.

PFN0504.qxd 5/3/05 8:40 AM Page 19

THE PENSION FORUM

20

replacing the old rules with the new, we kept theold and layered on the new.

The new rules proved their worth very quickly.However, they did need refinement. So we tin-kered. And we tinkered. And we tinkered. Wekept tinkering until today we have a mountain ofcomplexity. Section 412 of the Internal RevenueCode, the section setting down minimum fund-ing rules, is now over 12,000 words long! It hasbecome a monster practically unfathomable toanyone other than a pension actuary working full-time in this very narrow field of specialization.Enormous amounts of effort are spent in the pri-vate sector just to keep up with, and comply with,the rules. Equally enormous amounts of effort are(or should be) spent by the regulators in checkingto see whether compliance is taking place.

The problem has been exacerbated by the absenceof adequate regulatory guidance on application ofthese complex rules. It is increasingly obvious thatthe time has come to step back, rethink our objec-tives, turn over a clean slate, and rewrite the rules.But, first, we should analyze the mistakes that gotus where we are today. There were lots of them.

Mistake 1: New Rules Layered on Old

When we discovered that the old rules weren’tworking, we added new ones. We probablyshould have completely replaced the old with thenew. That simple step would have streamlined theprocess enormously.

Mistake 2: Emphasis on Smoothing

Even with the new rules, we placed too muchemphasis on smoothing—dampening year-by-year volatility in contribution requirements.Where did we smooth? Just about everywhere.

l

Consider the changes in calculated liability thatoccur when it becomes necessary to true-upassumptions—investment return, future paychanges, mortality, and the like. We didn’t

require immediate recognition of thesechanges. Instead, after we had concluded that anew liability level was the only correct one, wepermitted gradual grading to this new level. Wedid this by establishing the difference betweenold and new levels, and amortizing that differ-ence.

l

Consider the inevitable gains and losses thatoccur when year-by-year results fluctuate sothey aren’t always exactly what had beenexpected. We permitted amortization of thesegains and losses too. We failed to consider whatmight happen if business failure should occurbefore amortization was complete.

l

When it came to asset value fluctuation, wepermitted even more dampening.

l

We permitted amortization (although notnearly so gradually as before) of the deficits thatexist when benefit enhancements outpacedasset buildups.

l

We permitted still more dampening whenbenchmark interest rates changed. We wentfurther and told actuaries they didn’t have touse even the dampened rates as long as theyused rates that were within broadly specifiedtolerances of the dampened targets.

In short, whenever it becomes obvious that someaspect of the world around us has changed, we’vebeen telling sponsors they needn’t recognize thechange all at once.

Mistake 3: Poorly Conceived Interest Rate Rules

Finally, we went haywire in specifying the interestrates to be used for different purposes. Today,depending on the rule being satisfied, the statuto-ry rate might be the actuary’s best estimate offuture investment return. Or it might be 120% ofa weighted average 30-year Treasury bond rate, or110% of that rate, or 105% of that rate, or 90%of that rate. Or it might be 100% of “the weight-ed average of the rates of interest on amountsinvested conservatively in long term investment-grade corporate bonds”—or 90% of that rate.

PFN0504.qxd 5/3/05 8:40 AM Page 20

THE PENSION FORUM

21

There were no differences in the nature of liabili-ties being valued that might justify these differentpercentages. For other purposes, it might be175% of the federal midterm rate. For still others,it might be 150% of that rate. Finally, wonder ofwonders, for some purposes, it’s simply 5%.

Why We Have Statutory Requirements

Obviously the first step in designing any new setof requirements is to identify the reason or rea-sons for having requirements at all. Just abouteveryone agrees that one reason is to protectworkers from losing the pensions they’ve earned iftheir company should fail. There’s an excellentcase for the proposition that protecting theseearned benefits is the only reason for statutoryrules.

Many employers will continue to want to see sta-ble contributions and avoid surprises, just as theydid before ERISA. Employers belonging to thisgroup won’t need mandatory rules focused on thisstability. These employers will voluntarily followprocedures that produce it.

What about other employers? What if an employ-er will not voluntarily adopt procedures thatsmooth contributions and avoid surprises? Doessociety have any business forcing this employer toadopt such procedures? We already have a rulethat in general an ongoing employer can’t aban-don a plan unless all accrued rights are fully fund-ed. If we have funding rules that protect employ-ees of companies that fail, should we be seekinganything more?

Arguably one reason for seeking something morewould be to encourage sponsors to maintain theirplans even when the going gets tough. If societyfollowed this argument, it would be saying it’s notenough to ensure participants that their earnedrights are protected. We’d be saying we must pro-tect participants against the likelihood that volun-tary plan termination would occur, causing a lossof future accruals. This would be a strange

approach for a society in which the adoption of aprivate plan is a voluntary act in the first place.

A Basic Funding Rule

Suppose we accept the proposition that the onlylegitimate purpose for statutory funding rules isto protect employees of businesses that fail, andthat maintaining stable contribution levels issomething employers may want to do voluntarilybut should not be required to do. Given these twopremises, the indicated basic funding rulebecomes the ultimate in simplicity:

Adjusted assets must always be at least as great as accrued benefits.

All that is necessary is to specify the rules fordetermining adjusted assets and the rules fordetermining accrued benefits.

Adjusted Assets

First, consider adjusted assets. The challenge is toobtain protection from the possibility that even ifassets are sufficient to cover liabilities today, theymight become insufficient tomorrow.

With bonds and similar debt securities, there aretwo principal risks:

• First, the issuer may go broke, leaving bond-holders with an empty bag.

• Second, prevailing interest rates may change.

If interest rates go up, market values of existingbonds will go down. If rates go down, market val-ues of existing bonds will increase, but issuers maycall their bonds, forcing bondholders to reinvestat a lower rate. Even if the investor has good callprotection, interest received on the bond will haveto be reinvested at lower rates. By restricting pur-chases to high-quality investment-grade bonds,the pension fund manager can minimize the riskof default.

PFN0504.qxd 5/3/05 8:40 AM Page 21

THE PENSION FORUM

22

The manager also can obtain protection from theinterest rate risk. Suppose a fund’s only obligationis a lump-sum benefit to be paid in 12 years. Themanager who covers this liability with a non-callable zero coupon bond due to mature in 12years can be indifferent to interest rate changes.Achievement of this “duration matching” doesn’trequire that every benefit disbursement bematched with every income receipt. The key is toconstruct a portfolio of bonds whose market valuecan be expected to change to the same extent asthe present value of pension obligations, givenany particular change in interest rates.

All of this means that a duration-matched portfo-lio of investment-grade bonds will minimize boththe default risk and the interest rate risk. A port-folio of this type will come close to providingcomplete assurance that accrued benefits will becovered if the market value of assets equals thepresent value of accrued benefits. The residualrisks posed by defaults and interest rate changesseem inconsequential.

A portfolio of investment-grade bonds that’s notduration-matched will minimize the default riskbut not the interest rate risk. The investmentmanager may have good reason to eschew dura-tion matching. If interest rates seem very low, themanager may want to avoid locking in these lowrates. The manager may prefer to invest in bondsof very short duration. If interest rates seem veryhigh, the manager may want to take the oppositeapproach. The manager may want to lock in theseapparently high rates by investing in long-termbonds. Finally, the manager simply may feel thatthe then-current yield curve favors a particularbond duration.

It would be possible to determine different adjust-ments for differing degrees to which assets andliabilities are mismatched. However, a refinementof this nature would be difficult to apply and dif-ficult to police. It might be better simply to estab-lish a single rule for investment-grade assets that

are not matched to liabilities. It might makesense, for example, to have a rule that the adjust-ed value of nonmatched investment-grade assetswill equal 90% of market value.

With rules for investment-grade bonds and simi-lar debt obligations clearly established, thereremains a wide variety of other investments still tobe treated. This third group includes all forms ofequity ownership. It also includes non-invest-ment-grade debt obligations. In this third catego-ry, equity ownership probably offers the greatestchallenge. Here the risk goes far beyond the riskthat the issuer will become bankrupt. It includesthe risk of a temporary or permanent downturnin the issuer’s business operations. That downturnmight be unique to the issuer, or it might be epi-demic in the issuer’s industry. Or it might reflectgeneral economic conditions.

Worse yet, there is the risk of unpredictable andsometimes apparently irrational changes ininvestor attitudes. About the only thing that canbe said definitively about this third category isthat short-term fluctuations can (and probablywill) be profound. In this third category, it mightbe reasonable to establish that adjusted assets willequal, say, 60% of their market value.

To summarize, adjusted assets might be definedin three categories, Defining investment gradedebt as Moody’s AA, the categories might bedescribed thus:

Category Ratio of AdjustedValue to

Market ValueDuration-matched 100%investment-grade debt

Other investment-grade debt 90%

All other assets 60%

These percentages do not, in any way, reflect anattempt to smooth changes in asset values. They

PFN0504.qxd 5/3/05 8:40 AM Page 22

THE PENSION FORUM

23

simply reflect an acknowledgement that asset values do fluctuate. If an employer should failshortly after its pension plan has been subjectedto an annual test of funding adequacy, the rela-tionship between assets (at market) and liabilitiesmay have deteriorated.

This approach of discounting the value of certaintypes of plan assets has been the subject of someconsiderable criticism. The critics point out thatmarket values are determined in the marketplace,and it’s inappropriate to second-guess this deter-mination. But the exercise is not to find appropri-ate “true” values. Instead, the exercise is to estab-lish a method for ascertaining, with a reasonabledegree of assurance, that fluctuations in marketvalue will not cause values to fall below a level suf-ficient to provide expected benefits.

It seems worth pointing out that the same resultcould be obtained without mandating the dis-counting of assets. Instead, a margin or bufferzone could be required. This margin would be theamount by which the market value of assets mustexceed the value of accrued benefits. The degreeof excess could be related to the nature of theplan’s investments. To the extent investments arein duration-matched investment-grade debt secu-rities, a buffer zone might be deemed unnecessary.To the extent assets are in other investment-gradedebt, assets could be required to exceed accruedbenefit values by 11%. For all other assets, therequired margin could be 662⁄3%. The result ofthese surplus requirements would be identical tothe result of discounting assets.

This whole notion is not exactly a novelty. Duringthe 1980s there was a certain amount of activityin participating group annuity contracts with cus-tomer-selected investments. Investment results,determined explicitly by performance of the cus-tomer’s selected portfolio, were credited to thecustomer’s account. Pensioner reserves were calcu-lated using standard insurance company proce-dures. However, it was a requirement that assets

must exceed reserves by specified percentages.These specified percentages were determined in amanner analogous to the asset discount proce-dures being suggested here.

Accrued Benefits Defined

Establishing an accrued benefit definition to beused in the funding rule seems reasonablystraightforward. Congress and the regulators haveestablished a definition of accrued benefits to beused in determining whether a plan has sufficientassets to qualify for plan termination on a nondis-tress basis. From time to time, changes in this def-inition are proposed. For example, ongoing atten-tion is being given to the question of protectingdeath and disability benefits not considered partof the accrued benefit.

However, it appears reasonable at this time todefine accrued benefits as those benefits that mustbe covered by a sponsor wishing to terminate itsplan on a nondistress basis.

Valuing Accrued Benefits

Valuing accrued benefits requires assumptions asto interest, mortality, and expenses. Whereoptions such as early retirement or alternativebenefit forms are subsidized, it also requiresassumptions as to the likelihood that theseoptions will be exercised.

If accrued benefits are to be protected, the assump-tions used to value them need to satisfy this rule:

Use of the assumptions must produce liabilityvalues at least equal to the premiums thatwould be required under a contract availablefrom the commercial insurance industry to pro-vide paid-up annuities covering all accruedbenefits.

An annuity contract providing such benefits isoften described as a “group closeout annuity

PFN0504.qxd 5/3/05 8:40 AM Page 23

THE PENSION FORUM

24

contract.” There are a number of alternativesavailable to satisfy this assumptions rule.

PBGC Rates

The Pension Benefit Guaranty Corporation hasdeveloped procedures for determining and updat-ing assumptions that satisfy the rule. Theseassumptions are used for a number of purposes. Aprimary purpose is to value liabilities of plansundergoing “distress termination.” A distress ter-mination occurs when a plan has assets insuffi-cient to cover accrued benefits and is being termi-nated because the sponsor is bankrupt or suffer-ing extreme financial hardship.

The PBGC collects information each quarterfrom the insurance industry on the rates then incurrent use for group closeout annuities.Individual insurance companies have proprietaryinterests in maintaining the confidentiality oftheir current rate offerings. To protect these pro-prietary interests, information is furnished in away that masks the identity of each individualcompany and its rate bases.

From time to time, observers have compared liabilities based on these PBGC assumptions with premiums actually charged to plans termi-nating on a sufficient basis. These comparisonslead to the conclusion that the PBGC’s proce-dures for keeping its rate basis current areextremely effective. On balance, the comparisonshave shown remarkably little variance betweenliabilities based on PBGC rates and premiumsunder actual contracts.

So one approach to statutory funding is to man-date that accrued benefit values be based onPBGC rates for distress terminations.

A Procedure Parallel to the PBGCProcedure

Many observers have expressed concern over astatutory requirement that minimum funding

must always be based on PBGC rates. They pointout that times change, and PBGC rates may notalways be as closely related to commercial annuityrates as they are today.

An alternative procedure would be to establishmachinery that parallels the PBGC machineryand independently maintains an up-to-datestatutory rate basis. The organization or agencyresponsible for administering this machinerywould need to be one that has the trust of theinsurance industry. Members of the industrywould be understandably concerned if there wereany suspicion that confidentiality might bebreached. The organization would also need theconfidence of the regulators. Either the AmericanAcademy of Actuaries or the American Society ofPension Actuaries might be suitable.

A Statutory Interest Rule

Still another option would involve one treatmentfor the interest assumption and a different treat-ment for all other assumptions.

The interest assumption generally attracts moreconcern than the others do. Plan sponsors areconcerned that the rate (or rates) might be too low. Entities representing pensioner interestsare concerned that the rate (or rates) might be too high. Both factions might be more comfortable with a rule that’s automatic andeliminates discretion.

A rule that produces automatic results mightinvolve reference to a well-publicized index. Theindex might reflect swap rates. It might reflectbond rates used by an established mortgageagency such as Fannie Mae. Or it might reflectrates maintained by a nationally recognized commercial rating agency such as Moody’s orStandard & Poor’s.

The relevant rate would not necessarily be 100%of the index. It could be a fixed percentage of anindex if the consensus is that the index will vary

PFN0504.qxd 5/3/05 8:40 AM Page 24

THE PENSION FORUM

25

in sync with the interest rates underlying insur-ance company premium rates. The index itselfmight consistently be a fixed number of percent-age points higher or lower than the insurancecompany rate basis. The relevant rate could bedefined as a yield curve, matching shorter bonddurations to liabilities with shorter duration,and longer to longer.

It doesn’t appear feasible to establish a compara-ble automatic procedure for the other assump-tions. Attempts to establish automatic proce-dures respecting mortality assumptions have notproduced satisfactory results. For example, onestatutory mortality base is keyed to the ratesmandated for the valuation of annuity reserveswhenever a new table is mandated by a majorityof the 50 states. The problem is the time lag. Thenew table must be mandated by 26 states, andthe federal regulators must acknowledge themandate. While the world waits, the old tableremains in continued use long after it has becomedangerously obsolete.

One solution would involve an automatic proce-dure for the interest assumption, and a joint pub-lic/private committee empowered to update theother assumptions. The joint committee mightconsist of representatives from the private-sectorpension actuarial community together with repre-sentatives from the Treasury, Department ofLabor, IRS, and PBGC.

A Mistake to Avoid

At present, serious discussions are underwayregarding an automatic procedure for the interestassumption. At the same time, some participantsin the discussions are apparently assuming thatupdating the mortality assumption is not a prior-ity item. We could, indeed, get along for sometime with an obsolete mortality table. However,to do so safely we would need an offsetting adjust-ment to the interest rate. The interest rate wouldneed to be reduced to offset the inadequacy of themortality table. This need appears to have gone

unrecognized. Indeed, the PBGC has been chas-tised for its use of unrealistically low interest rates.The low interest rates are entirely appropriatewhen viewed as devices to offset the obsolete mor-tality table currently mandated.

Meeting Sponsor Needs for Smoothness

The statutory funding rules outlined here require just enough funding to ensure that bene-fits already earned will not be lost. They leave no room for smoothing. Amortization periodsand the use of averages are not part of the proposals. If these proposals were adopted, thesponsor who consistently contributes justenough to satisfy statutory requirements wouldbe in for a rough ride. The typical sponsor wouldfind this unacceptable.

Suppose an employer’s objective is a pension cost factor that’s a stable percentage of payroll ora stable amount per employee. In almost everysituation, assurance that this stability will occurwould require funding levels exceeding the statu-tory minimum level. Therein lies the secret. The sponsor seeking smoothing will elect to fundat a level greater than the statutory minimum.This sponsor can then be relatively indifferent to any lack of smoothing in the statutory minimum levels.

Statutory Rules to Accommodate HeavierFunding

The tax code and ERISA currently offer two road-blocks to this higher level of funding. First is thelimit on deductible contributions—and its com-plement, the excise tax on nondeductible contri-butions. Great strides have been taken in recentyears to make this a less serious roadblock, butmore needs to be done. We need to redouble ourefforts to persuade legislative planners that sub-stantially liberalized deduction limits for contri-butions to defined benefit plans do not constitutetax giveaways. We need to focus these legislativeplanners on the concept that with defined benefit

PFN0504.qxd 5/3/05 8:40 AM Page 25

THE PENSION FORUM

26

plans the long-term deductible cost is dictated bythe plan’s provisions. Amounts contributed anddeducted today will not be contributed anddeducted again tomorrow.

The second roadblock is more difficult, and itselimination will face greater opposition. Considerthe funding standards proposed here. There willbe a willingness to go voluntarily beyond the lev-els dictated by these standards if sponsors can begiven two new privileges:

l

First is the right to make trust fund with-drawals at will. This withdrawal right shouldapply to any amount by which assets exceed thenew minimum funding levels. As will be dis-cussed shortly, a withdrawal tax is appropriate,but it should not be punitive.

l

Second is the right, upon plan termination andafter all obligations have been satisfied, to with-draw any remaining assets. This too shouldinvolve a withdrawal tax, but not a punitiveone.

Consider withdrawals before plan termination.Current law forbids this-and with good reason.Under current funding standards, following therules does not provide an absolute guarantee of ter-mination solvency. If experience losses occur, cur-rent rules allow time to restore the balance. Theproposed standards don’t provide absolute guaran-tees-but they come much closer. And, when experi-ence losses do occur, the balance must be restored atonce. Fairness dictates that if shortfalls must be cor-rected at once, sponsors should be allowed to cor-rect overages to the extent they see fit.

Consider reversions upon plan termination. In a cynically conceived series of political decisions,we have allowed ourselves to become confusedover the status of excess plan assets. The spon-sor’s job is to provide benefits as promised.There’s no room for the notion that assetsbeyond those amounts needed to perform thisjob belong anywhere but back in the hands of

the sponsor. Our decision to impose punitiveexcise taxes on reversions has played an impor-tant role in weakening the funded status ofmany plans. Under current rules, no rationalsponsor will intentionally permit assets to exceedtermination solvency levels for any extendedperiod. The excise tax that would occur in eventof an unexpected need for plan terminationwould be too painful. The existence of this taxhas led to corporate combinations that wouldhave been deemed ill advised if not for the factthat they involved locked-up pension assets.

This is not to say that asset withdrawal taxes haveno role. Reference was made earlier to their legit-imacy. But their sole purpose should be to reversethe tax advantages that accrued while the with-drawn assets resided in the tax-exempt trust. Suchtaxes should apply whether the withdrawal isfrom an ongoing plan or a terminating one.

With these changes-higher deductible limits and access to excess trust assets-sponsors are likely tolook favorably on the additional funding necessaryto permit a smoothing of contributions. They’ll alsofind the asset adjustment aspect of the proposedminimum funding rules more palatable, knowingthat upon plan termination assets will be applied toprovide benefits, 100 cents on the dollar, and thesponsor will recover any surplus.

The sponsor who seeks contribution smoothingwithout exceeding minimum funding standardsdoes have another option. Much of the volatilitythat would be brought about by eliminatingsmoothed standards could be regained throughinvestment policy. By emphasizing investment-grade debt obligations, duration-matched to planliabilities, contribution volatility related to assetfluctuation could be virtually eliminated.

The important thing is that the sponsor, workingwith the plan actuary, will be able to focus onlong-term cost trends and accomplish

PFN0504.qxd 5/3/05 8:40 AM Page 26

THE PENSION FORUM

27

contribution smoothing to whatever extent thesponsor deems desirable.

Accounting Concepts to AccommodateHeavier Funding

Changes in the statute would be significant inencouraging heavier pension funding. Anothersignificant factor would be changes in how finan-cial analysts view a sponsor’s pension funding.Consider the possibility that financial analystsmight fully accept two concepts:

l

For accounting purposes, pension assets shouldbe viewed as assets of the employer.

l

For these same purposes, investment results onthese assets should, indeed, be reflected atonce, without smoothing. But they should notbe viewed as affecting results from operations.Rather, they should be viewed as items appear-ing “below the line.”

Acceptance of these concepts, coupled with thestatutory changes already discussed, would go farin eliminating disincentives for heavier funding.The advantage of heavier funding in terms ofsmoothing cash-flow demands would thenemerge as a powerful incentive without signifi-cant offsetting disincentives.

New Plans—and LiberalizingAmendments

Sponsors of existing plans can achieve smoothedcontributions by maintaining funding levels thatexceed the statutory minimum. However, thereremains the problem of a feasible approach tonew plans.

Consider the sponsor who establishes a new planproviding significant benefits for past service.Immediate compliance with the statutory fundingrules outlined here would require an initial contri-bution that most sponsors would find totallyunacceptable. Sponsors seeking to increase benefitsunder existing plans would face the same problem.

Providing a Temporary Unfunded Benefit

A solution to this problem would involve initialestablishment of a temporary unfunded plan.

These rules might apply:

l

The plan would not be permitted to remain ineffect for more than, say, five years.

l

Throughout the lifetime of this unfunded plan,employee notices would be required each year.These notices would state that l

The plan is unfundedl

There are no PBGC guarantees andl

The sponsor may terminate the plan and revoke all unpaid benefits at any time.

l

Throughout the lifetime of the unfunded plan,the sponsor would be permitted to contributeto a tax-exempt trust designed to fund benefitsupon termination of the unfunded plan.Deduction limitations respecting such contri-butions would be based on the benefit struc-ture of the unfunded plan.

l

At the end of the five-year period, or anytimesooner at the sponsor’s option, the sponsorwould need to discontinue the unfunded planand either abandon it completely or provide itsbenefits through the funded trust.

This temporary unfunded approach would beavailable to new plans. It would also be availablefor any benefit provided as an addition to anexisting plan.

Advantages of the Temporary UnfundedApproach

Of all the proposals set forth here, the temporaryunfunded plan almost certainly will be the mostcontroversial. The notion of a plan covering abroad spectrum of employees with no require-ment that there be assets backing up benefits will,indeed, require some thought.

PFN0504.qxd 5/3/05 8:40 AM Page 27

THE PENSION FORUM

28

However, consider the advantages:

l

Broken promises almost completely will beremoved from the picture. During the tempo-rary existence of the unfunded plan, there willbe no promises to break. Once the funded planreplaces the temporary one, funding standardswill virtually guarantee payment of all accruedbenefits.

l

Employees will fully understand their status.During the lifetime of the temporary plan,annual notices will communicate a very simplemessage: there are no promises. Unpaid benefits are subject to complete and retroactiverevocation. Once the temporary plan isreplaced by the permanent one, employees will have, with almost no possibility of excep-tion, the same assurance that participants infunded plans always think they have: full guarantees that benefits will not be lost due toemployer failure.

l

It will become possible to set PBGC premiumsat extremely low levels. There will be very fewcircumstances where plan assets will be insuffi-cient to pay promised benefits.

Shutdown Benefits

The unfunded arrangement could also serve as a roadmap for solving a problem that has plaguedplanners for many years: shutdown benefits.These are benefits that will never becomepayable—provided the sponsor never closes itsdoors. The proviso establishes the problem. Withmost plans, the likelihood that the sponsor will close its doors in the near future is remote.It’s less remote if, for example, the sponsor hasmultiple locations or plants. In these cases asponsor may decide to close some of its doors butnot all of them. It’s the multiple-plant scenariothat’s troublesome.

Given the remote likelihood that even some of the doors will really close, the plan actuarywill, quite properly, assign a realistically lowprobability to the likelihood that shutdown

benefits will ever be paid. Multiply the value ofthe benefits payable if shutdown should occur bythe probability that it will occur, and the result isa very low estimated liability. This means thatfunding against this liability is likely to be total-ly inadequate if even some of the doors shouldreally close.Suppose the shutdown benefit is part of a planproviding routine retirement benefits—benefitsnot contingent on shutdown. Actual shutdowncan precipitate payment of shutdown benefitsthat far exceed the reserve the actuary had estab-lished. The result can severely compromise thesecurity of benefit expectations of employees notaffected by the shutdown.

Assuming no insurance company will underwritethe risk of voluntary shutdown at anything closeto a reasonable premium, there’s only one way tosolve the problem. That is to establish the shut-down benefit as a completely separate plan andspecify that assets of the plan providing routineretirement benefits cannot be used to meet shut-down obligations. Sponsors will be unwilling tomake advanced funding contributions to a planproviding only benefits that are consideredunlikely to be paid. Hence, the unfunded planestablished on a temporary basis for nonshutdownbenefits will be a natural permanent vehicle forshutdown benefits.

Alternatives to Stronger FundingProposals

Clearly the rules suggested so far are not the onlyanswer. The rules proposed here don’t eliminatedependence on the PBGC, but they greatlyreduce it.

An alternative would be to move in the oppositedirection. We could increase our dependence onthe PBGC by eliminating all statutory fundingrules. We would define, in much more detail, therisk-related premium structure necessary to per-mit the PBGC to make up the shortfall wheneveran insufficient plan should terminate.

PFN0504.qxd 5/3/05 8:40 AM Page 28

THE PENSION FORUM

29

The amount of exposure would be evaluated usingtools similar to the tools we currently use to cal-culate variable premiums. Premium rates appliedto this exposure would be determined much moreprecisely than variable premium rates are nowdetermined. These new rates would be set at lev-els adequate to cover the risk. In general, theserates would be somewhat higher than the one cur-rently in use. Even more importantly, sponsoringemployers would be underwritten individually.Each sponsor would be assigned to a rating classon a basis that reflects the likelihood that the par-ticular sponsor will become incapable of fulfillingits pension promises.

Some observers point out that this type of ratingis already taking place in the private sector, where,for example, insurers routinely write performancebonds. Others argue that the underwriting of per-formance bonds has never been truly successful.They point out that most insurers are willing toinsure performance only, respecting those compa-nies that are so strong the likelihood of failure isvirtually nil. Still others will react unfavorably tothe notion that a governmental agency shouldenter the business of evaluating the creditworthi-ness of a private business. If the equivalent of per-formance bonds underwritten in the private sec-tor turns out to be feasible, the answer here maylie in transferring the role of the PBGC entirely tothe private sector.

Conclusion

The changes suggested here represent majordepartures from current rules. Existing players—sponsors, bargaining agents, rule makers, andexpert advisors—are likely to have difficulty, ini-tially, with the notion that changes this extremeare sensible.

In evaluating these proposals, two questions seemessential:

l

First, we must ask ourselves, again, why itmakes sense to have any externally imposed

funding rules. It is suggested here that the rea-son, the only reason, is to protect workers at alltimes from losing the pensions they’ve earned.

l

The second question is just as important, butless obvious. Suppose we didn’t currently havestatutory minimum funding requirements.Suppose we were writing rules where noneexisted. Knowing what we now know, what setof rules would we write?

There’s clearly a case for avoiding radical shifts.Gradual change often works better than precipi-tous shifts. But even with gradual change, it’simportant to establish a focus on where we’deventually like to be. With this focus, it becomespossible to make incremental changes withoutlosing direction.

This long-term target might run along the linesproposed here, in a way that minimizes the needfor a PBGC. It might run along alternative lines,also discussed here, in a way that would increasethe role of the PBGC. This alternative approachwould eliminate funding requirements and sub-stitute a more highly developed plan terminationinsurance structure. Or the target might be some-thing that falls between these two approaches.

The casual observer might say the in-betweenapproach is what we have now. This conclusionwould be wrong. Even with the in-betweenapproach, we need to reevaluate our methods ofdetermining adjusted assets, our methods ofdetermining liabilities, our approach to smooth-ing, and our methods of determining PBGC pre-mium rates.

In any event, the important thing is to define thelong-term goal. With the goal defined, the stepsto achieve it can be developed in a rational man-ner. We can thus develop the most efficientapproach to securing the pension expectations ofour nation’s workers—with the smallest possibleintrusion into the funding practices of ournation’s pension plan sponsors.

PFN0504.qxd 5/3/05 8:40 AM Page 29

THE PENSION FORUM

30

The pension actuarial community has been in theprocess of revisiting the fundamental principles ofpension actuarial science. Bader (2001) and Baderand Gold (2003) have raised important questionsabout the validity of the actuarial pension model.They have urged the profession to undertake amajor revision of the model in light of financialeconomics. The works of Bader and Gold, as wellas several other actuaries and economists, havebecome the subject of numerous discussions.

The paper of Bader and Gold (2003) offers acomprehensive list of grievances that the finan-cial economics community has had with variousmethodologies utilized by the pension actuarialcommunity. The negative role of ERISA enact-ment, numerous shortcomings of statementFAS87, the importance of understanding of finan-cial economics-these and several others points arevery well taken. However, the paper contains several declarations that should be disputed.

The actuarial pension model certainly needs further development, but it needs no reinvention.Misguided public policies, not actuarial models,produce meaningless and burdensome regula-tions. Faulty assumptions and unwise compro-mises, not deficient actuarial thinking, lead toopaque financial reporting. The alleged inabilityof the actuarial model to incorporate the emerg-ing science of financial economics has played no role in creation of the system that very fewconsider reasonable.

The purpose of this paper is to demonstrate thatthe actuarial pension model is in complete harmo-ny with the principles of the financial economics.The model is perfectly capable of answering thequestions raised in Bader (2001) and Bader andGold (2003).

Ultimately I believe that the actuarial pensionmodel will be reaffirmed as what it really is: avaluable quantitative methodology and an inte-gral part of financial economics.

1. Actuaries vs. Economists

Financial economists have been unhappy withthe contents of various actuarial reports for quitesome time. Actuaries have successfully ignoredthose complaints for quite some time as well. Tome, this story began when Lawrence Bader published a short paper titled “The Model HasNo Clothes” in 2001. I thought that the paperwas provocative enough to attract more atten-tion. It didn’t start a debate, but it came close.Finally, the article of Lawrence Bader and JeremyGold (2003) received broad attention and trig-gered a discussion that was later called “TheGreat Controversy.”

The paper jump-started a broad review of thefundamentals of the pension practice. Writtenwith great eloquence and style by authors ofimpeccable credentials, the paper contains a com-prehensive list of problems that the authorsbelieve “caused widespread, though rarely recog-nized, damage to pension plan stakeholders.” Thepaper presents several principles of corporatefinance that, as the authors believe, are applicableto pensions and assert that those principles are“almost … universally violated by the actuarialmodel.” Then the authors submit an extensive listof violations of the stated principles by currentactuarial practices. The very existence of thoseviolations makes the authors believe that they“have laid out the case for the obsolescence of theactuarial pension model.”

I disagree. I believe the “case for the obsoles-cence of the actuarial pension model” is based

Reaffirming Pension Actuarial Science


PFN0504.qxd 5/3/05 8:40 AM Page 30

THE PENSION FORUM

31

on questionable interpretation of financial eco-nomic principles in Bader and Gold (2003). Thegoal of this paper is to challenge several miscon-ceptions that lead to dubious conclusions aboutthe pension actuarial model.

2. Statement FAS87 Procedures

For the purposes of this paper, I arrange the argu-ments presented in Bader and Gold (2003) intotwo categories. The first category contains theauthors’ analysis of the existing procedures in thestatement FAS87. The second category containsthe authors’ questioning of pension actuarial sci-ence in general. In my view, the second categoryis much more controversial and important thanthe first one. However, it is convenient for me todeal with FAS87 first.

Bader and Gold (2003) make a great case againstmultiple risk-concealing procedures in FAS87.No disagreement here. Eliminate unnecessarysmoothings, unwarranted assumptions, andextended amortizations. Granted. The fair valueof liabilities should utilize the entire yield curve.Granted. Financial reporting must be transparentand “marked-to-market.” End of issue.

This part of the discussion is short and easy forseveral reasons. For the most part, FAS87 wascreated for the accountants and by the account-ants. Most actuaries never liked FAS87 to beginwith; FAS87 doesn’t serve the purpose it was cre-ated to serve; the investing public is increasinglyaware that the pension “earnings” should be sep-arated from the “core” earnings; therefore, nopension actuary should shed a tear when FAS87replaces its nuts and bolts. However, we shouldbe mindful of the potential damage that unrea-sonable and burdensome reporting requirementsmay inflict on the pension industry.

I would like to emphasize that the goal of thispaper is to defend pension actuarial science. Iadvance no arguments on the subject of whatmakes a good financial statement or a good public

policy. Those matters require more comprehensivetreatment than is suitable for this paper.Throughout this paper, when I indicate that a cer-tain idea on the subject of better reporting has mer-its, I don’t necessarily support the idea. I justacknowledge that there are good arguments for theidea that should not be taken lightly; there may beequally good arguments against the idea.

3. Actuarial Science vs. Regulations

It is important to separate pension actuarial sci-ence and its various implementations, augmenta-tions, and simplifications in existing actuarialpractices. Bader and Gold equate some proce-dures that practicing actuaries have to follow withpension actuarial science. It must be made clearthat the tenets of ERISA, Internal Revenue Code,and Financial Accounting Standards do notdefine or describe pension actuarial science.Actuarial textbooks and articles do. Several regu-latory and administrative organizations, includingthe IRS, PBGC, DOL, and FASB, may have con-flicting objectives and require applying the pen-sion actuarial model in questionable ways. It isunreasonable to blame a quantitative methodolo-gy called “pension actuarial science” for misguid-ed regulations that utilize that methodology.

A sizable majority of pension actuaries work inthe area of traditional pension valuation. That’swhere the demand is. That’s where the jobs are.The main responsibility of an actuary in that areais to certify the plan’s compliance with relevantregulations. Most pension actuaries are perfectlyconscious of the fact that those regulations areimperfect, to put it mildly. The adjective “insane”is not unusual when the actuaries discuss the regulations. Most “imperfections” of the existingregulations are well known.

Some of those imperfections are presented in Bader and Gold (2003). For the most part, I’msympathetic to the paper’s criticism of certainexisting practices. The authors lose me when theydeclare “The insights of financial economics have

PFN0504.qxd 5/3/05 8:40 AM Page 31

made our science obsolete.” I believe that decadesof wonderful developments in financial econom-ics as well as decades of disastrous legislative cre-ativeness have not made pension actuarial scienceobsolete. Moreover, I think that financial econo-mists would be well advised to learn the insightsof pension actuarial science when they ventureinto the areas that require forward-looking analysis.

4. FAS87 Bias

I would like to identify and discuss a commontendency in numerous publications devoted tothe asset-liability analysis of pension plans. For alack of a better term, I call it “FAS87 bias.” Iloosely define “FAS87 bias” as a set of the follow-ing assumptions.

1. The matching asset always exists. The match-ing asset is a portfolio of marketable securities(usually bonds) that has the same (or reason-ably close) payouts—in terms of timing, mag-nitude, and probability—as the benefit pay-ments of the pension plan.

2. There is only one legitimate concept of pen-sion liability: the one that’s equal to today’sprice of the matching asset.

3. Pension commitment1 is similar to debt.

Why do I believe that these assumptions are relat-ed to FAS87? FAS87 is a brainchild of theaccountants who deal with pensions. The defini-tion of liability under FAS87 is based on assump-tions 1 and 2: the liability is equal to the price atwhich the pension commitment can be settled, ordefeased. Assumption 3 represents a viewpoint ofaccountants and lenders (and, possibly, someinvestors)—very important groups of stakehold-ers with a vested interest in FAS87.

The proponents of the “FAS87 bias” may or maynot agree with the existing procedures employed

in FAS87. They may or may not restrict them-selves to the past service only, as in FAS87. Theymay or may not agree with the assumption inFAS87 that the promised benefits will almost cer-tainly be paid, and, therefore, the pension obliga-tion should be valued as a high-quality bond port-folio. Nonetheless, I believe that the proponentsof the “FAS87 bias” have, for the most part, thesame perspective as the creators of FAS87: theaccountants. What unites them is the belief thattoday’s asset prices are all we need to know. Inparticular, they believe that the only proper wayto discount future cash flows is to use today’s yieldcurve. However imperfect, I think the term“FAS87 bias” has some useful connotationsbehind it.

Here is my view of the assumptions that definethe “FAS87 bias”:

1. The Matching Asset. The existence of thematching asset is far from certain. The exactlymatching asset rarely exists. The only hope isto find a “reasonably close” matching asset,which is problematic even if we assume thatthe demographic assumptions are flawless.One of the reasons for that is today’s fixed-income instruments are not long enough for aportfolio that provides complete dedicationfor a conventional pension commitment. Foranother reason, I’m not convinced that there isa “reasonably close” matching asset for a bene-fit stream that is based on the five-year averageof the wage inflation in a particular industry.

2. The Liability Concept. The fact that there arenumerous liability figures out there is a reflec-tion of the fact that a well-organized pensionplan management involves numerous respon-sibilities. Different tasks may require differenttypes of liabilities. To comply with reportingrequirements, the plan must submit the priceof the matching asset (assuming it exists) as

THE PENSION FORUM

32

1 As introduced in Mindlin (2003), a pension commitment is the stream of benefit payments determined by the plan’s population andbenefit package.

PFN0504.qxd 5/3/05 8:40 AM Page 32

THE PENSION FORUM

33

the liability. Compliance with the reportingrequirements is one of the responsibilities of apension plan, but there are others. For example,one of the most important fiduciary responsi-bilities of a pension plan is to allocate theassets prudently in the best interests of theplan participants. That task may necessitateutilization of a different concept of liability.

3. Pensions vs. Debt. Here is a partial list of objec-tions. (a) If inflation rises unexpectedly, someliabilities rise as well. On the other hand, thevalue of debt always declines. (b) The similar-ity between pension commitment and debt isnot absolute. Pension commitment may besimilar to debt as far as some lender or investoris concerned. On the other hand, pension com-mitment and debt may be very dissimilar asfar as the plan participants are concerned. (c) Itmight be a good public and/or accountingpolicy to disclose pension obligation and debtin the same way. On the other hand, it may bea good corporate policy to treat pension obli-gation and debt differently. For instance, froma legal standpoint, a dollar owed to a pension-er must be prefunded in the best interests ofthat pensioner; a dollar owed to a creditorshould be managed in the best interests of theshareholders, possibly on a “pay-as-you-go”basis.

In short, “FAS87 bias” ignores some very impor-tant responsibilities and relationships that areindispensable parts of the pension industry

Identifying the roots of the “FAS87 bias” is noeasy task. Here is how McCrory and Bartel(2003) deal with a similar challenge: “If the finan-cial community wishes to regard pensions as debt,this is not an indication of any deep thought orarcane knowledge. Instead, it is just a natural ten-dency of people to extend concepts with whichthey are familiar to new situations, even when the

fit between the existing concepts and the new sit-uation is imperfect.” McCrory and Bartel (2003)may be on to something.

Bader and Gold (2003) is clearly “FAS87 biased.” The authors appear to view the pensionindustry from the standpoint of a single-employer corporate pension plan and its flagshipreporting document—the Statement of FinancialAccounting Standards No. 87. The paper dealsexclusively with the issues of “marked-to-market”financial reporting. A reporting—financial or anyother—is an inherently retrospective endeavor.Bader and Gold essentially argue that a properlydesigned statement FAS87 should have no for-ward-looking declarations. But expectations arethe core of the art of investing. They are just for-eign to the accounting mindset.2 This makes an“FAS87-based” and an “FAS87-biased” perspec-tive needlessly restrictive and incomplete.

Ultimately I believe that “The Great Contro-versy” will be viewed as an attempt of the“FAS87-biased” version of financial economicspresented in Bader and Gold (2003) to squeezepension actuarial science into the Procrustean bedof that version’s principles. I would like this paperto be considered as an attempt of pension actuar-ial science to resist the squeeze.

5. Principles Challenged

Of several great qualities of Bader and Gold(2003), I especially value its style and structure.The authors list several principles and analyzeexisting practices in light of those principles. Inparticular, that structure allows the reader to pindown the roots of any statement the reader findsquestionable. If one disputes a certain conclusionof the paper, the right way to do so is challengethe underlying principles. That’s exactly what I doin this section.

2 In fact, FAS87 contains several forward-looking assumptions. The most notorious of them is “long-term expected return on assets.”That assumption is highly controversial and has been criticized in a number of publications recently.

PFN0504.qxd 5/3/05 8:40 AM Page 33

THE PENSION FORUM

34

The heart of the discussion is Principle 4: “A liabil-ity is valued at the price at which a reference secu-rity trades in a liquid and deep market.” By itself,that statement is not incorrect. If one believes thatthe price of the matching portfolio is relevant, thenone should be at liberty to utilize that value as theliability. But once that statement is elevated to thestatus of a principle, it declares that no other meas-urement of the pension commitment can be legiti-mate (here is the “FAS87 bias” in action). This is abold claim that requires considerable justification.None is presented in the paper. The only potentialexplanation that I’ve been able to attribute to thePrinciple 4 is the law of one price. According toPanjer (1998), “two assets (or securities, portfolios,liabilities, and so on) with identical cash flows inthe future have the same current price in arbitrage-free market.” Note that both securities must betradable. Even if we assume that the matching assetexists, the stream of pension payments might beconsidered somewhat tradable only if the sponsoris about to settle its pension commitment by virtueof purchasing the matching asset or a group annu-ity contract.

The reality is that the overwhelming majority ofsponsors of ongoing defined benefit pensionplans choose not to settle their pension commit-ments at the present time. For better or worse,they invest in nonmatching assets. For thosesponsors, the pension commitment cannot beconsidered as a marketable security. Therefore,the law of one price is not applicable. To me,Principle 4, as presented in Bader and Gold(2003), is unsubstantiated.

However, I might have understood Principle 4 asan accounting standard. There might be merits inthe statement “A liability is disclosed at the price atwhich a reference security trades in a liquid anddeep market.” But the boundaries of accountingare too restrictive to accommodate interests of allstakeholders in the pension industry. There is lifeoutside of regulated reports in general and account-ing in particular. Different groups of stakeholders

should be allowed to measure the pension com-mitment according to their needs and risk toler-ance. For example, the “FAS87-biased” conceptof liability (see assumption 2) may be insufficientfor asset allocation purposes. As we’ll see inSections 7 and 8, some liability values are instru-mental in making the asset allocation decision,even though they don’t belong to any convention-al actuarial report and are calculated with no ref-erence to today’s yield curve.

In other words, financial economics may notrequire embracing the collective wisdom oftoday’s bond traders, as expressed in today’s yieldcurve, for all stakeholders for all purposes. Butfinancial accounting might require utilization oftoday’s yield curve for the reporting purposes.

One may argue that, in the case of a corporatepension plan, the plan is a part of the companyand, as such, gets priced every time the company’sshares are sold, and should be valued the sameway as a bond portfolio with similar payouts.That might be a good rationale for the purposesof transparent financial reporting. But pretendingthat the value of the company’s pension commit-ment has bondlike characteristics can be tremen-dously misleading. The “liability-is-a-bond”approach obscures the risks imposed on theinvestors by the sponsor’s decision to invest innonmatching assets. That approach is a risk con-cealment methodology by itself.

Let’s turn our attention to Principle 1 now: “$1 million of bonds has the same value as $1million of equities.” This statement appears tobe obvious. As we’ll see in the next section, thisprinciple is an easy corollary of basic principlesof pension actuarial science. The challenge is todetermine the meaning of the word “value.” Ifwe “value” the asset side only, the statement iscorrect. But if that is the case, then Principle 1changes the subject of the discussion and mis-represents the actuarial business. Pension actuar-ies are in the business of calculating pension

PFN0504.qxd 5/3/05 8:40 AM Page 34

THE PENSION FORUM

35

commitments and their present values. If wehave to “present value” a certain benefit stream,then a $1 million portfolio of bonds and a $1million portfolio of equities may produce con-siderably different results. While today’s valuesof the two portfolios are the same, their expecta-tions for the future may be quite different. Inthat context, Principle 1 is not so obvious.

In other words, the context is important.Principle 1 may be perfect for some accountingstatements. It is not so helpful when your job isto manage a variety of risks in the future.Principle 1 in the context of the asset allocationproblem may be interpreted as questioning therelevancy of the asset allocation. While asset allo-cation may be irrelevant as far as the plan’s audi-tor is concerned, it is exceedingly important for anumber of other reasons.

To me, Principle 1, while literally correct, is amanifestation of the “FAS87-biased” mindset thataccepts only one form of expressing our expecta-tions—as market prices of assets. This approachplaces severe restrictions on the methods at ourdisposal. I have serious reservations about theconceptual suitability and relevancy of Principle 1to this discussion.

Another problem is misinterpretation of thenotion of funding ratio. Here is an example fromBader and Gold (2003). Think of a pension planfor which the price of the Treasury matchingportfolio is $1 million. If the sponsor buys thatmatching asset, the funding ratio is 100%.However, a $628,000 equity portfolio wouldproduce a similar result (but only as far as the valuation report is concerned): the funding ratio is 100%; we assume that “equities are expected toreturn 10%.” The authors are unenthusiasticabout two different asset values ($1,000,000 vs. $628,000) that produce the same fundingratio. Based on Principle 1, the authors concludethat certain members of the plan “have beencheated of $372,000.” Speaking about the

$628,000 equity portfolio, the authors alsobelieve that “many pension actuaries wouldregard such a portfolio as fully funding the plan.” If that is true (which, to me, is extremely unlikely), those “many pension actuaries” areguilty of gross ignorance.

For the purposes of a conventional valuationreport, the plan’s actuary produces a schedule offunding progress. A liability is a measurementpoint on that schedule. The fact that the fundingratio is equal to 100% merely means the plan isright on schedule. The ratio provides no informa-tion about the riskiness of the schedule or thelikelihood of actually making the promised pay-ments; these matters don’t belong to the valuationreport. The considerations to invest in $1 millionof bonds or $628,000 of equities imply two dif-ferent funding schedules: one riskless and onerisky. There should be no surprise that thoseschedules require different amounts of assets as ofnow. “Fully fundedness” of a plan with the fund-ed ratio equal to 100% is little more than a con-fusing figure of speech.

In this example pension actuarial science providesno support to the decision to invest in equities. Ifthe decision makers for the plan believe that therole of asset allocation is to produce an acceptablelooking valuation report at minimum cost at thismoment, they are the ones guilty of gross igno-rance, not the actuary. In most cases the decisionto invest in the $628,000 equity portfolio is givento the actuary. Even if the intergenerational“cheating” does exist, pension actuarial scienceplays no role in its existence. In other words, don’tblame pension actuarial science for the decision toinvest in equities.

6. Some Principles of Actuarial Science

As I mentioned before, I happen to like the“from-principles-to-conclusions” structure inBader and Gold (2003). I attempt to use thesame structure in this section. Here are threefundamental principles that are universally

PFN0504.qxd 5/3/05 8:40 AM Page 35

THE PENSION FORUM

36

accepted among actuaries (or so I want tobelieve).

Principle 1: Present value of $1 at the end of aperiod of time is equal to the asset value thathas to be invested now to accumulate exactly$1 by the end of the period.

Principle 2: Present value of $1 at the end of aperiod of time is equal to 1 / (1 + R), where Ris the investment return for the period.

Dear reader, make up your mind now. Principles1 and 2 have profound consequences. Your sup-port or opposition to these principles will ulti-mately determine your position in this discussion.

It is important to note that if the return R inPrinciple 2 is uncertain, pension actuarial scienceempowers us to use our current expectations andmodel R as a random variable.3 If that’s the case,then the present value of $1 is a random variableas well. The concept of a random present valuebelongs to the mainstream actuarial thinking: seeBowers (1997) and Kellison (1991).4

Principle 3: Assumptions matter.This principle is as obvious as it is absolutelyessential. The assumptions must be disclosed andfollowed. Think of a checking account that guar-antees 1% annual return and a zero-couponTreasury bond that guarantees 2% annual return.In order to fund $1 at the end of the year, we need$0.99 if we choose to invest in the checkingaccount; we need $0.98 if we choose to invest inthe zero-coupon Treasury bond. If we choose toinvest in the checking account, the present valueis $0.99. It is incorrect to claim that the presentvalue is $0.98 if the money is in the checkingaccount. The fact that we have enough money to

purchase a readily available zero-coupon bond isimmaterial. Of course, we can always assume thatthe money is invested in the zero-coupon bond.Under that assumption, the present value is$0.98. But it is of crucial importance to makeclear that it is just an assumption. It must be dis-closed that in reality a different investment optionis selected.

As the first application of these principles, let’sconsider one of the statements in Bader and Gold(2003). Discussing a $1 million portfolio ofbonds and a $1 million portfolio of equities andtheir 10-year expected values, the authors declare,“Yet, the present values of the returns of the twoportfolios, when correctly discounted to reflectrisk, are equal.” Apparently, unimpressed by someimplementation of the actuarial pension model,the authors imply that the actuarial pensionmodel would have come to a different conclusion.Here’s how the actuarial pension model treats theproblem. If a portfolio of any assets is worth $Anow and R1, … , RN are the portfolio returns forthe next N years, then the investment return forthe N-year period is (1+R1) . ... . (1+RN)-1, and thefuture value of the portfolio is A . (1+R1) . ... . (1+RN).Then, according to Principle 2, the present valueis equal to A . (1+R1) . ... . (1+RN)

(1+R1) . ... . (1+RN)

regardless of the actual composition of the portfo-lio. In particular, $1 million of bonds has thesame value as $1 million of equities, which isPrinciple 1 in Bader and Gold (2003). The pen-sion actuarial model and financial economics arein complete agreement.

=A

3 See Klugman, Panjer, and Willmot (1998), section 2.1, principle 3.1.4 See Bowers (1997), chapters 3 and 4, for cases of deterministic discounting procedure and random payment timing, resulting

in random present values. See Kellison (1991), chapter 10, for a case of random discounting procedure and deterministic paymenttiming, resulting in random present value as well. Both Bowers (1997) and Kellison (1991) are standard actuarial textbooks.

PFN0504.qxd 5/3/05 8:40 AM Page 36

THE PENSION FORUM

37

7. Case Study

This section shows the principles of actuarial sci-ence in action as applied to a simple fundingproblem.

The plan has to make a single benefit payment of$100 20 years from now. The market value ofassets is $37.69. At the end of year 20, the com-pany will pay the shortfall or collect the leftoverassets. The assets are invested in a portfolio of 20-year zero-coupon Treasury bonds that pays exact-ly $100 in 20 years. We ignore taxes and transac-tion costs and assume that there’s no risk ofdefault by the plan sponsor. Essentially the sameproblem was presented in Bader (2001).

The assets and liabilities are exactly matched now.But let’s assume the plan sponsor has just movedthe assets from Treasuries to equities and intendsto keep the money there. Bader (2001) presentedthe following questions regarding the plan.

Question #1: How would shifting all the assetsfrom Treasuries into equities affect shareholdervalue?

Question #2: What discount rate should youuse: the Treasury yield, the expected return onplan assets, the company’s borrowing rate, thecompany’s weighted average cost of capital, orsome other rate?

Getting no satisfaction from his interpretation ofthe actuarial pension model, Bader also presentedhis third question.

Question #3: If traditional actuarial modelsand techniques stumble over questions of pen-sion cost and asset allocation for the simple casedescribed here, is there any reason to think that

they get it right for real-world pension plansand funding practices?

Here are the answers. The answer to Question 1is obvious—no change. The answer to Question2—it depends on the purpose of the calculation.For compliance purposes, use whatever isrequired by the regulations. Outside of the com-pliance, any fixed discount rate is wrong. Theinvestment return is not fixed for the choseninvestment option. The answer to Question 3 is“yes.” Yes, there are compelling reasons to thinkthat the actuarial pension model gets it right forthis case as well as “real-world pension plans,” asI demonstrate below.

For this case, I define the liability as the assetvalue that should be set aside now to fund thecommitment. According to Principles 1 and 2,the liability is equal to $100/(1 + R), where R isthe investment return for the period. Let’s assumethat the annual return of the portfolio of equitiesis lognormally distributed with the median 8%and standard deviation 17%.5 Then the liability islognormally distributed with the median 21.45and standard deviation 21.33. Keep in mind thatPrinciples 1 and 2 allow deployment of our cur-rent expectations for the capital markets, includingexpected return and risk, to calculate the liability.This liability reflects the risks the sponsor is exposed to on the asset side.6 This methodstands in a sharp contrast with another way to cal-culate the liability: to use the expected return as afixed discount rate, which eliminates all the riskson the asset side before the liability is calculated.

5 The conventional language is somewhat imprecise here. The assumption of lognormality of the investment return R usually meansthat 1+R is distributed lognormally, not R. That convention generally applies to investment returns only.

6 As defined in Mindlin (2003), this is the asset allocation-related liability associated with the pension commitment of $100 in 20years and the policy portfolio that has a lognormally distributed return with the median 8% and standard deviation 17%.

PFN0504.qxd 5/3/05 8:40 AM Page 37

THE PENSION FORUM

38

Let’s look at some risks and rewards the sponsorfaces now. If the sponsor can live with a 50%chance to fund the commitment, the fund musthave $21.45. If the sponsor wishes to have a 95%chance to fund the commitment, the fund musthave $66.90.

The market price for this pension commitment is$37.69—the price of the zero-coupon Treasurybond that pays $100 in 20 years. However, thefact that the sponsor can settle the pension com-mitment and all the risks associated with it (byvirtue of investing $37.69 in the matching asset)does not mean that it has to or wants to do so.The sponsor’s financial situation and risk aversioncharacteristics may allow the sponsor to take somerisk and enjoy (or regret) the results. In additionto the main objective of funding $100 in 20 years,the sponsor may have some other matters ofimportance (leftover assets, for instance).

Let’s deal with the actuarial report and statementFAS87 in this case.

A traditional actuarial valuation report wouldmost likely have the investment return assump-tion at 8% (= geometric mean return on assets),accrued liability $21.45, funded ratio 176%.7

Let’s not call the plan fully funded-there’s morethan a 20% chance that the fund will not haveenough money to carry out the promise.

Statement FAS87 would have discount rate 5%,ABO $37.69, funded ratio 100%. From theFAS87 perspective, the plan looks “fully funded.”The main problem with these figures is that it isan assumption8 that the assets are invested in thematching zero-coupon bond portfolio, while, in

reality, that’s not the case. The disclosure of$37.69 as ABO obscures the risks that the spon-sor has taken; there is, for instance, a 5% chancethat the sponsor is liable now for a series of addi-tional contributions that has the present value ofat least $29.21 (= $66.90 - $37.69) and, possibly,much higher.

Let’s conduct a very simplified asset allocationstudy. I want to make a minor modification toour example and assume that there are no assets inthe pension fund. The sponsor is considering itsinvestment options and an asset value to con-tribute now. To simplify the case, the only yard-stick for the policy selection is the probability tohave enough money at the end of the period. Tosimplify the case even more, let’s consider onlytwo investment options: the zero-couponTreasury bonds and the equities. If the sponsorwishes to have a 50% chance to have enoughmoney to make the payment, then the portfolioof equities looks better; it requires less upfrontmoney ($21.45 vs. $37.69). If the sponsor wishesto have a 95% chance to have enough money tomake the payment, then the Treasuries look bet-ter ($66.90 vs. $37.69).9

It is important to point out that the Principle 2-based liability concept is providing us with aquantitative methodology for the asset allocationproblem. The pension actuarial model doesn’tmandate a particular risk tolerance level. Giventhe risk level, the model provides quantitativetools to make an informed decision. Risk toler-ance is the key issue here—the issue that tradi-tional liability concepts can’t tackle.

7 Utilization of the expected return on assets as the discount rate is a common practice. I don’t think that mandating the use of theexpected return as the discount rate is a good idea. But it doesn’t mean that we should discard the expected return altogether. Tothe contrary, the expected return is a very important part of our expectations. It just shouldn’t always be used as the discount rate.

8 ASOP No. 27 specifically allows making an assumption about asset allocation: “the discount rate may be selected independentlyof the plan’s investment return assumption, if any. In such cases, the discount rate reflects anticipated returns on a hypothetical assetportfolio rather than on the plan’s expected investments”.

9 The breakeven risk tolerance is roughly 20.8%. In other words, if the risk tolerance is 20.8%, then both investment options requirethe same asset value.

PFN0504.qxd 5/3/05 8:40 AM Page 38

THE PENSION FORUM

39

8. Lessons from the Case Study

There are several important lessons to learn fromthe case study.

First, I would like to reiterate that, to realize apresent value, we have to be invested in the assetclass that delivers the return used in the presentvalue calculations. It is incorrect to claim that thepresent value of the commitment is $37.69 if themoney is in the equities. The fact that the spon-sor can buy the Treasury bond is immaterial. Thecommitment has the present value $37.69 only ifthe money is invested in the zero-coupon Treasurybonds.

Second, it may be important to know the “risk-free” market price of the commitment if you areconsidering your investment options. But if thesponsor has made the decision to invest in a non-matching portfolio of assets, the “risk-free” price islittle more than a “nice-to-know” figure. However,if the plan is required to report the “risk-free”price to a governing body, then the price is impor-tant. Compliance requirements may give signifi-cance to some figures that are irrelevant otherwise.But let’s be clear: the objective has changed. It’s nolonger the funding or asset allocation—it’s thecompliance. A figure produced for the compliancepurposes doesn’t have to be useful or even makesense in other areas. Of course, it would be greatif all the figures required to be reported by variousgoverning bodies were useful. It also would begreat if all the useful figures were required to bereported by various governing bodies. As we allknow, the reality is different.

Third, $66.90—the 95th percentile of the liabil-ity—was instrumental in making the asset alloca-tion decision, although $66.90 doesn’t belong toany conventional actuarial report. The accruedliability—one of the key elements of the valua-tion report—was not helpful at all in the assetallocation decision. The ABO was helpful, butonly because one of the investment optionsunder consideration was the matching asset.

9. The Controversy

Let’s deal with the most contentious part of “The Great Controversy.” In short, the problemis the following. According to widely acceptedactuarial practice on the funding side, higherequity allocation implies higher expected return,which, in turn, requires a higher discount rate,which, in turn, implies lower liability. Thus,higher equity allocation implies lower liability.That is controversial.

It is crucial to understand that the liability con-cept used in that logic violates Principle 2. A fixeddiscount rate is selected for the discounting pro-cedure, while the assets are invested in risky assets.The underlying liability may have a nontrivialprobability distribution, but the actuarial reportanalyzes its expected value only. Bader and Gold(2003) states: “The actuarial pension model dis-counts liabilities at the expected return on theassets held to fund those liabilities; it ignores therisk.” That statement is plainly incorrect. As I’vedemonstrated in the prior sections, the actuarialpension model doesn’t ignore the risk. That riskanalysis simply doesn’t belong to a conventionalactuarial report.

That is the core of the controversy. The assump-tion “the same returns in all years” in a conven-tional actuarial report is unacceptable for a prob-lem that involves comprehensive asset and liabili-ty sides. The fixed discount rate assumption pro-duces an internally inconsistent asset-liabilitymodel. The main problem is the benefit paymentsare discounted at actual returns on the asset sideand at a fixed discount rate on the liability side.To illustrate this point, think of a contribution of$X now. Imagine that the accumulated value of$X is used in N years to pay a benefit payment $Y.If R1, … , RN are the returns for the next N years,then

Y=X . (1+R1) . . . . . (1+RN).

PFN0504.qxd 5/3/05 8:40 AM Page 39

THE PENSION FORUM

40

Therefore, the present value of $Y in N years onthe asset side is

Y(1+R1) . . . . . (1+RN)

On the other hand, the present value of $Y in Nyears on the liability side is

Y(1+D)N,

where D is the fixed actuarial discount rate. Thediscounting procedures on the asset and liabilitysides are clearly different.

This is a classic “apples-to-oranges” case.10 Tomake the model internally consistent, one has touse the same discounting procedure on both the assetand liability sides. Since the discounting proce-dure on the liability side is simply inadequate fora meaningful risk analysis, we should calculate theliability by discounting the payment stream thesame way we do it on the asset side, namely, atactual returns. That’s exactly what we did in thecase study. The actuarial pension model providedus with liability concept that could deal with risk-related issues.

Why do the conventional actuarial reports look atthe expected values only (by virtue of utilizationof fixed discount rates)? Well, it is simple, and, atthe same time, the regulations allow the simplifi-cation of a fixed discount rate. The last thing thepension industry needs now is more complicatedcalculations. However, the resulting asset-liabilitymodel can’t handle any risk-related problem. Theconventional valuation and FAS87 reports havemultiple objectives, but the risk analysis is notone of them. Those reports are not designed tohandle asset allocation problems—their purpose

is to certify the plan’s compliance with relevantregulations. That fact is generally recognized inthe pension industry: the risks on the asset andliability sides are usually analyzed outside of theconventional reports.

The main conceptual error in Bader and Gold(2003) is that the authors try to solve a risk-relatedproblem by using the contents and concepts of theconventional actuarial reports. Having run intotrouble with those concepts, the authors turned tothe principles of financial economics for an answer.They conclude that it must be assumed that the fundis invested in the matching asset; the liability must bedefined as the price of that asset. The answer given bypension actuarial science is more comprehensive.11

If the fund is invested in the matching asset, the lia-bility is equal to the price of that asset. Otherwisethe liability doesn’t have a fixed value; it has a dis-tribution that reflects the risks on the asset side;that distribution should be of paramount impor-tance for the actuarial analysis.

Here’s the key difference between this paper andBader and Gold (2003) in a nutshell. Bader andGold see an imperfect actuarial report and blamepension actuarial science for its existence. I see animperfect actuarial report and blame the imper-fect actuarial report. That’s what “The GreatControversy” is all about.

10. Questions to Economists

I would like to present a few questions to the pro-ponents of the “financial economics” approach topension plans, as defined in Bader and Gold(2003).

10 Unless the goal is to project the contents of the actuarial report in 10 years. But that’s a completely different story.11 Mathematicians call it a generalization.

X=

PFN0504.qxd 5/3/05 8:40 AM Page 40

THE PENSION FORUM

41

Let’s forget about the regulations for a momentand think of a public pension plan that choosesnot to buy the matching asset (assuming it exists).

Is the current market price of the matching assetrelevant to a pension plan that is not interested inexercising that price? Do you know the perma-nent solvency requirement’s price tag?

Now think of a corporate pension plan that choos-es not to buy the matching asset. Let’s also assumethat the sponsor has no intention to terminate theplan or file for bankruptcy. The previous questionis still valid: Is the current market price of thematching asset relevant? However, it is not unrea-sonable to think that, as long as shares of the cor-poration are publicly traded, the value of the pen-sion plan is somehow built into the share price.But does that value always have to be the “termi-nation value” for all intents and purposes? Do theproponents of financial economics advocate valu-ing every segment of the company’s business on atermination basis? Do the proponents of financialeconomics believe the markets do it that way?Would an “economically enlightened” CFO sendher accountants to a local garage sale to determinethe “marked-to-market” value of used file cabinetsand water coolers?

But for the sake of argument, let’s assume that wedo want to value the pension plan on a termina-tion basis. Under the assumption of the plan ter-mination, are the proponents of financial eco-nomics prepared to price the following: loweremployee morale, cost of additional compensa-tion demanded by the employees in exchange forthe lost pension, cost of compliance with PBGCregulations, potential loss of key employees, andpotential cost of legal actions against the sponsor?Do the proponents of financial economics believethat the ABO already includes all the costs associ-ated with the termination?

11. After Bader and Gold (2003)

I would like to mention a couple of developmentsthat have happened after the publication of Baderand Gold (2003).

First, the debate has been defined as “PensionActuarial Science in Light of FinancialEconomics.” A separation of pension actuarialscience from financial economics is as fictional asit is unhelpful. I believe pension actuarial scienceis a part of broadly defined financial economics. Ifexperts in geometry had a symposium entitled“Geometry in Light of Mathematics,” whatwould you think about those “experts”?

Second, several publications have portrayed thepension actuarial community as being split intoso-called “traditionalists” and “financial econo-mists.” The “financial economists” are the onesthat agree with Bader and Gold (2003). The restare the “traditionalists.” Since the “traditionalists”are separated from the “financial economists,” the“traditionalists” appear to be against principles offinancial economics. Such a split is a myth; itwould leave no place under the sun for people likethis author.

Here’s where I stand.

1. I believe pension actuarial science is a finequantitative methodology.

2. I believe in principles of financial economicsas the science of financial markets, asset pric-ing, and related subjects.

3. I recognize numerous deficiencies in conven-tional actuarial reports and agree (for the mostpart) with their critique presented in Baderand Gold (2003).

4. I strongly disagree with the paper’s conclu-sions regarding pension actuarial science.

I think that a majority of pension actuaries andfinancial economists will eventually support mypositions. Or so I hope.

PFN0504.qxd 5/3/05 8:40 AM Page 41

THE PENSION FORUM

42

12. Final Remarks

Bader and Gold (2003) is a great paper. The actu-arial community should appreciate it for severalreasons. First, it started a stimulating discussion.The call for better understanding of the principlesof financial economics is well timed and sensible.Second, it made an important contribution to thedebate about better reporting procedures. Third,and most importantly, by virtue of being so good,the paper clearly demonstrated severe limitationsof the version of financial economics the authorspresented in the paper. The need for pensionactuarial science is greater than ever.

As far as the liability calculations are concerned,the pension actuarial community has a clearchoice: to embrace the “fair value of liability” con-cept outlined in Bader and Gold (2003) as “theliability,” or to recognize the multitude of chal-lenges the pension plan stakeholders face and themultitude of liabilities related to those challenges.Eventually the marketplace of ideas will sorteverything out.

Larry Bader titled his 2001 paper “The ModelHas No Clothes.” There’s little doubt that thename was inspired by the Hans ChristianAndersen fairy tale “The Emperor’s NewClothes.” I happen to greatly appreciate the anal-ogy between the emperor and the actuarial pen-sion model as well as Bader’s witty allusion. I alsobelieve that the fairy tale is a clever illustration toour story, although with one small exception. It isnot the emperor, but the emperor’s accountantwho has no clothes. The emperor is dressed mag-nificently, thank you very much.

References

Bader, L. 2001. “Pension Forecasts, Part 2: The Model Has No Clothes.” Pension SectionNews no. 46 (June 2001): 14-15. http://library.soa.org/library/sectionnews/pension/PSN0106.pdf.

Bader, L., and J. Gold. 2003. “ReinventingPension Actuarial Science.” The Pension Forum15(1): 1-13. http://library.soa.org/library/sectionnews /pension/PFN0301.pdf.

Bowers, N. L., et al, 1997. Actuarial Mathematics,2nd ed. Schaumburg, IL: Society of Actuaries.

Kellison, S. 1991. The Theory of Interest, 2nd ed.Homewood, IL: Richard D. Irwin.

Klugman, S., H. Panjer, and G. Willmot. 1998.Loss Distributions. New York: John Wiley & Sons.

McCrory, R., and J. Bartel. 2003. “ReinventingPension Actuarial Science: A Critique.” PensionForum 15(1): 17-23. http://library.soa.org/library/sectionnews/pension/PFN0301.pdf.

Mindlin, D. 2003. “Discount Rate Revisited.” In the online monograph “The GreatControversy: Current Pension Actuarial Practicein Light of Financial Economics Symposium.”http://library.soa.org/library-pdf/m-rs04-1_13.pdf.

Panjer, H., et al. 1998. Financial Economics.Schaumburg, IL: Actuarial Foundation.


PFN0504.qxd 5/3/05 8:40 AM Page 42

THE PENSION FORUM

43

I must begin by thanking Mr. Mindlin for hiskind remarks about “Reinventing PensionActuarial Science” (Bader and Gold 2003, here-after Reinventing). His generous comments andwell-organized exposition make me regret that Icannot agree with more of his arguments.

Is Mr. Mindlin Really a Friend ofFinancial Economics?

He begins by laying a foundation for harmonizingthe actuarial pension model with financial eco-nomics. He correctly observes that the FASB andERISA do not define actuarial pension science,and he professes admiration for “decades of won-derful developments in financial economics.” Hethen defines and criticizes an approach to pensionfinance that he calls the “FAS87 bias,” of whichhe finds Reinventing guilty.

Mr. Mindlin mentions that he uses “FAS87 bias”for lack of a better term. I will suggest a betterterm for the approach that he criticizes: “finan-cial economics.” In Reinventing, Jeremy Goldand I use the principles of financial economics toprice pension obligations in a rules-free environ-ment. We do not directly address FAS87 orERISA; we aim to identify the true economiccost of pensions to those who bear the costs—shareholders and taxpayers—and to illustrate thepoor decision making that can result from mis-measuring those costs.

Challenging Principle 4

In Sections 4-5 Mr. Mindlin challenges the finan-cial economics principles enumerated inReinventing. He disputes Principle 4, which statesthat the value of a liability is the market price of a“reference portfolio” that matches the liabilitycash flows. He asserts that such pricing by marketvalues is a “retrospective endeavor.” In fact, it is a

forward-looking approach, because market pricesreflect consensus valuations of future risks andrewards.

Principle 4 is not, as Mr. Mindlin states, based onthe law of one price. It is based on the simplernotion that like liabilities have like values, and adollar owed to a pensioner is very like a dollarowed to a creditor. A company should be indiffer-ent between those two obligations, although onemay be tradeable and one not.

Suppose that a company with a one-year borrow-ing rate of 3% must pay $103 in one year. Thatobligation has a value of $100 (ignoring taxes),whether it is owed to a creditor or a pensioner.The value remains $100 even if the companyinvests $95 in equities that it expects will sufficeto pay the obligation. (I assume that the compa-ny has enough other assets that the performanceof the equities doesn’t materially affect the defaultprobability.) Actuaries following ASOP 27 wouldprice the obligation at $95, while investors wouldpay $100 for it. An actuarial model says you canmake one-year bonds for $95 and sell them for$100 can fairly be called broken.

Mr. Mindlin may point out that pension liabili-ties are far more complex than this simple illustra-tion. True—but if the actuarial model is wildlyincorrect for a simple one-year obligation, whyshould we even bother to inquire how it handlesmore complex problems?

Mr. Mindlin refers to differences between pensionliabilities and debt that make it problematic toidentify a reference portfolio. One is the effect ofwage inflation on pensions. However, only theaccumulated benefit obligation, which is not sub-ject to wage inflation, falls within the definition

A Critique of Reaffirming Pension Actuarial Science

by Lawrence Bader, F.S.A.

PFN0504.qxd 5/3/05 8:40 AM Page 43

THE PENSION FORUM

44

of a liability (Bader and Gold 2003, Authors’Response, footnote 19; see also Bader 2003). Ofcourse, pension liabilities differ from bonds invarious ways. But marketed bonds carry maturi-ties from immediate to 100 years, default riskfrom 0% to 100%, and a wide range of collateral.We find uncertain payment schedules in floatingrate notes, Treasury inflation-protected securities,and mortgage-backed securities. This variety isquite adequate to value accrued pension benefitswith reasonable accuracy: we do not require basis-point precision. Mr. Mindlin also observes thatpension obligations are not marketed, but he doesnot explain why tradeability changes the value ofan obligation from the obligor’s viewpoint.

Mr. Mindlin asserts that “different groups ofstakeholders should be allowed to measure thepension commitment according to their needsand risk tolerance.” What need or risk tolerancewould lead a shareholder to value a dollar paid toa pensioner differently from a dollar paid to acreditor? Why would a pensioner value a pensionpromise differently from an equally secure non-transferable bond of the same company?

Financial obligations vary widely in their legal sta-tus, guarantees, time horizons, and payment con-tingencies. We accommodate such differences byjudgment and arithmetic, not by completely dif-ferent valuation techniques. It is not enough forMr. Mindlin to point to differences between pen-sions and other debt instruments. A serious chal-lenge must show why those differences affectvalue in fundamental ways that cannot be han-dled within Principle 4.

Challenging Principle 1

Mr. Mindlin also challenges our Principle 1,which states that $1 million portfolios of stocksand bonds have the same value. He accepts this principle as “literally correct” but a “manifestation of the ’FAS87-biased’ mindsetthat ... places severe restrictions on the methodsat our disposal.”

The range of methods to justify a preferencebetween stocks and bonds is capacious but notunlimited. Taxes, risk tolerance, the PBGC, peergroup comparisons, and surplus ownership areamong the factors that we may consider. But aminimal constraint on a discounting method isthat it produces equal values for the cash generat-ed by equal amounts of stocks and bonds. Wemay conclude that stocks are preferable for a par-ticular purpose, but we cannot say that $1 millionof stocks has a higher first-order present valuethan $1 million of bonds.

Mr. Mindlin returns to Principle 1 in his Section6, demonstrating that the actuarial model pro-duces equal values for $1 million bond and stockportfolios. He concludes that “the pension actuar-ial model and financial economics are in completeagreement,” contrary to Reinventing. In ourAuthors’ Response, we stated, “Of course, actuar-ies who anticipate risk premiums do not literallyvalue a $1-million equity portfolio more highlythan a $1-million Treasury portfolio. Theyachieve the same result indirectly, however, whenthey value liabilities financed by equity morecheaply than the same liabilities financed bybonds.” Both Mr. Mindlin and ASOP 27 firmlysupport the latter practice.

Three Fundamental Principles?

In Section 6 Mr. Mindlin explains that actuariesshould measure the present value of pension obli-gations by discounting at the expected return R ofthe invested assets. He explains this process interms of “three fundamental principles that areuniversally accepted among actuaries (or so I wantto believe).”

I must disappoint him. Neither Jeremy Gold norI, nor any bond investor or private lender, acceptshis principles. We cannot value a future dollarwithout knowing its payment probability. Norcan we take the discount rate R as the return of anarbitrary security. We must know the targetedpayment probability distribution; then we define

PFN0504.qxd 5/3/05 8:40 AM Page 44

THE PENSION FORUM

45

R as the yield of a security that has that sameprobability distribution. Perhaps no marketedsecurities fit the bill precisely, but we surely couldbracket the discount rate tightly enough toexclude the use of an equity return as R.

The fact that a plan sponsor expects his riskyassets to earn more than the matching asset is notrelevant to liability measurement. High-assetreturns are always welcome, but they don’t reduceliabilities. A successful equity investment, a win-ning lottery ticket, or a big year-end bonus givesyou more resources to pay your mortgage or buyyour groceries, but it doesn’t lower their cost.

Ignoring Principle 5

In Sections 7-8 Mr. Mindlin responds to the threequestions raised in Bader (2001). That articleaddresses these issues, and I will add here only acomment on Mr. Mindlin’s discussion of assetallocation strategy. He suggests that the asset allo-cation should reflect the plan sponsor’s “financialsituation and risk aversion characteristics” (“if thesponsor can live with a 50% chance to fund thecommitment”). Here he ignores Principle 5 inReinventing, “Risks are borne and rewards areearned by individuals, not by institutions.” The“sponsor” does not have to live with the 50%chance: the shareholders do. (Also the partici-pants and PBGC, of course—but Bader [2001]inquired solely about shareholder value.)Shareholder value is determined by how the mar-kets, not the sponsor, price risk. The market isindifferent between equal dollar amounts ofbonds and stocks, and the sponsor cannot affectshareholder value by trading marketed securities(apart from second-order effects like taxation).

In Section 9 Mr. Mindlin warns of inconsistencyin discounting the asset and liability sides. He calls it “a classic ’apples-to-oranges’ case ...one has to use the same discounting procedure onboth the asset and liability sides.” Financial economics achieves consistency here, not by“using the same discounting procedures,” but by

using market prices for both assets and liabilities(using reference securities if the actual securitiesare not traded).

Questions for Financial Economists

In Section 10 Mr. Mindlin poses a few questionsfor financial economists.

First he considers a public employees’ pensionplan that chooses not to buy the liability-match-ing asset. He asks whether the current marketprice of the matching asset is relevant to that plan.The answer is yes: the price of the matching assetis the value of the obligation borne by the taxpay-ers, regardless of how the assets are deployed.

He then turns to a corporate plan that the spon-sor has no intention of terminating. He asks, “Isthe current market price of the matching asset rel-evant?” He labels this price the “terminationvalue” and asks whether financial economistsadvocate valuing all business assets on a termina-tion basis, “[sending their] accountants to a localgarage sale to determine the ’marked-to-market’value of used file cabinets and water coolers.”

Financial economists find the value (not the “ter-mination value,” just the “value”) of marketedfinancial securities by looking at market prices.Those market prices impound the market consen-sus about the future earning power of those secu-rities. It is no more appropriate to call these mar-ket prices “termination values” than it is to regardmutual fund net asset values, or stock pricesthemselves, as “termination values.”

As for valuing a corporation’s nonfinancial assets,that is done by estimating their collective powerto generate future earnings—not by the pricesthat they would fetch at “a local garage sale.”

PFN0504.qxd 5/3/05 8:40 AM Page 45

THE PENSION FORUM

46

Conclusion

In conclusion, I again thank Mr. Mindlin for his thoughtful treatment of Jeremy Gold’s andmy work. His paper will benefit readers who wishto assess the actuarial side of “The GreatControversy.” But I must add a cautionary note.In his concluding sentence, Mr. Mindlin suggeststhat, contrary to the title of Bader (2001), thepension model, and therefore the actuaries whorely on it, are magnificently attired. I advise actuaries who accept this sartorial advice to useplenty of sunscreen.

References

Bader, L. 2001. “Pension Forecasts, Part 2: TheModel Has No Clothes.” Pension Section News no. 46 (June): 14-15. http://library.soa.org/library/sectionnews/pension/PSN0106.pdf.

________. 2003. “Treatment of Pension Plans ina Corporate Valuation.” Financial AnalystsJournal vol. 59 (May/June).

Bader, L., and J. Gold. 2003. “Reinventing Pension Actuarial Science.” The Pension Forum15(1): 1-13. http://library.soa.org/library/sectionnews/pension/PFN0301.pdf.

PFN0504.qxd 5/3/05 8:40 AM Page 46

THE PENSION FORUM

47

I congratulate Dimitry Mindlin on a useful, well-thought-out, and highly entertaining paper. Inparticular, I felt that its FAS87 mindset portrayalis justified and quite true. As someone who has noknowledge of U.S. pension accounting rules, Ioften have trouble communicating to U.S. actu-aries about both traditional and financial eco-nomics ideas. Having said this, I am not qualifiedto comment on the FAS87 elements of the paper.

My personal experience of being educated as anactuary is very different to his. In particular, sever-al traditional actuarial concepts that I learned areat odds with basic financial economics concepts:

l

What is value? My actuarial training causes meto think of value in terms of utility rather thanmarket price. In particular, the present value ofa thing is whatever an actuary calculates it to berather than the price at which it can be tradedfor or manufactured.

l

What is a liability? Again, whatever an actuarydefines it to be rather than the wider use of theterm to mean a debt, a negative asset, or a con-tingent claim.

l

What is a stakeholder? I cannot recall if the word “stakeholder” appeared in the course notes.

l

What does arbitrage-free mean? On qualifica-tion, I thought arbitrage-free was a statementabout market efficiency rather than a concept ofwhat a good financial model should strive to be.

I would like to discuss each of these concepts inthe context of points raised in the paper.

Value

There are several items of a semantic nature that Ifeel are important. First, I quote two definitionsof value (from the Macquarie dictionary):

l

Value (version 1): The worth of a thing is meas-ured by the amount of other things for whichit can be exchanged, or as estimated in terms ofa medium of exchange.

l

Value (version 2): That property of a thingbecause of which it is esteemed, desirable, or useful.

These are very different definitions, and I believemuch confusion results from semantic confusionabout the term “value”: to a first degree, tradition-al actuarial science tends to focus overly on thesecond definition (value as usefulness or utility toan individual), and financial economics focuseson the first (value as market price).

Take, for example, the value of water. Using thefirst definition, the value of water (i.e., the marketprice) is virtually zero; I can leave a tap runningfor hours, and it would cost me less than a cent.Under the second definition, the value of water(i.e., its usefulness or utility) is very high: withoutwater I wouldn’t last long.

Now, the tension between these values is obvious (and a real source of concern for anyoneliving on the driest continent on earth in themidst of a prolonged drought). For example, Iimagine that, as a buyer of water, I am getting agood deal, in that it is not being properly priced,taking into account total costs of manufacture,environmental costs, future scarcity, and so on.Many people reflecting on my situation woulddeclare me a rich man, and I would agree withthem. But would a bank lend me more money—perhaps so I could install more taps and increasemy wealth even more?

I believe that actuaries often look at defined ben-efit (DB) schemes and see utility propositions:

Comments on “Reaffirming Pension Actuarial Science”by Tony Day, B.A., F.I.A.A.

PFN0504.qxd 5/3/05 8:40 AM Page 47

THE PENSION FORUM

48

massive transfers of risk away from members tocompanies, enforced consumption deferral, help-ful equity prices. The problem is that they thenparticipate in the conversion of these items into amarket price that reflects utility rather thanexchange. As a result, banks (and others) lendmoney based on “good deals” that have approxi-mately the same value as water (in both senses ofthe word).

Liability

To continue the semantic theme, I would like tocompare a dictionary definition of liability (againfrom the Macquarie dictionary) with the defini-tion used in the case study:

l

Liability (dictionary): an obligation, especiallyfor payment; debt or pecuniary obligations(opposed to asset).

l

Liability (case study): the asset value thatshould be set aside now to fund the commitment.

I don’t think this redefinition of a common termis fair enough. In addition to the common-senselink between liability and debt, the dictionarydefinition suggests, quite simply, that a liability isthe opposite of an asset. The case study definitionis one step removed: in addition to the “commit-ment” (read: liability), we also take into accountthe “value” of an unrelated “asset.”

To my mind, the definition is equivalent to thefollowing: liability is defined as the dictionarydefinition of liability less the “good deal”obtained from investing in equities (or whatev-er). The case study is thus all about whether thesponsor should take on the “good deal” of equi-ty ownership—it has nothing to do with liabili-ties as commonly defined.

Arbitrage and Stakeholders

The author makes much of the fact that pensionassets and liabilities are not usually tradable, or at

least that closely matching assets to pension liabil-ities may not exist. Yes, of course. He then sug-gests that this somehow means that arbitrage con-cepts therefore do not apply. I disagree.

Financial economics is not dependent on the exis-tence of arbitrage opportunities in the world.Rather, it suggests that we should create arbitrage-free models of the world because not to do soinvites wrong specifications of value (in the mar-ket price sense). It is a failure of traditional actu-arial science to recognize the fact that, most often,arbitrage opportunities can be found within thesystem that is DB pension schemes and do notrequire outside agents.

I think this failure can be pinned down to anabsence of stakeholder analysis in standard actuar-ial practice. If we do not “see” the various stake-holder groups that form a DB plan, we will, ofcourse, fail to see the potential free lunch given toone party at the expense of another. A quick wayto remedy this would be to insist that every pieceof actuarial advice should, first, seek to identifystakeholders and, second, declare what interestthey have in each financial decision, recommen-dation, or assumption.

“The majority of pension actuaries disagree withthe basic tenets of financial economics. Theymust disagree given their advice and actions asagents of DB plans. With the exception of thispaper, their silence in the face of substantive crit-icism has been deafening. I share many ofMindlin’s concerns and agree with much of whathe says: actuaries do ask questions beyond valua-tion of liabilities; they should strive to define andadd utility value for stakeholders; they shouldsearch for and advocate “good deals.” However,whatever the purpose of a calculation or specificsof a question, we should always clearly statewhich definition of value we use and for whomwe are valuing something.

PFN0504.qxd 5/3/05 8:41 AM Page 48

THE PENSION FORUM

49

Mr. Mindlin’s paper is a welcome addition to theongoing discussion on actuarial practice. I hopethat a couple of brief comments might be useful.

Whose Risk Is It Anyway?

In reading Mr. Mindlin’s paper, I was struck by the different approaches to risk in his paperand in the Bader and Gold paper and the surrounding discussions. Briefly, in Bader andGold and related papers we seem to be told thatrisk is to be avoided or hedged at any cost. Anumber of reasons are cited for this: optimal cor-porate structure (the Modigliani and Millerpaper), generational equity (applied to eithershareholders or taxpayers), accounting rules,effects on the PBGC, reduction or death of theequity risk premium, and the idea that stocksheld by corporations in pension plans are subop-timal economically.

In each case I think a counterargument can beformulated, demonstrating that investment riskcan reasonably be taken by those responsible forthe management of a pension plan. However,that is a discussion for another time.

The most important point to be made here isthat the decision about whether or not to takeinvestment risk is not ours to make. The actuari-al profession has not been charged with defend-ing generational equity, protecting the PBGC,guarding corporate or national economic struc-ture, or any of the rest of it. We can offer advice,but we do not make the decisions.

Above all, we must not under any circumstancesmake decisions for our clients implicitly in ouractuarial methods, assumptions, or calculations.

Common Ground?

There must be some common ground in thisdebate.

I think many of us would agree that we canimprove our communication of risk to ourclients. It is in this area that Mr. Mindlin’s paperis especially useful. He suggests how we can move away from point estimates based onexpected values and focus the attention of ourclients on the distribution of possible outcomes,and on the risk-reward tradeoffs inherent inthose distributions.

Using the approaches outlined by Mr. Mindlin,we can provide our clients with better informa-tion concerning the magnitude of the risks theyare taking. We can furnish a balanced picture ofrisks and rewards, and then allow our clients tomake their decision.

It is their decision.

Comments on “Reaffirming Pension Actuarial Science”

by Robert McCrory, F.S.A., M.A.A.A., E.A., F.C.A.

PFN0504.qxd 5/3/05 8:41 AM Page 49

THE PENSION FORUM

50

I would like to thank Mr. Bader for further clarifications of the views he and Mr. Gold pre-sented in “Reinventing Pension Actuarial Science” (Bader and Gold 2003), also known asReinventing. I will follow the same convention and call my paper Reaffirming. The purpose ofthese comments is to clarify and amplify the differences between Reinventing and Reaffirmingin light of Mr. Bader’s remarks. It is also my unfortunate duty to correct several misquotationsand misinterpretations of my views by Mr. Bader.

1. The Great Controversy Once Again

In the conclusion to his comments, Mr. Badernoted that Reaffirming would “benefit readerswho wish to assess the actuarial side of ’The GreatControversy’” (italics added). Apparently the dis-cussion has a nonactuarial side that I have failedto acknowledge. Since it’s desirable to considerthe subject of any debate in its entirety, the ques-tion is: What’s the subject of this debate? Indeed,what are we talking about?

“The Great Controversy” is first and foremostabout the risk management of pension plans andits most important component: the asset alloca-tion decision. Bader and Gold (2003) put existingactuarial practices—the conventional valuationand FAS87 reports—in the context of the realworld of risky assets and demonstrated that, as faras the “marked-to-market” financial reporting isconcerned, the conventional reports contain seri-ous problems. However, those problems were wellknown to economists and practicing actuarieslong before Bader and Gold’s paper, so that partwas relatively uncontroversial. It is the authors’call to reinvent the pension actuarial science inlight of their version of financial economics thatattracted a lot of attention and became the start-ing point of “The Great Controversy.” But asbold and controversial as that call is, it is an off-shoot of even bigger claim. Bader and Golddeclared that the only economically legitimate

way of measuring a series of future cash flows is tovalue it as the price of the matching bond portfo-lio. But we don’t measure pension commitmentsfor the sake of measuring. We measure pensioncommitments primarily to fund them—to deter-mine how much to contribute and how to investthe assets in order to have enough money to makethe promised payments.

So let’s be clear about the subject of this debate.It’s the asset allocation, dear reader.

2. Reinventing’s Principles 4 and 1

Principle 4 is at the front and center of the debate.Unfortunately, in his description of my views, Mr.Bader produced a perplexing mixture of misquo-tations and misinterpretations. Contrary to Mr.Bader’s assertion, I dispute not the language ofPrinciple 4, but the endorsement of that state-ment as a general principle of financial econom-ics. The declaration “A liability is valued at theprice at which a reference security trades in a liq-uid and deep market” is not incorrect by itself.Contrary to Mr. Bader’s assertion, Reaffirming isperfectly happy with the pricing by market values.If one has a good reason to value a stream offuture cash flows as the market price of thematching asset (assuming it exists), then oneshould be at liberty to do so. But once that state-ment becomes a general principle, it takes awayour ability to make forward-looking assumptions.

Author’s Response to Mr. Bader’s Comments


PFN0504.qxd 5/3/05 8:41 AM Page 50

THE PENSION FORUM

51

It also reduces our playing field from financialeconomics to a version of financial reporting.

By definition, the goal of a reporting—financialor any other—is to report what has happened upto this point. Any reporting (not “pricing by mar-ket values,” as Mr. Bader attributes to me) is aninherently retrospective endeavor. It is challeng-ing for any reporting to accommodate future cashflows of uncertain magnitude and timing—think offuture pension contributions or payments tostock option holders.1 Every substantial forward-looking assumption—an expectation—has apotential of creating a serious problem in finan-cial reporting. (Remember the notorious long-term expected return on assets in FAS87?) Unlikefinancial reporting, financial economics has noproblem with expectations. In particular, variousasset allocation and option-pricing2 methodolo-gies require forward-looking assumptions.

Principle 4 requires us to value long-term non-transferable pension benefits as a portfolio oftradable bonds with similar payouts. Messrs.Bader and Gold also claim that “the vast majori-ty of thought leaders in the financial communi-ty” would share that view. It appears that thesame logic should apply to employee stockoptions: those should be valued the same way asstock options with similar characteristics tradable“in a liquid and deep market.”3 Well, here’s whatLawrence Lindsey and Marc Sumerlin wrote inan op-ed piece in The Wall Street Journal on June21, 2004.4 “Employee stock options are longterm and nontransferable. The fact that they can-not be sold means they cannot be measured by

market-based option calculators.” In otherwords, according to Messrs. Lindsey andSumerlin, an employee stock option is not neces-sarily valued “at the price at which a referencesecurity trades in a liquid and deep market.” Itdoesn’t sound like much of a consensus on thatissue even among economists.5

Contrary to Mr. Bader’s assertion, I did not statethat Principle 4 was based on the law of oneprice. The law of one price came up as my bestattempt to justify Principle 4. Reaffirming isabsolutely clear about that principle. As a gener-al principle of financial economics, Principle 4 isunsubstantiated. Period.

Principle 1 is not as controversial as Principle 4.But Principle 1 is not a principle of financial eco-nomics, it’s a principle of financial accounting.When Messrs. Bader and Gold say “$1 million ofstocks is the same as $1 million of bonds,” theysimply change the subject of the discussion. Theproblem of “present valuing” of a stream offuture payments is not the same as the problemof asset auditing. Contrary to Reinventing’s asser-tion, the pension actuarial model neither values a“$1-million equity portfolio more highly than a$1-million Treasury portfolio” nor values “liabil-ities financed by equity more cheaply than thesame liabilities financed by bonds.” At some risklevels, the present value of the pension commit-ment6 funded by stocks is lower than the presentvalue of the pension commitment funded bybonds; at some other risk levels, the opposite istrue. Messrs. Bader and Gold believe thisinequality implies “indirectly” that today’s values

1 The relationship between pensions and options requires more thorough treatment than is appropriate for this discussion.2 See note 1.3 See note 1. But if you insist, here’s one incredible similarity between pensions and options: accountants hate both of them.4 Lawrence Lindsey and Marc Sumerlin are well-known economists. Mr. Lindsey was director of the National Economic Council in

2001-2002 and is now president and CEO of the Lindsey Group, of which Mr. Sumerlin, former deputy director of the NationalEconomic Council, is managing director.

5 It remains to be seen if the arguments of Messrs. Lindsey and Sumerlin on the issue of “employee stock options vs. tradable stockoptions” apply to the issue of “pensions vs. tradable bonds.”

6 As before, the pension commitment is defined as the stream of benefit payments determined by the plan’s population and benefit package.

PFN0504.qxd 5/3/05 8:41 AM Page 51

of $1 million of stocks and $1 million of bondsare not the same, but their logic is flawed. Whiletoday’s values of the portfolios are the same, thepresent values of a series of payments funded bystocks and bonds are different because our expec-tations for stocks and bonds are different. Onceagain, our expectations are the central issue.

3. $103 in One Year versus $100 inTwenty Years

Mr. Bader brings up a case of one payment of$103 in one year. He claims that this author “maypoint out that pension liabilities are far morecomplex than this simple illustration.” No, Iwould not. I absolutely agree with Mr. Bader thatif a “model is wildly incorrect” for a simple exam-ple, we shouldn’t “even bother to inquire how ithandles more complex problems.” I also happento believe that the actuarial pension model is per-fectly capable of handling his example.

Reaffirming analyzes a similar case of one paymentof $100 in 20 years (see Reaffirming, Section 7).Apparently Mr. Bader believes that his example isconceptually different than the example presentedin Reaffirming. Although that conceptual differ-ence eludes me, here are my arguments fromReaffirming in somewhat abbreviated form.

For Mr. Bader’s example, an FAS87 statementmay show the ABO as $100.98 (assuming 2% asthe Treasury one-year rate) or something close tothat value. Let’s assume the entire fund is invest-ed in equities (please don’t blame the actuary forthat decision). If the actuary believes that the geo-metric return for the equities is 8.42% anddecides to use that value as the interest rateassumption (contrary to Mr. Bader’s assertion, Idon’t think this practice is a good idea; seeReaffirming, note 8), the valuation report willshow $95 as the accrued liability. But this valuehas nothing to do with any “price”; it has every-

thing to do with compliance with relevant regula-tions. The model is not broken. Mr. Bader simplymisuses the contents of the valuation report. Idon’t think it’s reasonable to believe that the valu-ation report should be helpful in bond trading.

For his internal purposes, a stock analyst mayvalue the commitment to pay $103 in one year as$95, $100, $100.98, $103, $200, or any othervalue, depending on the analyst’s risk toleranceand the purpose of his calculation. Today’s bondtraders value a similar bond commitment as$100, but I’m not sure what to do with this fig-ure: the idea of funding a company’s pensioncommitment with the same company’s bondsdoesn’t appeal to me that much.

The issue of “a dollar owed to a pensioner vs. adollar owed to a creditor” is not central to thisdebate, but, since Mr. Bader wishes to revisit theissue, let me touch on it briefly. There are severalreasons to treat the pensions and corporate (orstate or municipal) bonds differently; some ofthose reasons are mentioned in Reaffirming. Thereare also good reasons to believe that the pension-ers have a better chance of receiving that dollarthan the creditors, but these issues are moreinvolved than suitable for this discussion.

4. The Law of One Price

Mr. Bader concedes Principle 4 is not based onthe law of one price; my hat is off to Mr. Bader.7

He claims the principle “is based on the simplernotion that like liabilities have like values.” Butthat “simpler notion” is simply incapable of vali-dating Principle 4. When there’s no way to profitfrom a seeming arbitrage opportunity, like objectsdo not necessarily have like prices. Here are a cou-ple of examples of that phenomenon.

THE PENSION FORUM

52

7 There are economists out there who advance similar arguments and insist that their views are based on the law of one price.

PFN0504.qxd 5/3/05 8:41 AM Page 52

THE PENSION FORUM

53

1. An insurance company can review the marketprices of a particular product and either offerthe product at a lower price, or offer the prod-uct at the same price as the competitors, ordecline to offer the product altogether. Inother words, an insurance company doesn’thave to accept existing market prices and mayuse its own expectations to value a product.

2. Disagreements with the market prices are asignificant part of the market itself. Eventhough the market price of a particular stockis $10 now, investor A’s model may value thestock at $5 and recommend to sell it shortand/or buy a put option on the stock.Investor B may value the stock at $9 and donothing. Investor C’s model may value thestock at $15 and recommend to buy the stockand/or a call option on the stock.

Market prices are just a tip of the iceberg. Behindthe market prices, there is an incredible variety ofexpectations and valuations that range from asimple gut feeling to the most sophisticated quan-titative models. Any valuation depends on someuncertain future events, so the market partici-pant’s risk tolerance plays a major role in the val-uation. Objects of our interest (assets, liabilities,etc.) can be valued differently for different organ-izations, different governing bodies, and differentpurposes. For internal purposes, any organizationshould be able to value its or anyone else’s assetsand liabilities any way the organization wants anduse the results of the valuation to achieve anyobjective the organization has. That’s what freemarkets are all about.

Pension liabilities should be no exception, butPrinciple 4 disagrees and orders everyone to valuethe pension liability as the price of the hypothet-ical matching asset. According to Principle 4,there cannot possibly be any need to value pen-sion liability differently, even though the objec-tives and perspectives of different stakeholderscan be entirely different. One of my goals is todemonstrate that there are real problems out there

that require different stakeholders to value pen-sion commitments differently.

5. Why Value Differently?

Mr. Bader brings up the issue of “pensioner vs.bondholder” and asks directly: “Why would apensioner value a pension promise differentlyfrom an equally secure nontransferable bond ofthe same company?” This is a great question aswell as a clear manifestation of the “FAS87 bias.”

Here’s the answer: because the goals of pensionfunding and bond pricing are entirely different.One of the most important aspects of the pensionfunding is the asset allocation decision. We haveto make the decision now; the decision involvessome future benefit payments; therefore, we haveto put the payments on the same footing andbring them to the present. In other words, inorder to make the asset allocation decision, wehave to calculate a present value of the pensioncommitment. At this point we don’t know whichdiscounting procedure would be appropriate forthe present value calculation because the assetallocation objective has not been specified yet.

Let’s look at the objectives of the asset allocation.Let’s take the plan participants’ side. Naturally theplan participants would like to be absolutely cer-tain about the safety of their benefits. However,absolute certainty hardly ever exists; some risksare simply unavoidable. Higher contributionswould help, but the sponsors’ budgets are notunlimited. Therefore, it’s not unreasonable tobelieve that it is in the best interests of the planparticipants that the assets are allocated in a waythat maximizes the likelihood that the benefitswill be paid at a given level of assets (which includethe market value of assets and present value offuture contributions). “[T]he policy is to providethe investor with the highest probability of beingable to pay for the groceries when the timecomes,” writes Peter Bernstein, a prominent econ-omist and historian (see Bernstein 2003).

PFN0504.qxd 5/3/05 8:41 AM Page 53

Now let’s take the plan sponsor’s side. “For a givenpublic or private pension plan benefit structureand for a given funding level, a plan sponsor maywant to choose an investment strategy to mini-mize the present value of future contributions tothe plan.” I could easily imagine a group of economical purists expressing their rightful indig-nation at this statement. (Don’t you know thatthe present value of future contributions is equalto the market price of liabilities minus the marketprice of assets—it’s a number! A number is anumber—it can’t be minimized.) But the groupwould have a tough road ahead. The quotationbelongs to Fischer Black (see Black 1995), a dis-tinguished economist and one of the foundingfathers of option pricing.8

The two asset allocation objectives raise numerousquestions.9 Here’s a small sample. Under the existingasset allocation, what is the probability that a givenasset value is sufficient to pay the promised benefits?What is the highest probability that can be achievedby reallocating assets that a given asset value is suf-ficient to pay the promised benefits? It may be inthe best interests of a plan participant to know thatprobability. If that probability is below, for example,60%, what is the lowest additional contribution tothe fund required to increase the probability to60%? It may be in the best interests of the plan spon-sor to know the value of the additional contribu-tion. Ask the same question for the safety levels70%, 80%, or any other. For every level of risk,there is the lowest asset value required to fund thepension commitment at that level of risk. Thisrequired asset value at a given risk level can be con-sidered a liability, although it doesn’t belong to anyconventional actuarial report. It exists only as a toolfor risk management purposes.

Would Messrs. Bader, Gold and like-minded econo-mists allow these and similar questions to be asked

and answered? If yes, is it their position that themarket price of the hypothetical matching asset isthe only present value we need to answer thosequestions? Is it their position that those questionscan “be handled within Principle 4”? Count me asa skeptic. Fortunately our good old friend thepension actuarial model is readily available.

6. Two Worlds

We live in a financial world that contains a multitude of risks. That world is governed by theprinciples of financial economics. Risk manage-ment is an indispensable part of life in that world.Risk elimination is a valuable component of therisk management, but it’s not always available.Financial markets play arguably the most impor-tant role in the risk management. Market pricesare of great consequence, but everyone has theright to disagree and enjoy or regret the conse-quences of that disagreement. There are manyways to express one’s disagreement with the market prices. Countless participants of the mar-kets apply their unique expectations to takeadvantage of opportunities the markets create.There’s no artificial bias toward any asset class.Certain market participants may designate thebonds as their preferred asset class. But that’s achoice, not an obligation: they are free to employevery available funding vehicle and allocate theassets in their best interests.

Market participants consider each asset on itsmerits in terms of return, risk, relationships withother assets, and, most important, the capabilityof the asset to serve the financial objectives of thisparticular market participant. The market partic-ipants have a great appreciation of what may hap-pen in the next instant, day, year, or century. Theyuse their expectations and risk tolerance to priceuncertain future events. Since the market partici-

THE PENSION FORUM

54

8 You’ve got to read note 1; it’s not even funny anymore.9 The asset allocation objectives specified by Peter Bernstein (2003) and Fischer Black (1995) lead to the same set of optimal

policies, but this remarkable fact is outside of the scope of this discussion.

PFN0504.qxd 5/3/05 8:41 AM Page 54

THE PENSION FORUM

55

pants have different expectations and risk toler-ances, trades happen.

Pension plan sponsors are no exception. In thisworld the asset allocation decision for a pensionplan is of paramount importance. The papersBlack (1995) and Bernstein (2003) belong concep-tually to this world; they acknowledge the presenceof risks and articulate the objectives accordingly.

Reinventing belongs to a different world, or, moreaccurately, to a small segment of the world offinancial economics. In that smaller world, as wellas in the statement FAS87, risk elimination isassumed to be always available. There’s the pre-ferred asset class: the bonds. Mr. Bader declaresthat “marketed bonds carry maturities fromimmediate to 100 years, default risk from 0% to100%, and a wide range of collateral. We finduncertain payment schedules in floating ratenotes, Treasury inflation-protected securities, andmortgage-backed securities. This variety is quiteadequate to value accrued pension benefits withreasonable accuracy.”

Those “default risk” and “uncertain payment sched-ules” in various fixed-income securities are contin-gent events in the future. Today’s bond traders usetheir expectations to price those bonds. Theseexpectations and resulting bond prices are elevatedto the status of unquestionable tenet in the smallerworld. The only asset class for which expectationsare allowed is the fixed-income securities. The onlygroup of people allowed to have expectations aretoday’s bond traders. Everyone else must follow theexpectations built into today’s bond prices. Just likein FAS87, only the bonds can be used to value pen-sion commitments in the smaller world.Expectations about equities must be excluded: asMr. Bader puts it, “we surely could bracket the dis-count rate tightly enough to exclude the use of anequity return.”

Just like in FAS87, what may happen in the nextinstance is, for the most part, of no interest. Just

like in FAS87, the actual asset allocation is irrele-vant in the smaller world; after all, $100 of bondsis the same as $100 of stocks. Just like in FAS87,we must utilize the imaginary investment in thehypothetical matching bond portfolio for the dis-counting purposes in the smaller world. That’s thementality Messrs. Bader, Gold, and like-mindedeconomists are presenting as the true financialeconomics. Taking this logic one step further, Mr.Bader is proposing now to make this way ofthinking compulsory. According to Bader (2004),“mandatory full funding … is the only practicalprevention for the diseases that can afflict a guar-antee system.”

I’ve tried to come up with an appropriate term forthe smaller world. As Mr. Bader noticed, I intro-duced “FAS87 bias” for lack of a better term. Thetwo other finalists were “accounting economics”and “extreme form of marked-to-market para-digm.” Having settled on “FAS87 bias,” I stillthink the term carries useful associations with it.

7. Reaffirming’s Principles 1 and 2

Once again, I’d like to give Messrs. Bader andGold a lot of credit for their well-thought-outstructure “from-basic-principles-to-conclusions”in Reinventing. My intention was to follow suit inReaffirming. It’s great that we can express our dif-ferences at the level of basic principles. The read-ers will have a clear choice: to accept or rejectthose principles.

Unfortunately Mr. Bader’s presentation of my views is inaccurate. The statement he attrib-utes to me—“actuaries should measure the pres-ent value of pension obligations by discounting atthe expected return R of the invested assets”—isincorrect in two ways. First, my position on meas-urements of pension commitments is, in short,we use different measurements for different pur-poses. Reaffirming recognizes “the multitude ofchallenges the pension plan stakeholders face andthe multitude of liabilities related to those chal-lenges.” Some measurements are related to today’s

PFN0504.qxd 5/3/05 8:41 AM Page 55

THE PENSION FORUM

56

yield curve; some measurements are related to theexisting policy portfolio; some measurements arebased on someone’s wishful thinking-the pensionactuarial model is unbiased and can accommo-date everyone. Second, the term “expected return”traditionally means a fixed discount rate equal toeither geometric or arithmetic return on assets.Reaffirming’s Principles 1 and 2 do not requireutilization of the “expected return” or any otherfixed rate.

I agree that “we must know the targeted paymentprobability distribution.” We may also “define Ras the yield of a security that has that same prob-ability distribution,” but only if either the assetsare invested in that security or we are willing toassume that the assets are invested in that securi-ty (e.g., for the purposes of financial reporting).Mr. Bader claims that “financial economicsachieves consistency … by using market prices forboth assets and liabilities (using reference securi-ties if the actual securities are not traded).” So themodel that Messrs. Bader and Gold propose isinternally consistent for the few plans that havepurchased the matching bond portfolio.

What about the ones—and that’s almost everyoneelse—that have not? Don’t they need an internal-ly consistent asset-liability model? Apparentlythey must assume that they have purchased thematching bond portfolio and take it from there.However, there are important problems thatrequire utilization of the actual policy portfolio. Idon’t think it’s reasonable to believe that theassumption of investment in the hypotheticalmatching bond portfolio provides sufficient ana-lytical tools to solve for all the problems thatrequire a present value calculation.

It is disappointing that, according to Mr. Bader,“neither Jeremy Gold nor I, nor any bondinvestor or private lender, accepts” Reaffirming’sPrinciples 1 and 2. I’d like to believe that Mr.Bader is mistaken regarding the “any bondinvestor or private lender” part of that statement.

8. It Is All about Future Earnings

In Section 10 of Reaffirming I posed a few ques-tions for financial economists. The purpose ofthose questions was to illustrate and amplify oneof the biggest problems in Reinventing. I’d like tothank Mr. Bader for giving excellent answers tosome of my questions and highlighting the pointI was trying to make.

Mr. Bader declares that “only the accumulatedbenefit obligation, which is not subject to wageinflation, falls within the definition of a liability.”By definition, the ABO is the termination liabili-ty: it is calculated assuming the plan has been ter-minated. Clearly Reinventing’s position is pensionplans should be valued on the termination basis.Dear financial economists, I asked, are youinstructing us to value every segment of a partic-ular business on the termination basis? Mr.Bader’s answer is perfectly clear and correct. No,the basis for the valuation is future earnings. Sohere’s what Mr. Bader appears to suggest: everysegment of a particular business should be valuedon the basis of its “power to generate future earn-ings,” except the pension plan! We should apply our expectations to future revenues, but not tothe future cost of running the pension plan. Shallwe skip the future cost of other employee bene-fits? What about the future cash compensation?Many stocks will surely look a lot more attractivewithout the future labor cost.

My views are perfectly clear as well. The purposeof the water cooler is not to appreciate in valueand contribute to the bottom line, but to keep thelabor force hydrated and productive. A particularbusiness should be valued as a whole (includingthe pension plan, needless to say) on the basis ofits future earnings. Future earnings are equal tothe difference between future revenues and futurecosts. A pension plan, as part of the labor cost,belongs to the latter. Present value of future con-tributions should be at the front and center of thevaluation, but we won’t find that figure in FAS87.Accounting in general and FAS87 in particular are

PFN0504.qxd 5/3/05 8:41 AM Page 56

THE PENSION FORUM

57

not terribly helpful with the analysis of future cashflows of uncertain timing and magnitude (likepension contributions or stock option payouts).Reinventing is a great paper in many respects, butthe direction of future developments of the pen-sion actuarial science is not one of them. An“FAS87-biased” version of financial economics, aspresented in Reinventing, is inadequate for a com-prehensive analysis of a pension plan.

9. Conclusion

This discussion has uncovered a number of seri-ous problems in the actuarial analysis of pensionplans, particularly in the area of risk manage-ment. Messrs. Bader and Gold believe the solu-tion is to employ the version of financial eco-nomics they presented in Reinventing and sur-rounding discussions. In their view, the pensionactuarial model is “obsolete.” I disagree. The pen-sion actuarial model is perfectly capable ofaddressing all issues raised in Reinventing and, Ibelieve, will be eventually recognized as a valu-able part of financial economics.

We use the pension actuarial model to calculatethe ABO—Mr. Bader’s preferred liability con-cept—as well as other components of statementFAS87. We use the pension actuarial model tocalculate liabilities for the traditional valuationreport. We use the pension actuarial model toanswer the questions raised in Section 5 of thisresponse. The need to “reinvent” the pensionactuarial model appears to be rather exaggerated.

In conclusion, let’s get back to our emperor. Hewas introduced indirectly in Bader (2001) andmade brief appearances in Reaffirming and Mr.Bader’s comments. Let me also bring up investorA from Section 4 of this response, who believesthat the intrinsic value of a particular security is$5, while the security’s market price is $10. To

demonstrate that “The Emperor Has NoClothes,” Messrs. Bader, Gold, and like-mindedeconomists should prove that the investor can’tvalue the security by himself; any deviation fromthe market price should be prohibited; the securi-ty’s market price is all he needs to know. Here’s achallenge for them. Dear proponents of theextreme form of marked-to-market paradigm,please demonstrate how to profit from the osten-sible arbitrage opportunity investor A has created.When you show the money, it will be obvious toeveryone that the emperor is naked.

While they’re in the process of designing thatstrategy, it would be prudent for the emperor’saccountant to follow Mr. Bader’s advice regardingthe sunscreen. The emperor shouldn’t be both-ered: he is perfectly safe in his magnificentclothes, thank you very much.

References

Bader, L. 2001. “Pension Forecasts, Part 2: TheModel Has No Clothes.” Pension Section Newsno. 46 (June): 14-15. http://library.soa.org/library/sectionnews/pension/PSN0106.pdf.

________. 2004. “Pension Deficits: AnUnnecessary Evil.” Financial Analysts Journal vol.60 (May/June).

Bader, L., and J. Gold. 2003. “ReinventingPension Actuarial Science.” The Pension Forum15(1): 1-13. http://library.soa.org/library/section-news/ pension/PFN0301.pdf.

Bernstein, P. 2003. “Which Policy Do YouMean?” Economics and Portfolio Strategy(August 15).

Black, F. 1995. “The Plan Sponsor’s Goal.”Financial Analysts Journal vol.51 (July-August).


PFN0504.qxd 5/3/05 8:41 AM Page 57

I’m delighted that Mr. Day finds Reaffirming “highly entertaining.” However, I wouldn’tadvertise the “entertainment” component of this discussion. A young aspiring actuary mightthink, “If these guys consider this stuff entertaining, their day jobs must be unbearably dull.”We don’t want to make it that obvious.

Mr. Day takes this discussion in a very interesting direction. He is correct to emphasize theimportance of the language we utilize. The “items of a semantic nature” present a valuablecontribution to this discussion. I respond to Mr. Day’s comments in the same order as theyappear in his piece.

THE PENSION FORUM

58

Value

Mr. Day gives us an intriguing example of anobject that can be of insignificant value in onesense and highly valuable in another. But as fasci-nating as the “running water” example is, it is animperfect illustration for the subject of this dis-cussion; certain components of the running waterexample are not readily quantifiable. What wouldbe the value of a glass of water if you were dyingof thirst in a desert? It’s a great question, but wemay want to leave it there for now: the answer haslittle to do with financial economics. Fresh airwould be the ultimate example of an object thathas no value: it’s free, but, at the same time, it isimmeasurably valuable to any breathing being. Itremains to be seen if this example is helpful in thisdebate; count me as unconvinced.

We are in the business of valuing of future cashflows of uncertain timing and magnitude. Thevalues we deal with are inherently quantitative.The desire to stay within the category of quanti-tative values leads me to a slightly different exam-ple, which, to me, serves as a good illustration ofthe two versions of the value concept.

Think of an automotive vehicle. Version 1 of thevalue concept from Mr. Day’s dictionary is the

price of the vehicle: it is “the worth of a thing asestimated in terms of a medium of exchange.”This value is expressed in monetary units.

Let’s consider some other properties of the vehicle:

1. 0 to 60 mph acceleration time2. Towing capacity3. Number of passengers4. Mileage5. Safety grade (1 to 5 stars).

Each property on this list is an example of version2 of the value concept: it is a “property of a thingbecause of which it is esteemed, desirable, or use-ful.” All these properties are quantifiable.

While version 1 of the value concept is expressedin monetary units, version 2 of the value conceptmay be expressed in seconds, pounds, number ofpeople, miles per gallon of fuel, or vaguelydefined “stars.” Mr. Day believes that the two ver-sions “are very different.” While some differencesdo exist, the two versions are fundamentally sim-ilar. Both versions are measurements of the objectunder consideration. The object can be measuredin monetary or other units, but the result is thesame—we assign a number to the object. “The

Author’s Response to Mr. Day’s Comments


PFN0504.qxd 5/3/05 8:41 AM Page 58

THE PENSION FORUM

59

tension between these values” is not caused by thedifference in the value concepts. Rather, it’s areflection of the fact that an object can be valueddifferently for different purposes and by differentstakeholders.

For example, the same vehicle can have highmileage and little towing capacity—it can behighly valuable to a student and of little value toan owner of a boat. The difference in the values ofwater in Mr. Day’s example exists because Mr.Day plays two different roles: a customer of thewater service in version 1 and an individual whoneeds water for his survival in version 2. It comesas no surprise that two different perspectives gen-erate two different values of the same object. Mostobjects allow multiple measurements. Even cashmay occasionally exhibit this quality.

Now let’s get back to the actuarial models. Here’sthe difference between Reinventing and Reaffirmingin a nutshell. Reinventing declares that the price ofthe hypothetical matching asset is the only legiti-mate measurement of a series of future cash flows.Reaffirming argues that a series of future cash flowscan be measured in many ways. The problem offunding a series of future cash flows is more compre-hensive than the problem of pricing the series. Oneof the reasons for the existence of multiple measure-ments of a series of future cash flows is the presenceof risk: different risk tolerance levels may requiredifferent measurements.

Liability

We look at the issues in this discussion from differ-ent perspectives. Actuaries, economists, investmentprofessionals, accountants, et al.—we all havestrong views shaped by our education and experi-ence. As we operate in different areas of the pensionand other industries, it’s no wonder that we perceivecertain matters differently. The liability concept is agood example of this phenomenon.

I am convinced that the word “liability” has beenlargely overused in the actuarial and financial

literature. There are simply too many objectscalled a “liability.” Conventional actuarial reportshave no shortage of those liabilities: accrued lia-bility, current liabilities, ABO, PBO, and severalothers. But some authors recognize other liabili-ties as well. “The investor’s problem is to fund astream of liabilities,” writes Bernstein (2003). Itsounds like that liability is an individual paymentin a series of future cash flows. In this context, “astream of liabilities” most likely means “a streamof payments,” but it is disconcerting that we haveto guess about the meaning of the word. Kneafsey(2003) employs a similar definition: “Construct aseries of cash flows needed to successfully reachone’s goal. This is the liability.” In Ryan andFabozzi (2002), the word has two distinct mean-ings in the same sentence: “the liabilities are val-ued as the present value of future liabilities.” Afterall, what is the liability: a series of future cashflows or a single value?

Let’s assume for a moment that we have agreed thatthe liability is a single value. Principle 4 inReinventing declares that the price of the hypothet-ical matching asset is the only economically correctmeasurement that can be called a liability. If thatprice is greater than the market value of assets in thepension fund when the plan is terminated, thesponsor is liable for the difference in the most strin-gent interpretation of the word “liable.”

But what are we discussing: the issues of financialeconomics or the desired stringency levels of theword “liable”? Imagine that the sponsor has createda schedule of contributions to fund its pensioncommitment (e.g., as a fixed percentage of payroll).If the scheduled value of assets at the present isgreater than the market value of assets in the pen-sion fund, the sponsor is “liable” for the differencein a sense that is stringent enough to compel a rela-tively large group of relatively reasonable people tocall the scheduled value an accrued liability. So whatis the crux of the issue: the scheduled value of assetsmust not be named a “liability” because the term isreserved for something else, or the scheduled value

PFN0504.qxd 5/3/05 8:41 AM Page 59

THE PENSION FORUM

60

of assets must not be named a “liability” because thevalue is of no interest to anyone?

Reinventing proposes to use the term “liability”exclusively for the purposes of financial account-ing. Of course, we can define the liability as theprice of the hypothetical matching asset, butthat’s a definition, not a principle. Even if every-one accepted the terminology Reinventing advo-cates, we’d still have to come up with a new ter-minology for other measurements of pensioncommitments that are of vital interest for thestakeholders in the pension industry. We canargue about wording conventions, but this argu-ment has little to do with financial economics.The argument is an exercise in semantics.

Even if all the governing bodies that regulate thepension industry reinvented themselves and collec-tively eliminated all definitions of liability exceptthe one that’s equal to the price of the hypotheticalmatching asset, we would still have a multitude ofmeasurements of pension commitments becausethey are needed for the prudent pension plan manage-ment. That “reinvention” would move the pensionfunding from the tightly regulated area of tradi-tional actuarial valuations to a loosely regulatedarea of broadly defined asset-liability management.Even if no governing body is interested in lookingat the benefit levels the plan participants will get atretirement (the future service included), the rest ofus should look at those levels anyway. Even if allthe governing bodies are interested in the termina-tion valuation only, the rest of us should stillattempt to create a disciplined funding methodol-ogy for pension plans that can be reasonably con-sidered as “ongoing concern.” Doing otherwisewould be less than prudent.

As far as the semantics is concerned, we have twooptions on the table. Reinventing proposes to reservethe term “liability” for the financial accounting pur-poses only; the terminology for other objects of ourinterest must be reinvented. Reaffirming proposesthe most general definition of liability that can

accommodate everyone. A stream of benefit pay-ments (which may or may not include the futureservice) is called “a commitment.” To measure (orvalue) a commitment, we employ a discountingprocedure. A broadly defined liability is a commit-ment discounted via any discounting procedure. Noreinventing is required.

In light of this definition, the liability, defined inMr. Day’s dictionary as an obligation to make aseries of payments, is a commitment. In that senseMr. Day is correct: Reaffirming has redefined theterm. But we do need two different expressions todifferentiate series of future cash flows and theirmeasurements in our terminology. The utilizationof the same term (liability) to two objects of dif-ferent nature (a series of payments and a singlevalue) is confusing and highly undesirable.Traditionally pension liability is a single value(e.g., accrued liability, current liabilities, ABO,PBO, etc.), which gives me no choice but to comeup with a different term for the promised series ofpayments. In the case of a pension plan, the spon-sor makes a commitment to make a series of pre-defined payments to the plan participants. That’sone of the reasons behind the choice of the term.To me, the term “commitment” has certain usefulqualities associated with it.

Mr. Day is correct to note that “the case study isthus all about whether the sponsor should take onthe “good deal” of equity ownership.” Reaffirming isperfectly open about this. The case study is a quin-tessential funding problem: the objective is to deter-mine how much to contribute and how to allocatethe assets. The asset that Mr. Day perceives as“unrelated” is an indispensable component of thesolution to the fding problem. However, I under-stand why the asset may be perceived as “unrelated”:it’s truly unrelated to any accounting statement.

I disagree with Mr. Day that “the case study … hasnothing to do with liabilities as commonlydefined.” It is the other way around. Those “com-monly defined” accounting liabilities have nothing

PFN0504.qxd 5/3/05 8:41 AM Page 60

THE PENSION FORUM

61

to do with the case study. In the case study, the roleof the policy portfolio is not to manage the spon-sor’s accounting statements, but to minimize costfor the stakeholders (shareholders or taxpayers) andmaximize safety of the benefits for the plan partici-pants. The matching bond portfolio happens to beone of the investment options. The price of thisportfolio is roughly equal to one of the “commonlydefined” liabilities—the ABO.

Arbitrage and Stakeholders

When told to think (believe, value, etc.) exactlylike someone else, we should demand extraordi-narily good reasons to do so. Had it not been forour natural tendency to think independently,most of us would have still been outspoken mem-bers of the Flat Earth Society.

Some economists insist that we must value aseries of pension benefit payments exactly liketoday’s bond traders value a similar series offuture cash flows. What if someone refuses tocomply? The punishment for doing so is sup-posed to be the presence of an arbitrage opportu-nity, but is it there?

Mr. Day is correct that “we should create arbitrage-free models of the world.” But if two ana-lysts assign different values to an asset (or a liability) because their objectives and/or risk budg-ets are different, this world is still arbitrage-free. Ifan investor believes that the market price of an assetis higher than the asset’s intrinsic value, it doesn’tnecessarily mean that the investor’s methodologymust be reinvented. A difference of opinions does-n’t necessarily create an arbitrage opportunity.

I agree with Mr. Day that “every piece of actuari-al advice should, first, seek to identify stakehold-ers and, second, declare what interest they have ineach financial decision, recommendation, or

assumption.” Most importantly, we must identifythe purpose of the valuation. As we’ve seen in thisdiscussion, failure to do so may lead to assigninginappropriate values to highly valuable objects.

Conclusion

My biggest disagreement with Mr. Day is about hisstatement that “the majority of pension actuariesdisagree with the basic tenets of financial econom-ics.” I am not qualified to comment on opinions ofnon-U.S. actuaries. As far as U.S. actuaries are con-cerned, their professional activities are driven, forthe most part, not by their knowledge of financialeconomics, but by the existing regulations imposedby a number of governing bodies. Even if “arbitrageopportunities can be found within the system thatis DB pension schemes,” their existence has very lit-tle to do with pension actuarial science. There is atendency out there to blame the actuaries and theactuarial education process for the unfortunate stateof affairs in the pension industry. I believe that, forthe most part, the blame is misplaced.

The areas of our agreement appear to be muchbigger than the areas of our disagreement. I’d liketo thank Mr. Day for his witty and thought-pro-voking comments.

References

Bernstein, P. 2003. “Which Policy Do You Mean?”Economics and Portfolio Strategy (August 15).

Kneafsey, K. 2003. “Solving the Investor’sProblem.” Investment Insights (The InvestmentResearch Journal from Barclays Global Investors)vol. 6, issue 5 (August).

Ryan, R., and F. Fabozzi. 2002. “RethinkingPension Liabilities and Asset Allocation.” Journalof Portfolio Management vol.28, no. 4 (summer):7-15.


PFN0504.qxd 5/3/05 8:41 AM Page 61

THE PENSION FORUM

62

Mr. McCrory is correct to recognize the dif-ferent approaches to risks as one of the mainthemes of this discussion. There appears tobe no consensus in this area. But I also agreewith Mr. McCrory that the appropriatenessof investment risk in a pension plan is “a dis-cussion for another time.” That discussionhas been around for quite some time, and itcontinues in this issue of The Pension Forum;see Mr. Bader’s “Pension Deficits: AnUnnecessary Evil” and related comments.

It is true that the actuaries do not makeinvestment decisions for their clients. Thejob of an actuary is to “furnish a balancedpicture of risks and rewards” based on actu-arial models. What makes a good actuarialmodel? This is one of the biggest challengesthat actuaries face. Mr. McCrory makes agood point that “we must not under anycircumstances make decisions for ourclients implicitly in our actuarial methods,assumptions, or calculations.” A good actu-arial model should neither contain a hiddeninvestment advice nor be biased toward anygroup of professionals that claim to havesuperior knowledge at this moment. Today’saccepted wisdom may look imprudenttomorrow. But above all, a good actuarialmodel should reflect reality. For example, ifa sizable group of plan sponsors invest innon-matching risky assets, a good modelshould be able to demonstrate the risks and

rewards that this investment decision maybring about.

There’s no doubt that certain implementa-tions of the pension actuarial model haveserious problems. But at its core, the pensionactuarial model is a good one. So far, I haveseen neither convincing evidence to the con-trary nor a credible alternative.

Author’s Response to Mr. McCrory’s Comments



PFN0504.qxd 5/3/05 8:41 AM Page 62

THE PENSION FORUM

63

Procedure for Preparing Articles for The Pension Forum

Pension Section News is intended as a medium for the timely exchange of ideas and information of inter-

est to pension actuaries. The Pension Forum is for the publication of full papers and is issued on an ad-

hoc basis by the Pension Section.

All articles will include a byline (name, with title and employer, if you wish) to give you full credit for

your effort. The Pension Forum is pleased to publish articles in a second language if the author provides

a translation.

So that we can efficiently handle articles and papers, please use the following format when submitting

articles and papers to either Pension Section News or The Pension Forum.

Send either a Microsoft Word file or mail the article on diskette. Headlines are typed upper and lower

case. Carriage returns are put in only at the end of paragraphs.

Please send a copy of the article to:

Arthur J. Assantes

Editor, Pension Section News / The Pension Forum

The Pension Section

Society of Actuaries

475 Martingale Road, Suite 600

Schaumburg, IL 60173

PFN0504.qxd 5/3/05 8:41 AM Page 63

NOTES

64

PFN0504.qxd 5/3/05 8:41 AM Page 64

Date post:	16-Feb-2019
Category:	Documents
Upload:	doanbao
View:	222 times
Download:	0 times

THE PENSION FORUM - SOA · booklet of The Pension Forum for personal, internal, classroom or other...

Documents