"[F]irst, the accounting for defined-benefit pension plans deviates from the accounting required for other business and compensation arrangements, even when the economics are similar. While issues such as how to most appropriately measure the pension obligation and report pension items in the income statement should be considered, the Staff believes that work on accounting for defined-benefit plans should also focus on those areas that are inconsistent with the accounting for similar items in other areas, including:

• Consolidation…there is not an obvious reason why the plan should not be consolidated…
• Deferral of Actuarial Gains and Losses – It is not clear why changes in estimates related to retirement obligations should not be treated in the balance sheet the same way as changes in estimates related to other obligations…
• Valuation of Asset…As the sponsor of a defined-benefit plan is affected by the gains and losses on pension plan assets in almost the same way as it is affected by gains and losses on other investments, this distinction appears questionable.

[W]hile the disclosures are quite detailed, the Staff notes that it has long been accepted that ’good disclosure doesn’t cure bad accounting.’"
                                                                          - SEC report, pp. 107-108

For those keeping score at home, accounting for retirement benefits was only the second most serious accounting concern relating to the understatement of corporate liabilities. The SEC estimated that there might be "approximately $1.25 trillion in non-cancelable future cash obligations committed under operating leases that are not recognized on issuer balance sheets."

Given that subtle prodding by the SEC, it is little wonder that FASB voted unanimously to overhaul accounting standards FAS 87 (Pensions) and FAS 106 (Other-Post Employment Benefit Obligations (OPEB) aka: Retiree Healthcare) in November. The structure of the reform will take two phases:

The two-step approach addresses SEC’s highest concerns, but leaves the dirtiest and most contentious work for later. For some sponsors, Phase One will dramatically alter the appearance of their balance sheet and reduce shareholder’s equity. According to a study by CSFB, if Phase One had been in effect at the end of 2004, eighteen companies would have seen shareholder’s equity reduced by 25% or more and seven would have seen it wiped out completely.⁴

While the one-time effects of the change are significant, reducing shareholder’s equity of the S&P 500 by some 7%, the impact of the change will be felt years into the future. Because asset values and PBO will be valued at market on the balance sheet annually (quarterly mark-to-market will probably come in Phase Two), sponsors of large pension plans who do not address asset-liability mismatches will see significant volatility in book value much as FAS 133 added volatility to the balance sheet of large derivative users. This may draw heightened attention from ratings agencies, bondholders, and those few equity investors who look beyond the income statement.

Changes to how pensions are presented on the income statement will be far more contentious. Three of the major issues to be addressed in Phase Two:

1. Presentation of Net Periodic Pension Cost (NPPC)
   NPPC currently aggregates the cost of additional benefit provision (service cost), with financing costs (interest cost), income (return on assets), and amortized gains and losses. Revision of the standard may keep service cost as an operating expense while moving the other pieces to below the EBIT line.
   
   
2. Amortization of Gains and Losses
   It is unclear how this piece will be reflected on the income statement if the entire amount of gains and losses will be reflected on the balance sheet in Phase I. To preserve coherence between the two, some flow through the income statement is required.
   
   
3. Use of an Expected Return on Assets or Actual Realized Returns
   Accounting currently uses an expected rate of return rather than realized returns. This yields a considerable reduction in volatility, but there is some debate whether doing so gives an accurate picture of the company. It does enable a truer picture of core operating businesses, but not the total picture of the sponsor. A move away from expected returns would likely be the single greatest factor leading to major changes in sponsor asset allocation.

The arguments for expected return center on the idea that plans are sponsored for the long run and year-over-year changes represent non-material noise. Financial economists would contend that using expected returns without adjusting for the risk of the assets (or possibly the relative risk of assets versus liabilities) is flawed. The reason that expected returns for equities or risky bonds are higher is precisely because risk is borne. Recognizing income attributable to a risk premium before the actual returns are realized is tantamount to counting your chickens before they are hatched. It is like writing insurance or making loans without providing reserves for possible losses.

The natural tendency using expected returns is to bias asset allocation toward higher returning assets to enhance earnings. Because equity returns are volatile and uncertain, the question arises whether expectations may be too aggressive on a forward-looking basis. Figure 2a, below, shows the yield on the Moody’s long-term AA corporate index (a proxy for liability discount rates), 10-year Treasury yields, and the average spread between the discount rate and expected return on assets for S&P 500 companies with defined benefit pensions. Note that the spread between assumed returns on assets rises while the bond yields fall. Assuming a simple split of 65% equities and 35% fixed income, Figure 2b shows the implied equity risk premium (ERP). While the first year of the sample appears conservative with an ERP of 2.4%, recent implied ERP’s may appear high considering current high levels of liquidity from global sources of capital and street estimates for the ERP clustering in the 3-4% range.⁵

If using expected returns is a "little white lie" in corporate GAAP, it is a much bigger issue among public sector pensions. While FASB discounts liabilities commensurate with high quality corporate bonds, GASB (Government Accounting Standards Board) discounts liabilities for public plans using the assumed return on assets. Making adjustment to reported funding ratios on a market-value-of-assets basis paints a bleak picture. One state plan that we spoke with recently shows a funding ratio of 72% on its latest actuarial report. Adjusting liabilities to reflect the FASB basis, which is roughly where the state could issue taxable bonds to fund the plan, would reduce the funded ratio to approximately 55%. With a larger envelope and sharper pencil, making allowance for more conservative mortality tables, accurate valuation of cost-of-living adjustments, and other major factors, that number could easily fall below 50%. This is a typical plan, probably near the middle of the pack if ranked by funded status. Others are in far worse shape. It is hard to know the aggregate status of the country’s state and municipal plans on the same basis as corporate America’s, but the adjusted deficit is likely in excess of one trillion dollars.

Some would argue that the use of expected returns for public plans is more suitable as they have infinite lives and thus longer planning horizons. While this may have some merit if applied to the federal level, it is weaker when applied at the state or municipal level. The states and municipalities may endure, but assuming their taxpayers—in the face of rising income, sales, or property taxes—will continue to pay for benefits that do not accrue to them is a bigger stretch.

Another issue with state and municipal plans is the increasing tendency to put off addressing the funding problem. While localities have historically amortized gains and losses over periods on the order of ten years, they are increasingly extending these to thirty or even forty years. The sponsors may endure, but the elected officials who passed the buck onto future generations will be long gone. With benefits earned over thirty years and mistakes corrected over another thirty, it is not our children who will be forced to pick up this check, but our grandchildren.

As of this writing, the Senate (S1783) and House (HR2830) have both passed bills to reform pension funding and are yet to have a conference committee to iron out differences. The arrow below indicates the relative stringency.  There is some debate of the relative stringency among current law and the House and Senate bills, with all three still short of the Administration’s proposed rewrite.⁷ At the far right are the FTK’s risk-based capital rules that come into effect at year’s end in the Netherlands. The Dutch effectively regulate pensions with a slightly toned-downed version of the capital adequacy standards for life insurers. Plans must maintain solvency (assets > liabilities) and be able to demonstrate solvency with a 95% probability looking forward one year into the future. If they fail the test they must have a remediation plan in place and benefit accruals are frozen. Promises made are promises kept.

All versions of reform (House, Senate, Administration) have a number of common elements; the differences are mostly a matter of degree. All would amortize unfunded liabilities over seven years. All would use some form of modified yield curve to value benefits, with the Administration wanting a full AA-quality yield curve and both the House and Senate substituting a simplified three-point approximation and possibly weaker credit standards. All would have some phase-in of the new law with full implementation not coming until 2008 or 2009. All would raise PBGC per participant premia to $30 per head from the current $19. All allow for additional tax deductibility of contributions for plans that are currently over-funded. The principal differences lie in smoothing asset values or discount rates through time, treatment and determination of "at risk" plans, PBGC variable rate premia, special dispensation for the airline industry, and the use of credit balances – still many horses to trade.

• Time Smoothing
  Current law allows for a four-year smoothing of discount rates and up to five-year averaging of asset values to reduce volatility. Critics would argue that averaging gives an unnecessarily conservative picture of plan health in good times and a dangerously rosy picture in bad times – it conflates the valuation and capital budgeting processes needlessly. HR2830 would reduce both smoothing periods to a maximum of three years. The Senate and Administration would use fair market values of assets at valuation date. S1783 would smooth interest rates over one year and the Administration version would smooth over the 90 business days preceding the valuation date.
  
  
• At-Risk Plans
  Given the troubles faced by the PBGC and the characteristics displayed by plan sponsors who have put pension plans to the PBGC, all proposals have more stringent funding requirements and limitations on grant of additional benefits for financially weak sponsors with poorly funded plans. All would force actuarial valuation assumptions that increase the present value of liabilities beyond that seen in healthy running plans (termination versus ongoing liabilities). The principal difference lies in determination of "at risk" status with the Senate and Administration versions relying on credit ratings assigned by the recognized ratings services. Weak sponsors would be those who are non-investment grade and have been so for a period of time. The House version does not use outside agents to determine risk status.
  
  
• PBGC Variable Rate Premia
  Both the Senate and House versions would maintain the current 90 bps (0.90%) premium on unfunded benefits, but the liability basis used would be more onerous than the one used currently for variable rate premium determination. Funding would also be determined using current market value of assets in both cases. The Administration believes that these remedies are not sufficient to fill the PBGC’s deficit in a reasonable time frame and would point to the fact that the credit spreads for many of the companies who have to pay variable rate premia are sizable multiples of the 90 bps level. They would give the PBGC Board, with representation from the Departments of Treasury, Labor, and Commerce, the right to set variable rate premia at their discretion. Current revisions to the PBGC premium structure require Congressional and Administration approval.
  
  
• Special Dispensation for the Airlines
  The Senate Bill would extend the seven-year amortization of unfunded liabilities to twenty years for the airline industry given their current and unique hardships. Both the House Bill and Administration proposal allow for no special dispensation.
  
  
• Credit Balances
  Under current law, if a sponsor makes a contribution to a pension plan in excess of the minimum required by law, this creates a credit balance that can be used in the future to offset future minimum required contributions. The House and Senate Bills would maintain credit balances albeit in more restricted forms and eliminate their use entirely if a plan fell below 80% funded. The Administration proposal would eliminate them completely.

Now that Senate has filled the Supreme Court, and confirmed Ben Bernanke as the new Chairman of the Federal Reserve Board, look for the pension debate to re-emerge in Washington sandwiched between debate about the budget and other pressing issues. Some form of compromise legislation should arrive on the President’s desk by Easter.

It is amazing to see the eager acceptance of the long-shunned tools of leverage and derivatives. Five years ago, the most difficult sale for a Wall Street derivatives salesman was to a pension plan. Now in the quest to do more with less or to be the first pension plan in your town to have a portable alpha strategy⁷, sponsors are embracing derivatives. For those who understand what they are doing and have carefully thought through the decision and have procedures in place to monitor and manage risks, these can be very powerful and useful tools. But they are tools only and cannot solve the underlying problems in any underfunded plan. Moreover, they do nothing to alleviate the fact that with higher expected return must come higher risk and price volatility.

For those who have not proceeded with care and thoughtfulness, caveat emptor. Like anything else in life, promises that seem too good to be true usually are, and risks overlooked at strategy inception can crop up at the most inopportune times.

One example of an increasingly popular trade is a duration extension overlay through swaps or Treasury Futures. The pitch focuses on the plan sponsor’s twin goals of reducing tracking error and keeping a high enough equity allocation to maintain their current return on assets assumption. There is a psychological element involved: keep your asset allocation the same, and reduce risk through this derivative overlay. Sounds appealing. However, a plan invested in 60% equities and 40% fixed income with a duration completion swap or futures overlay, is almost economically identical to an allocation of 100% duration-matched fixed income backing a 60% equity futures overlay. In fact, a strong argument could be made that the latter strategy is actually superior when considerations such as collateral management and liquidity are taken into account. It is interesting to note that many sponsors will look at these trades as very different, with different perceptions of leverage, although the reality is that they are the same.

Another factor that seems to be under-appreciated is that when you move into levered strategies—even if they are used to control asset-liability mismatch—you need to still be cognizant of absolute volatility. The figure below presents plots of absolute volatility (asset only) versus tracking error (assets – liabilities) for a variety of leverage levels. Each curve traces the range from 100% fixed income to a 100% equity allocation. The two curves furthest to the upper left represent unlevered strategies with bond allocations to either intermediate bonds (e.g.: Lehman Brothers Aggregate) or a liability duration-matching fixed income portfolio (14 years duration). Curves to the right of the unlevered strategies represent increasing levels of effective leverage in the duration overlay. These curves are truncated as it is assumed that the plan sponsor would not be interested in total asset durations much in excess of liability duration. The markers represent equi-equity allocations—the fraction of the underlying portfolio held in equity is held constant—of forty, sixty, or eighty percent equity.

First, note that the only way to decrease both absolute volatility and tracking error is to move money from equities into fixed income. Rule one: there is no free lunch. Second, note that the move from an unlevered intermediate bond portfolio to an unlevered liability duration-matching portfolio trades off absolute volatility for reduced tracking error. However, in the range of equity allocations commonly seen (40-80%), the ratio of the reduction of tracking error to increased absolute volatility varies from 1-5x. Note that as we move further and further into increased leverage (bigger overlays), the slope of equi-equity allocation curves flattens—at some point the overlay is just adding volatility with no tracking error benefit.

Sponsors who do want to reduce tracking error without changing absolute volatility do have some options. For example, the 80/20 portfolio with fixed income in intermediate bonds has very similar absolute volatility (11.3% vs 11.0%) to the 40/60 portfolio with a 50% overlay - essentially a 21-year duration portfolio - but tracking error drops from 16.4% down to 6.7%. The question is what the sponsor is willing to give up in expected returns to achieve the risk reduction. The 40/60 portfolio with 21-year duration is very similar to something that a few sponsors implemented in the late eighties and we are seeing more and more commonly with clients and a few consulting firms. For some clients it may best suit their needs, but it is not optimal for everyone.

If pension funding law and accounting are both moving towards a focus on tracking error, do we still need to care about absolute volatility? Absolutely when we are managing a leveraged asset base! Remember that while pension benefits are economic liabilities, they are not tradable financial liabilities. If interest rates rise sharply, liabilities may go down but they cannot be liquidated for gain. Making good on a levered hedge to satisfy margin collateral calls requires liquidating financial assets. If assets collateralizing the overlay sell off sharply at the same time, it can mean a world of hurt despite falling liabilities.

It is important to remember that the plan in and of itself is a form of economic leverage. This is explicitly recognized in a regulatory framework such as the Dutch FTK or alternatively any modern banking or insurance capital regulation. The sponsor is long the assets and short the liabilities. The surplus represents the residual excess of assets over liabilities and is the only thing that may not be leveraged. Without a surplus (i.e., any underfunded plan), the entire position is leveraged. Even a plan with a funding ratio of 100% or 120% or more can be in fact highly leveraged if one looks through the assets (and possible overlays) to establish the core underlying economics.

Treasury could follow the lead of France, the UK, and others by issuing debt longer than 30 years. France and Britain both initiated 50-year issuances last year to strong demand. These long-long bonds are prized both by pensions and life insurers for their greater stream of cash flows to better match liabilities and by investors seeking convexity. Comparing the French OAT’s of 2035 to the issue maturing in 2055, duration increases by roughly one-third, extending from 17 years to 23 years. Convexity nearly doubles. Because it is prized, it is priced. At issuance last February, the 50-year OAT yielded 3 basis points more than its 30-year counterpart. Today it yields 3 basis points less. The story is similar for the UK issues, which have richened by 4 basis points for both 50-year nominal and index-linked bonds versus 30-year counterparts. In fact, demand for long paper in the UK is so strong and the cost of funding so advantageous for the Crown that there is conjecture that the UK’s debt management office will issue nearly two-thirds of debt sold in 2006 at 30 years or longer.

The figures below show the flattening of the long-end of U.S., French, and UK yield curves for 2005 and January 2006. Figure 4a. shows the flattening in the 10’s – 30’s yield spread in the United States. Figure 4b. shows this for the recently created 30’s – 50’s yield spread in France and the UK.

¹ Forecast rolls forward 2005 estimate for 2006 and 2007. Service cost is estimated at 2.5% of previous year’s liabilities. Benefit payments grow at 5% per annum. Interest cost is based on previous year discount rate. Contributions are assumed at $60B for 2006 and $80B for 2007. Actuarial gain/(loss) – attributable to change in discount rate – and return on assets are stochastic with the following assumptions:

Durations are assumed as 4.5 years for sponsor bond portfolios and 14 years for liabilities. The asset mix is assumed as 55% US Equities, 10% International Equities, and 35% intermediate US Bonds. Note that a negative correlation between interest rate changes and US equity returns implies a positive correlation between equity and bond returns.

² Based on PIMCO estimates for defined benefit contributions and estimates for dividend payments and CAPEX from Morgan Stanley Research.

³ The report is available at http://www.sec.gov/news/studies/soxoffbalancerpt.pdf

⁴ "Let the Games Begin: FASB to Tackle Pensions & OPEB," CSFB, November 10, 2005.

⁵ Sources: Goldman Sachs (3.5%), Morgan Stanley (3.0%), UBS (3.0%).

⁶ For those not familiar with Washington acronym speak: ERISA – Employee Retirement Income Security Act (1974), TEFRA – Tax Equity and Fiscal Responsibility Act (1982), OBRA – Omnibus Budget Reconciliation Act (1986), NAFTA – North American Free Trade Agreement (1994).

⁷ Given their differences as well as the complexity of the current law and whether one defines current law as including the 2004-2005 amendment for determination of discount rate or not, it is impossible to rank-order the three on the left for all plans in all circumstances. The PBGC has evaluated the plans using their stochastic exposure model to evaluate the potential systemic risk that each proposal presents - http://www.pbgc.gov/docs/Reforms10-26.pdf. Their analysis and relative ranking differ from analysis of specific examples by actuarial consulting firms.

⁸ Committee on Investment of Employee Benefit Assets. See press release at: http://www.afponline.org/pub/pdf/PressRelease-Impact_of_Pension_Reform_final_0905.pdf

⁹ James Moore and Olivia Mitchell, "Projected Retirement Wealth and Savings Adequacy," Ch.3 in Forecasting Retirement Needs and Retirement Wealth, University of Pennsylvania Press, 2000.

¹⁰ See Coming Up Short: The Challenge of 401(k) Plans, Brookings Institution Press, 2004 for a thorough treatment on the economics of defined contribution plans.

¹¹ See http://www.theonion.com/content/node/44679