Yesterday, Andrew Biggs of AEI presented the above-named paper. (Title at page 1 of the link appears to be different, but it actually is indeed "Public Employee Pensions: Investment Risk and Contribution Risk.")
The paper addresses the funding issues posed by defined benefit (DB) plans for state and local government employees around the country. This issue is a well-known time bomb that will either explode or else not, possibly in the near future, depending not only on what's done but also on the plans' good versus bad luck with their investments, many of which are in risky equity.
Currently the DB plans are estimated, if I recall correctly, to have about $1 trillion in present value of promised future benefits that remain unfunded. However, this is based on discounting future benefits at a relatively high discount rate, which makes them look smaller. The high discount rate (say, 8%, although it varies with the plan) is based on the expected rate of return from fund investments, which often are quite risky.
Given this actuarial convention in making the estimates - which isn't permitted for private sector DB plans that are subject to ERISA - suppose a DB plan is investing its assets at a very safe 2% rate. It will appear to have very high unfunded liabilities, since they will be valued using this 2% discount rate. The plan's managers decide, therefore, to bet the ranch. They shift to very risky investments with an expected return of, say, 8% or 10%. But this is not free money, of course. The market values the underlying assets the same as the stodgy old 2% assets, because these items could hit a home run, on the one hand, or blow up, on the other.
Was this a wise investment choice by the plan? Very possibly not, since the downside would leave it with gigantic unfunded liabilities. But if I understand correctly how things are being done, the shift would cause the plan to appear as if it were much better-funded, because it could now use the 8% or 10% rate to discount its future liabilities. This can both create dubious incentives, and mislead the audience for funding estimates to overlook the risk of the downside scenario in which the pension plan would end up being very short-handed.
The paper focuses on the question of how investment risk should lead us to think about the DB plans' future prospects - in particular, the risk that, if the plan investments do badly, there will either be defaults (or at least unexpected cuts) or a need for greatly increased contributions by workers or employers. In particular, based on a Monte Carlo simulation where assets such as stock are assumed to have an 8% expected return (or lower in one of the scenarios), with a 12 percent standard deviation, it evaluates probabilities for insolvency, required contribution volatility, etc. In particular, it offers parameters for the unsurprising, if disappointing, conclusion that you can't keep current contributions low and use risky assets to juice the expected return without creating significant contribution volatility and default risk. In other words, using riskier assets fails to offer a free lunch in the paper's simulations (although it does of course turn out nicely for DB plans' stakeholders in the set of cases where the investments do well rather than poorly).
The quatrilemma whereby you can't have low contributions, plus long adjustment periods to make up shortfalls, plus low contribution volatility, plus low default risk, is based on presuming that a theory to the contrary that the DB funds' proponents and stakeholders sometimes advance is not in fact correct. This is the theory that mean reversion in stock market returns means that long-term players, such as state and local governments if they can weather the interim storms, actually do get a free lunch in the form of higher expected returns without greater long-term risk.
To illustrate, suppose that we thought the stock market really was pretty much a lock to earn, say, 8% over time. Then, if you can wait long enough, you can capture the higher return without bearing the variability that torments more short-term players. For this to be true, however, periods of lower or higher returns must not be merely diluted, but positively offset. E.g., if the market's been returning only 6% for a while, it's now likely to earn more than 8% for an offsetting period that gets it back on track.
Conceptually, it's as if an honest coin that yielded 5 straight heads is now likely to even things by running more tails for a while. But admittedly the stock market is a sufficiently complicated beast (e.g., given the role played by market psychology) that we can't rule it out as decisively as we could in the coin example. Still, it would require a rather odd market inefficiency that you would expect savvy investors to exploit. (E.g., if the market is due to start earning an unusually high rate of return, the price level ought immediately be bid up to the point where we are back to the normal rate of return.)
I'm certainly open to theories of market inefficiency. But, while this really is not my area, mean reversion does not appear to me to be among the more plausible candidates. (And indeed I gather that Bob Shiller, who has gotten a Nobel Prize for writing about market inefficiencies, does not believe in it.) For now it's just worth noting that, with mean reversion in stock prices or even a little bit of it, the dilemmas (or, rather, trilemmas and quadrilemmas) discussed in the Biggs paper might become at least slightly less binding.
Insofar as the existence of some or any mean reversion in relevant asset prices remains uncertain, I suppose you could say that supporters of the DB plans' current investment strategies and actuarial reporting conventions are taking a risk that they actually aren't taking on as much risk as they seem to be. And even if the resolution of the second-order risk remains ambiguous, we will certainly learn at some point how their first-order risk-taking has come out.