All entries for Friday 10 March 2023

David Miles' Response

Follow-up to Miles and Sefton on USS capital exhaustion from Dennis Leech's blog

Dear Dennis

Thanks for your comments on the various notes James Sefton and I wrote in 2021.

You make two main points: First, in our simulations we look at how the USS finances might evolve in the absence of new contributions and with liabilities (accrued pension rights) taken at some point. Second, we assume that pensions are paid out of the existing assets with a constant portfolio allocation. The evolution of that stock of assets depends on the return (investment income generated plus the capital gain or loss on the assets) net of pensions paid.

In practice the USS is still open – though in some sense its openness is declining because more new pension accruals and inflows will be significantly into a DC fund and perhaps because the rate at which younger academics join the DB scheme looks likely to be less than that of earlier cohorts.

What seems likely is that whether the scheme gets new DB members depends on the deal they are offered and that deal will be more or less attractive depending on whether there are enough assets to pay existing pension liabilities that exist as time passes. This means that to assume the scheme continues to have a steady flow of new DB members joining and that existing members can have pensions paid out of new contributions (as well as from investment income) certainly understates the extent of inter-generational risk sharing. The more existing pensions are paid out of new contributions which are in excess of what is needed to match the present value of new liabilities the worse a deal is offered to active members and the less likely that they remain active.

Our calculations ask a question about whether existing liabilities are such that they can be paid in full out of existing assets. It calculates the probability that they can and – in cases where those assets are more or less than is needed to pay existing pension promises - what is the size of the ultimate deficit or surplus. The calculated probability of running out of money before existing pension liabilities are paid is the probability that future employees or employers will have to make good the hole. If this comes from future active members it will means the present value of extra pensions they get from their contributions are lower than the money they put in. That seems likely to make new members look elsewhere. That generates a risk . Our calculations calibrate the scale of that risk.

You are right that if equities display significant mean reversion – which I think you implicitly assume – then an assumption of independent returns will exaggerate risk. This is why we did simulations with a degree of mean reversion that some studies suggest might exist.

All best

David

Dennis Leech : 10 Mar 2023 16:55 |  Tags: Pensions;Uss |  Comments (0) |  Close comments |  Report a problem

Miles and Sefton on USS capital exhaustion

Writing about web page https://www.niesr.ac.uk/publications/how-much-risk-uss-taking?type=policy-papers

Commentson the paper “How much risk is the USS taking?” by David Miles and James Sefton

The paper addresses the question posed in the title using stochastic simulations to model the likelihood of the scheme’s capital assets becoming exhausted. In this note I argue that the method they adopt, in particular their specific chosen stochastic model, is not a reliable basis for making such a judgment because it exaggerates risk.

It is crucial to understanding whether a pension scheme has “enough money”, to be clear about what that would mean in practice, what assumptions are being made about the progress of the scheme over time, and what kind of money is needed anyway. Thus any stochastic modelling that tries to answer that question should distinguish between, on the one hand, the market valueof the investment assets and, on the other, the investment incomethose investments bring in. These are not equivalent and which is more relevant depends on the framing of the question.

If the scheme is being considered for possible closure, the former view is important, since one must have regard to runoff, where assets are sold to pay pensions as the beneficiaries dwindle. But if it is being seen as possibly able to continue open indefinitely, with the support of a strong employer covenant and a steady influx of new members, then the latter perspective is required since asset sale is not the primary concern. It is important that trustees carry out analyses from both points of view, which are not equivalent.

Miles and Sefton fail to make this distinction. Instead their approach is simply to model the time path of total returns, running together investment income and capital gains/losses - whether realised or not. This means that, from the point of view of answering the question whether the scheme can remain open sustainably, their simulations overstate risk and their conclusions are overly pessimistic. Their stochastic model allows in, and bakes in, too much variation coming from excessively volatile asset markets, whether stock, bond, real estate markets.

Risk will also be overstated as a result of using percentage yields(%)instead of cash income(£)as the random element. Yield expressed as a percentage is inherently excessively volatile because market prices appear in the denominator. On the other hand, absolute investment income in money terms is driven by real economic factors and much less susceptible to this volatility.⁴

The paper’s basic idea is that the investment portfolio can be seen as comprising two classes of assets: ‘risky’ equities, which produce a high return on average but one that is also highly variable, and ‘safe’ bonds, with a low but fairly constant return. The question is essentially whether the equity returns, on average, are sufficiently high to outweigh the combined effects of the low bond yield and that the likelihood of the equity returns underperforming by too much.

It uses a simple model applied to actual USS data starting from the assets at the 2020 valuation, £66.5 billion. and the projected cash flow of pension payments each year up to 2102 by when all accrued benefits will have been paid.

Their conclusion is that, while on average the scheme is well enough funded that it will be able to pay all the pensions promises, there is still a significant probability that the money will run out and the scheme will be unable to pay the promised pension benefits without needing to ask the employers for further contributions. This probability is of the order of over 30 percent for a portfolio with a majority of growth assets. These findings are worrying.

The paper addresses a straightforward empirical question that turns mainly on the relative risk and return parameters of the two asset classes. However there are a number of major issues with this study that may well point to different conclusions emerging if addressed. Most importantly there is an important, basic, methodological issue unaddressed: the relation between asset prices and pensions funding for the USS, with its unique characteristics as a very large, open, immature, multi-employer scheme.

Detailed Comments

1. The methodology employed is questionable because it focuses solely on asset valuesrather than specifically on pension funding, which is likely to lead to an overstatement of risk and therefore biased findings. It examines the likelihood (as a probability estimated in a Monte-Carlo simulation) that the assets will become exhausted - not the likelihood that not all the projected pensions will be paid, without further funding. These two questions are not the same and the difference is crucially important to the stochastic modelling: the former methodology focuses on the market value of the investment portfolio where the randomness mainly comes from the stock market through prices. For the latter methodology, asset values are not of primary concern, since the scheme will only need to sell capital to pay pensions whenever there is a shortfall in income. The study should therefore model the probability of such a shortfall separately.

By ignoring this distinction Miles and Sefton are implicitly making an unreal assumption that is actually key to their results. They end up using an inappropriate risk metric for determining funding - in all circumstances whether the scheme needs to sell assets to pay pensions or not - related to the excess volatility of stock market prices. It would surely be more relevant to use a measure of variation in investment income, that is cash receipts, which is known to be much less volatile, since it is mostly independent of the animal spirits, irrational exuberance, booms and busts, etc to which the stock market is prone. When talking about risk it is necessary to study the second moments of the probability distributions but also to be clear about the metric: are we looking at randomness of capital values or investment receipts?

continue reading

Dennis Leech : 10 Mar 2023 16:39 |  Tags: Pensions;Uss |  Comments (0) |  Close comments |  Report a problem