The state of USS: some observations
David Miles, Imperial College London
This is a USSbrief, published on 24 July 2018, that belongs to the OpenUPP (Open USS Pension Panel) series. This USSbrief represents the views of the author only, and is a summary of evidence given by the author, in person to the 6th meeting of the UCU-UUK JEP (Joint Expert Panel) on 9 July 2018.
The USS ‘Test 1’ is central to its assessment of the sustainability of the scheme. Test 1 compares the maximum extra contributions which universities might have to make to ‘ensure’ that pensions are paid with their ability — in extremis — to do so. This payment is potentially required to achieve ‘self-sufficiency’ of the scheme. It is compared with the greatest payment that could be made without real damage to the university sector.
The way that Test 1 is presented invites one to believe that if it is satisfied, USS members can be sure that they will receive their full promised pensions. There are many problems with that interpretation and they mean that satisfying Test 1 is much less reassuring than it seems. The sustainability of the scheme is much less solid than satisfying Test 1 would seem to imply. Some people have suggested that passing the hurdle of Test 1 is too stringent a test of the sustainability. I believe that view is misguided. I reach that conclusion based on several points.
1. So-called ‘self-sufficiency’ of the scheme is in fact very well short of a situation that scheme members can be sure of receiving their pensions without extra contributions from universities. This is rather fudged in the way the USS defines self sufficiency. The USS says:
Test 1 aims to ensure that the schemes promised benefits can always be funded … (‘Methodology and inputs for the 2017 valuation: initial assessment’, February 2017, italics added).
But in fact ‘self-sufficiency’ means only that pensions can be paid, without extra support, with ‘a high degree of confidence’. How high is that? This is not entirely clear from the documents USS make easily available, but I believe the answer is with a probability of 95%. That would be considered scandalously risky for a bank and completely unacceptable to bank regulators. I suspect that USS scheme members would not consider the scheme to be on a solid foundation if they got their promised pensions 19 times out of 20.
2. Test 1 relies upon an estimate of ‘Technical Provisions’. This appears to be calculated as the present value of expected future pension payments that have been promised, discounted at some rate. The choice of discount rate is absolutely crucial. The rate the USS use is very clearly risky — there is no sense in which it is a rate than can be safely assumed to be earned on the assets. It appears to be calculated by taking a view (a rather optimistic one) on the expected return on asset classes and then making an ad hoc downward adjustment for ‘prudence’.
The expected return on bonds used is not that currently anticipated by holders of bonds in the market. Instead it is assumed that in the future there is a substantial and persistent upward adjustment in yields which greatly helps the solvency of the pension scheme. The justification for this adjustment is weak. It appears to be based on a hunch that yields have been driven below their equilibrium level. Real yields on government bonds have in fact been on a fairly steady and persistent downward trend for at least 30 years and there is no strong reason to expect this to be reversed soon. Of course it might, but to assume that it does is the opposite of prudent planning.
3. The prudence adjustment to other expected yields (for example that on equities) is not well explained in USS publications. It clearly does reduce the chances of underperformance relative to the assumed central expectation (or best guess). But by how much? Does the adjustment reduce the probability of under-achievement over a long horizon, relative to the expected yield, to 1% or 5% or 10% or 25%? In one USS document it seemed to be suggested that the prudence adjustment reduced the chances of underperformance, relative to the mean, to 33%. If that is what ‘prudence’ means it leaves huge risk. Even leaving that much risk would require a large prudence adjustment. Just how large is worth considering with some simple calculations.
Consider, for example, the adjustments you would need to make to the expected return on a portfolio of equites. Let us assume log returns on equities are normally distributed. As an example, suppose the expected annual log return is 4% in real terms and the annual standard deviation of log returns is 15%. These are plausible estimates. Suppose also that returns are independently distributed over time. Finally suppose that the relevant horizon for assessing the risk of the portfolio is 25 years. The following is then true:
The mean log return over 25 years is 25 x 4% = 100%
The standard deviation of log returns is (25 ^ 0.5) x 15% = 75%
The expected LEVEL of the value of the portfolio (whose initial value of 1) is:
Exponential [25 x 0.04 + 25 x (0.15 ^ 2) / 2] = 3.60
The probability that the value of the portfolio is less than its mean value is approximately 65%. This probability is well above 50% because the value of the portfolio is NOT normally distributed — it is log normally distributed. There is a large right hand tail in the distribution and the chances of the portfolio being below its mean value is much greater than 50%. The probability that the portfolio would actually have a loss over the 25 year horizon is about 9%.
How much would we need to reduce the assumed expected return below 4% to give a small chance of underperforming that downward-adjusted mean value if the true expected log return (per annum) were in fact 4%? The answer of course depends on how low you wanted to make that small chance be. To make it a 30% chance you would need to adjust down the annual return from 4% to a “prudent” lower yield of only 1.25%. To get the chance of underperformance down to 10% you would need to reduce the expected yield down from 4% to a prudent level of -1%. So for a common sense meaning of prudence — meaning only a 1 in 10 chance of underperforming the mean outcome — and with an ACTUAL expected annual return of 4% and standard deviation of 15%, you need to adjust down the expected annual return to MINUS 1%.
I do not believe that the USS prudence adjustments are of this order of magnitude. Table 7 of the February 2017 document ‘Methodology and inputs for the 2017 valuation’ shows much smaller prudence adjustments.
I conclude from the above that passing Test 1 should give members of the USS and the sponsoring universities little comfort. To put it another way: the position of the scheme is more precarious than the apparently ultra-prudent USS calculations suggest.
This is a USSbrief, published on 24 July 2018, that belongs to the OpenUPP (Open USS Pension Panel) series. This USSbrief represents the views of the author only, and is a summary of evidence given by the author, in person to the 6th meeting of the UCU-UUK JEP (Joint Expert Panel) on 9 July 2018. The author believes all information to be reliable and accurate; if any errors are found please contact us so that we can correct them. We welcome discussion of the points raised and suggest that discussants use Twitter with the hashtags #USSbriefs36 and #OpenUPP2018; the author will try to respond as appropriate. This work is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International License.
