USS’s valuation rests on a large and demonstrable mistake

When corrected there is no deficit as at 31 March 2018 and no need for detrimental changes to benefits or contributions

[UPDATE 29 Oct: Here I provide further documentation, via reference to statements and a graph in a May 2018 video by their Chief Risk Officer, that USS has made a large and demonstrable mistake regarding Test 1.]

1. Introduction and overview

It is hard to overstate the importance of USS’s confirmation yesterday of the accuracy of union-appointed JNC member Sam Marsh’s modelling of some implications of their valuation. Of greatest significance is their confirmation that Marsh’s calculation of £78.2bn overstates the following by less than 5%: the value that the assets in the scheme have a prudently adjusted 67% chance of reaching or exceeding in 2037 (20 years from now) in the absence of any ‘de-risking’ of the portfolio, any increase in the existing 26% level of contributions, and any decrease in the existing level of DB and DC pension benefits.

USS has thereby provided ‘new material evidence’ which confirms the key figure underpinning Marsh’s analysis that, on a proper understanding of Test 1 — in fact, as I show below, USS’s own understanding — no de-risking of the DB portfolio is called for. If, moreover, such de-risking is cancelled, then the scheme is nearly fully funded on a prudent basis as at March 2017. Given the subsequent improvement in funding position (see Annex 9 of the Joint Expert Panel’s report), the scheme would almost certainly be in surplus on a prudent basis as at March 2018. It is also very unlikely that any increase beyond the current 26% level of contributions would be required in order to continue to prudently fund existing DB and DC benefits.

In other words, when USS’s mistake with regard to Test 1 is corrected, there is no deficit as at 31 March 2018 and no need for detrimental changes involving any decrease in pensions benefits or increase in contributions.

In four words: No Deficit! No Detriment!

This is the key message of this blog post. In what follows, I provide a more detailed and technical defence of the points above. (Those who would like to skip over the technical details should jump now straight to the Conclusion in Part 3 and the Addendum below it.

2. How exactly did USS get things so wrong?

USS is, in effect, confirming that they made a serious mistake regarding Test 1. They assumed an equivalence, which does not obtain, between the value of the scheme’s assets that is 67% likely in a given year and the valuation of the liabilities on a prudent (technical provisions) basis in that year. In other words, USS assumed an equivalence between the figures in the first and the second columns of the table below. Marsh has demonstrated that these figures are not equivalent. USS’s own recent calculations, which they have shared with Marsh, also confirm these figures to be non-equivalent.

Sheffield Working Group’s draft submission to UUK’s JEP consultation

As noted above, USS maintains that Marsh’s column 1 calculations of asset values overstate things, but by less than 5%. (See this tweet for Marsh’s explanation of the discrepancy between his figures and USS’s.) Even if, however, we reduce the row 1, column 1 figure by 5% (from £78.2bn to £74.3bn), the value of the assets remains sufficiently high so as not to trigger any Test 1 mandated ‘de-risking’ of the assets in the DB portfolio. It does not trigger any de-risking even under USS’s current highly restrictive interpretation of the test. This is because, even under this restrictive interpretation, the gap between the level of the assets (column 1) and the self-sufficiency liabilities (column 3) remains within the £10bn ‘reliance on the employer covenant’ that is now permitted by Test 1 (see p. 30).

2.1. USS’s failure to calculate figures for column 1

Until Marsh requested it (and only after repeated requests, extending over the period of a year), USS claims not to have made their own calculation of the value of the assets provided in column 1 above. As revealed by correspondence with Marsh, USS maintains that they did not initially seek “to confirm these [column 1] figures”:

They did not initially seek to confirm these figures, since they simply assumed that they would be equivalent to the figures in column 2:

2.2. Mind the gap

Having calculated the value of the technical provisions liabilities in column 2, and assumed that the asset values in column 1 would be the same, USS then measured the gap between the figures in the second and third columns of row 1 (which assumes no de-risking). That gap is huge. At c. £20bn, it is far in excess of the maximum £10bn gap that Test 1 permits, under its current restrictive interpretation. For that reason, USS mandated the ‘de-risking’ of the scheme’s assets (i.e., a shift from growth assets such as equities and property into bonds) in order to reduce the gap between the second and third columns to the £10bn permitted by Test 1. (The row 3 ‘November de-risking’ begins immediately, whereas the start of the row 2 ‘September de-risking’ is delayed by 10 years.)

2.3. How columns 1 and 2 might diverge

As indicated in tweets above, USS mistakenly assumed that the values in column 1 would be the same as the values in column 2. This tweet by Marsh and the ones above and below it provide an explanation of how, contrary to USS’s assumption, the values in columns 1 and 2 might diverge:

2.4. Why Test 1 mandates application of column 1 rather than column 2 when the two diverge

In claiming that “the trustee’s valuation methodology does not require the projection of assets based on current contribution rates” and therefore they have not “sought to confirm these figures” [in column 1], USS reveals that they have designed Test 1 in a manner that fails to capture its underlying rationale.

This underlying rationale is as follows: that the value of the assets held by the scheme 20 years from now be high enough that purchase of a self-sufficiency portfolio 40 years from now be affordable via a +7% increase in employer contributions from year 20–40. (See this blog post and this blog post for further explanation of this underlying rationale for Test 1.)

Moreover, as I show in this tweet and the ones surrounding it, USS’s own statements affirm a measure of reliance on the employer covenant in terms of differences in asset values:

3. Conclusion

USS’s 2017 valuation rests on a demonstrable mistake of fairly epic proportions. They failed to measure reliance on the employer covenant in a manner that conforms to their own statements of what such reliance involves. Moreover, these are not dispensable misstatements. They capture the essence of the underlying rationale for Test 1 itself.

Had the Joint Expert Panel received Marsh’s analysis in time to influence their report, they would have been provided with strong grounds to recommend further adjustments to the valuation, which bring contribution rates all the way back down to the current 26%, rather than merely down to 29.2%.

Had USS provided Marsh and other members of the University of Sheffield USS Working Group with the column 1, row 1 figure they had requested way back in September 2017, USS’s error over Test 1 would have come to light much sooner. The case would have been made in the autumn of 2017 that Test 1 does not mandate any further de-risking of the pension fund beyond the considerable de-risking that has already occurred during the past decade. This error would have been corrected before it provided Universities UK employers with a rationale for endorsing deep cuts to pensions that prompted industrial action of historic proportions.

Now that this further, serious problem with the valuation has come to light, above and beyond those identified by the Joint Expert Panel, USS must, at the very least, embrace all of the JEP’s recommended adjustments to the valuation in full. The governing majority of UCU and UUK trustees must instruct their executives and the scheme actuary to go at least that far, now.

Addendum: further diagnosis of USS’s mistake

Test 1 triggers false alarms

USS’s mistake comes down to this: they designed an unsound test of whether the scheme is within affordable reach of the safe harbour of a low-risk self-sufficiency portfolio. A sound test is one that delivers no false positives (i.e., false alarms) and no false negatives. USS’s Test 1 is poorly designed because it delivers false alarms. It indicates that the safe harbour is beyond affordable reach when it is not.

This is because the test is blinkered in measuring only whether the following gap is so large that it cannot be bridged by +7% contributions over 20 years: the gap between (i) the liabilities discounted at the low rate of return of a low-risk self-sufficiency portfolio and (ii) the liabilities discounted at the higher (prudently adjusted to 67% chance of success) rate of return x of the actual assets.

This is blinkered because it fails to acknowledge that this gap might be reduced by the scheme’s being in surplus in 20 years’ time because returns of the same prudent rate x on the actual assets generate a level of assets that is higher than the liabilities that have been discounted at that prudent rate x. Such a surplus might boost the level of assets to the point where a self-sufficiency portfolio is within affordable reach via +7% contributions over 20 years.

Marsh’s calculations, as confirmed by USS, establish that there would be just such a surplus, which would be great enough to place a self-sufficiency portfolio within affordable reach. There would be such a surplus in the absence of any de-risking of the current portfolio or changes to the current level of contributions or benefits.

USS’s mistake was in falsely assuming that there would not be such a surplus, but rather that the value of the assets would equal that of the liabilities. Having made this false assumption, its Test 1 sets off a false alarm that the safe harbour of a self-sufficiency portfolio is beyond affordable reach when it is, in fact, within reach.

[UPDATE: USS has posted a response to this blog post, and here is my explanation of how precisely it fails to answer my charge of “large and demonstrable mistake”.]