Why ‘Test 1’ must be dropped: a critique of its design and implementation

Number 68: #USSbriefs68

Andrew Chitty, University of Sussex
Tim Wilson, University of Dundee

Click for printable PDF

This is a USSbrief that belongs to the OpenUPP (Open USS Pension Panel) series.

There are many problematic aspects to the USS’s 2018 valuation consultation document (USS, 2 Jan 2019). It is full of weak and poorly expressed arguments. Its tone is obdurate and even truculent. But at root what is objectionable in it is its intransigent response to the six proposals made in the Joint Expert Panel (JEP)’s first report in September (JEP, 13 Sep 2018, p. 53). These proposals were intended as adjustments to the 2017 valuation, but after USS announced in November that they would replace this by a 2018 valuation both UUK and UCU called on USS to apply the JEP’s recommendations to the new valuation (Jarvis, 14 Nov 2018; Bridge, 7 Dec 2018).

The document’s default position is to accept only the two most anodyne out of the JEP’s proposals, adding in two more of its own. On this basis it calculates that, in order to maintain benefits, contributions will need to go up from the present 26% to 33.7% of salary. This is not far off the crippling figure of 36.6% that USS have meanwhile mandated as the third stage of their rule 76 ‘cost sharing’ arrangement based on the 2017 valuation, which is due to come into force in April 2020 if nothing better can be agreed before then (USS, 10 Dec 2018).

The document adds that it is willing to accept two more of the JEP’s proposals, but only if employers sign up to unspecified extra ‘contingent contributions’, to be triggered in the coming three-year cycle if certain indicators of ‘risk’ go above a pre-agreed level. On this basis it calculates that regular contributions would only need to go up to 29.7%. Meanwhile the document rejects the last two JEP proposals out of hand.

For comparison, the JEP calculated that if all six of its proposals were applied to the 2017 valuation then contributions would only need to go up to 29.2%, and USS have calculated that if all six were applied to the 2018 valuation they would go down to 25.5% (Aon, 14 Jan 2019, p. 4).

Of the four proposals that the document rejects, or else accepts only on condition of ‘trigger contributions’, two refer to the role of USS’s notorious ‘Test 1’ in the valuation:

  1. Defer the start of ‘de-risking’ by 10 years
  2. Increase ‘target reliance’ from £10bn to £13bn

What is Test 1? In a nutshell, it requires that the USS scheme’s investment strategy be such that in 20 years’ time there is enough money in the pension fund so that, if the scheme were to close to new accruals and contributions, the employers could afford to top up the fund to the point where it would be almost certain of fulfilling all the pensions promises made up to that date without further aid. It is this Test that drives USS’s strategy of progressively ‘de-risking’ the scheme over the next 20 years, that is, moving its fund out of equity-like investments with higher returns but higher ‘investment risk’ (risk of not producing the expected returns) and into bond-like investments with lower returns and lower investment risk — a strategy which has the effect of increasing the scheme’s ‘deficit’ and forcing up contribution rates. The JEP’s above two recommendations aim to reduce the ‘de-risking’ effects of Test 1.

The term ‘de-risking’, in contrast with a more neutral alternative such as ‘equity-bond shifting’, is a tendentious and effective piece of spin. For, while ‘de-risking’ reduces the investment risk of the portfolio, it also drives contributions progressively up and increases the chance of their becoming so high that the Defined Benefit (DB) scheme is forced into closure, with future pension provision likely then based on a Defined Contribution (DC) scheme in which employees would have to bear all the pension risk. The former risk is largely born by the employers and the latter largely by employees. For lack of an established alternative we shall use the term ‘de-risking’, but with inverted commas to remind the reader that it represents a transfer of risk, not an elimination of risk.

In this USSbrief we explain Test 1 and show how it works to drive ‘de-risking’. Out of the many criticisms that have been made of Test 1 we focus on one: that it is unfit for its own underlying purpose. We point out that both the JEP and the Pensions Regulator have explicitly or tacitly endorsed this criticism. We develop the criticism by showing that in fact Test 1 is designed to fulfil two underlying purposes, and that it is unfit for either of them. We argue that, both as incapable of fulfilling its own purposes and as an architect of crisis in the USS scheme, Test 1 must be dropped from USS’s valuation methodology. At a minimum, in the 2018 valuation USS should accept both of the above two JEP recommendations aimed at mitigating the effects of the Test, without strings attached.

1. USS’s basic funding method

To understand Test 1 we need to first summarise the main steps of USS’s basic valuation methodology or ‘funding method’: its method for assessing the level of salary contributions needed to cover pension payments. The Pensions Act 2004 and the Occupational Pension Scheme Funding Regulations 2005 together state that DB schemes must use what is called an ‘accrued benefits’ funding method (UK Gov, 2004, sec. 222; UK Gov, 2005, para. 5). It works as follows.

Every three years the scheme managers must calculate a figure for the scheme’s ‘technical provisions’ on a particular date, called the ‘valuation date’. The Pensions Act defines technical provisions simply as ‘the amount required, on an actuarial calculation, to make provision for the scheme’s liabilities’ (UK Gov, 2004, sec. 222). More exactly it is the sum of money on the valuation date which according to actuarial calculations would be enough, if invested, to meet all the scheme’s ‘accrued liabilities’: its obligations to pay future pension benefits that have resulted from service by scheme members up to that date. To calculate this sum the scheme managers must use life expectancy, salary and inflation predictions to estimate the accrued liabilities, and then use a ‘discount rate’ so as to convert these liabilities back into a single present-day sum of money such that, if it were invested with that discount rate as the rate of interest, it would be just enough to meet all the liabilities as they fall due. They must use this sum as the figure for the scheme’s ‘technical provisions’. Discounting liabilities that will fall due in the future in this way so as to convert them into a present-day sum of money is called ‘valuing’ them, so the technical provisions figure is sometimes called the ‘technical provisions value of liabilities’.

The Scheme Funding Regulations do not specify how scheme managers must settle on the discount rate they use to calculate a technical provisions figure. They say only that the discount rate must be chosen ‘prudently’ and taking into account either the expected rate of return on the scheme’s assets, the yield on high-quality bonds, or both. Beyond that ‘it is for the trustees or managers of a scheme to determine which method and assumptions are to be used in calculating the scheme’s technical provisions’ (UK Gov, 2005, para. 5).

Once they have decided on a discount rate and used it to calculate a technical provisions figure, the scheme managers must calculate how far the market value of the scheme’s assets on the valuation date falls short of (or exceeds) the technical provisions figure. If the two figures are the same the scheme is ‘fully funded on a technical provisions basis’. If the value of the assets is lower (or higher) than the technical provisions figure then the difference is the scheme’s ‘technical provisions deficit (or surplus)’. If there is a technical provisions deficit then the managers must initiate a ‘recovery plan’, financed by ‘deficit recovery contributions’, to bring the value of the scheme’s assets up to the technical provisions figure within a certain number of years (UK Gov, 2005, para. 8).

Finally, the managers must look ahead to the coming years, predict the new pension liabilities that will be created each year through members’ service, and use a discount rate, normally the same as the one above, to calculate the annual level of contributions needed to meet those liabilities. These are called ‘future service contributions’, since they are needed because of members’ service done after the valuation date.

USS use this ‘accrued benefits’ funding method, but within it they have adopted a specific method for deciding on the discount rate. They begin by assuming a certain investment strategy for the fund over the long term (meaning, say, the next 30 or 40 years), with a certain, possibly changing, portfolio of different assets. Then they calculate an ‘expected’ or ‘best-estimate’ prediction of the rate of return on that portfolio. From that they calculate a ‘67% prudence-adjusted’ prediction of the rate of return: one such that there is a 67% chance that the actual rate of return will turn out to be at least as high as that. Then they use this last figure as the discount rate to calculate a technical provisions figure, a technical provisions deficit (if any) and levels of deficit recovery and future service contributions.

Then it is up to UCU and UUK, negotiating through the Joint Negotiating Committee (JNC), to decide how to divide the contributions between employers and members, although since 2014 contribution increases have been allocated 65% to employers and 35% to members. The JNC can also decide to reduce the level of benefits earned by future service so as to keep future service contributions down. If the JNC can’t reach a decision, then under its rule 76 USS can maintain the existing benefit levels and impose the necessary contributions needed in a ‘cost sharing’ 65%/35% ratio, as it is currently doing (USS, Feb 2018, rule 76.4).

2. ‘De-risking’

It is crucial to see that given USS’s basic funding method the discount rate, and therefore the technical provisions figure and contribution levels are all ultimately determined by the long-term investment strategy the scheme managers adopt from the valuation date. For the discount rate is derived from the 67%-prudent expected rate of return on investments, and different types of investments have different expected rates of return. Since scheme managers are free to vary their long-term investment strategy, they are free, within limits, to vary the discount rate and thus all the figures which depend on it. To that extent the technical provisions figure, and so the size of the technical provisions deficit, is decided rather than discovered.

The managers’ most fundamental decision in planning a long-term investment strategy is what proportion of the fund to hold in the form of equities (and similar ‘higher return, higher risk’ investments) and what proportion in high-quality bonds (and similar ‘lower return, lower risk’ investments). The more they revise their strategy towards holding more (high quality) bonds in future, that is towards a ‘de-risking’ investment strategy, the lower the resulting discount rate must be, and therefore the higher the technical provisions figure, the larger the technical provisions deficit (if any), and the higher both the deficit recovery and future service contributions.

In fact there has been a generalised move by British DB fund managers to engage in ‘de-risking’ in the last 15 years (Cowling et al, 2017, p. 19). The pressures on them to do this are complex but we can mention three. First, a global drive towards market-based methods of valuation has led to accounting standards that count a company’s pension scheme liabilities as a debt on the company, worsening the terms on which it can issue shares or borrow, and incentivising companies to reduce the risk on such debts by funding their pension promises with bonds (Cowling et al, 2017, pp. 10, 26). Second, the Pensions Act 1995 and especially the Pensions Act 2004 have embedded the same drive towards market-based valuation in law, in that they demand that companies replace the older ‘cash flow forecast’ funding method by the ‘accrued benefits’ funding method described above, which discounts the scheme’s liabilities to assign them a present-day ‘value’. And third, when the Pensions Regulator (TPR) was established by the Pensions Act 2004 as the body to oversee pension schemes the key objectives given to it were to protect members’ pension benefits and to reduce the risk of pension schemes having to enter the Pension Protection Fund (UK Gov, 2004, sec. 5). In line with this, TPR has from its start adopted what might be called an ‘insolvency mentality’: it has been consistently preoccupied by the possibility that the company sponsoring a DB pension scheme may become insolvent at any time, leading to the pension scheme having to close and go into ‘run-off’. As a result TPR has focused on ensuring that should this happen the scheme would be certain of being able to meet its accrued liabilities out of its own resources and without government help. For that the scheme’s fund would have to be sufficiently large, but ideally it would also have to be invested in ‘matching’ assets, assets that would guarantee a stream of income to pay benefits as they fall due from day one. Since the assets which come closest to doing that are high-quality bonds, the ideal from TPR’s point of view is to see all DB pension funds ‘de-risked’.

It has been argued that these pressures, by forcing private DB schemes to adopt ‘de-risking’ investment strategies, have driven up their contribution rates and led ultimately to their closure on a large scale (Bridgen, Aug 2018). The move from the ‘cash flow forecast’ to the ‘accrued benefits’ funding method has been especially heavily criticised. (Leech, Jul 2016; Wilkinson and Curtiss, May 2018). However the fact is that none of the above three pressures (with the possible exception of the first) needs to apply to USS, which is a scheme funded not by a single private company but jointly by 350 semi-public institutions that are extremely unlikely to become collectively insolvent.

Despite this USS have followed the general trend of DB schemes. They have been steadily ‘de-risking’ the scheme since around 2007, the year the Pensions Act 2004’s regulatory framework came into full operation. See the graph below.

The figures for 2008 and 2009 may be affected by the financial crisis which led to a crash in equity prices, but these prices had recovered by 2015 so it does not affect the overall picture. Unfortunately in their annual reports USS do not disaggregate what we call in the graph ‘other’ assets into equity-like and bond-like types. However even if all the rise in these investments since 2007 was equity-like in character the graph would still show a consistent ‘de-risking’ process at least since 2010.

Still, until 2014 USS’s ‘de-risking’ was essentially a strategy without an argument. The only rationale given for it in the annual reports was the occasional reference to ‘diversification’. USS provided their first systematic justification for ‘de-risking’ with the launch of ‘Test 1’ in their 2014 document ‘An integrated approach to scheme funding’ (USS, Jul 2014). This was less than a year after Bill Galvin moved from CEO of TPR to become CEO of USS (Rainey, Mar 2013; Otsuka, Aug 2018). So the suspicion must be that Galvin brought an ‘insolvency mentality’ with him from TPR and determined to apply it systematically to the management of a scheme where that mentality is perhaps less appropriate than in any other outside the public sector. Meanwhile, coincidentally or not, at this time UUK were engaging in a coordinated drive to persuade employers of the benefits of closing the DB scheme and transferring members into a DC scheme (Callard, Apr 2018).

3. Self-sufficiency and reliance

The ‘An integrated approach to scheme funding’ document lays out three ‘guiding principles’ for scheme funding, and three supporting ‘tests’, but it is Guiding Principle 1 and Test 1 that take pride of place. To understand them we need a few concepts.

First, a pension scheme is generally called ‘self-sufficient’ if there is a very low probability that it will need to call on the sponsor for additional funding (Shaw, n.d.). USS have a more specific definition: they define the ‘self-sufficiency value of liabilities’ on a given date as the sum of money which if it were invested in a suitable portfolio (assumed to be predominantly in bonds) would be 95% certain of meeting all the scheme’s liabilities accrued up to that date without outside aid. For the scheme to be ‘self-sufficient’ is for the value of its assets to be equal to this figure.

The self-sufficiency value of liabilities is a parallel concept to the technical provisions value of liabilities. The difference is that, whereas the discount rate used to calculate the technical provisions figure is based on the 67% prudence-adjusted expected rate of return on the planned portfolio mix over the long term, the discount rate for calculating the self-sufficiency value of liabilities is based on the best-estimate expected rate of return on a hypothetical portfolio held over the long term composed mainly of bonds (USS, Sep 2017, pp. 19–20). Although the technical provisions figure incorporates a 67% prudence margin, the fact that an actual planned portfolio would normally have a significant proportion of equities means that the discount rate used to calculate the self-sufficiency value of liabilities will be lower than the one used to calculate the technical provisions value of liabilities, and correspondingly the self-sufficiency value of liabilities figure will be higher than than the technical provisions figure. In fact it will often be much higher. In the current consultation document USS use a discount rate of gilts +0.75% (0.75% above the yield on long-dated gilts) to calculate the self-sufficiency value of liabilities and a discount rate of gilts +1.2% to calculate the technical provisions value of liabilities (USS, 2 Jan 2019, pp. 25, 18).

The scheme’s ‘self-sufficiency deficit’ is then defined as the gap on a given date between the self-sufficiency value of the liabilities and the value of the assets, or SS minus A. So it is parallel to the technical provisions deficit, the gap between the technical provisions value of the liabilities and the value of the assets, or TP minus A. Again, the self-sufficiency deficit will be larger than the technical provisions deficit.

We shall say that the scheme is ‘quantitatively self-sufficient’ if it has no self-sufficiency deficit, and ‘quantitatively and qualitatively self-sufficient’ if it has no self-sufficiency deficit and furthermore it is already invested in a ‘self-sufficiency-type portfolio’, thus one composed mainly of bonds. The difference between the two types of self-sufficiency is significant because the scheme is so large that it may take some time to shift its assets from one type to another without its own market operations driving the price of the first type down and the second up. So on a given date the scheme may be quantitatively self-sufficient but, because it is overly invested in equities, some time away from becoming also qualitatively self-sufficient.

The second concept is the scheme’s ‘reliance on the covenant’, or simply ‘reliance’. Unfortunately at different times USS have used ‘reliance’ to mean two different things. Since early 2017 they have defined it to mean the self-sufficiency deficit, or SS minus A:

Reliance is the difference between the assets held at a given point in time and those assets required to allow the scheme to be self-sufficient […] (USS, Sep 2017, p. 26, cf p. 34; cf Feb 2017, p.10; 2 Jan 2019, p. 12)

However up until early 2017 they defined it to mean the gap between the self-sufficiency value of liabilities and the technical provisions value of liabilities or SS minus TP:

The reliance on the covenant is measured by comparing the value of the liabilities on a technical provisions basis with a calculation of the liabilities on a self-sufficiency basis […] (USS, Oct 2014, p. 11; cf Nov 2016, pp. 11, 13; Feb 2017, p.27)

Let us call these two ideas reliance in the ‘primary’ and in the ‘secondary’ sense. Reliance in the primary sense is the amount for which the scheme would have to ‘rely’ on employers to get the value of its assets up to the self-sufficiency value of its liabilities. Reliance in the secondary sense is the amount for which the scheme would have to rely on employers to get the value of its assets up to that level if it were fully funded on a technical provisions basis, i.e. if it were the case that A = TP. If we look to a point far enough ahead in time and assume that by that time any technical provisions deficit or surplus has been eliminated, so that A = TP, then the figures for reliance in the primary and secondary sense will be the same. However they are clearly distinct ideas. For clarity, wherever possible we shall use ‘the self-sufficiency deficit’ for reliance in the primary sense and ‘the self-sufficiency technical provisions gap’ for reliance in the secondary sense. In the context of a culture that celebrates individual independence ‘reliance’ suggests weakness and inadequacy, so it is another tendentious term.

To put some flesh on these concepts, the new consultation document estimates the following figures at the 31st March 2018 valuation date (USS, 2 Jan 2019, p. 18):

In general:

Self-sufficiency deficit = self-sufficiency technical provisions gap + technical provisions deficit

The third concept is that of future discount rates, technical provisions, self-sufficiency values of liabilities, and values of assets. So far we have thought of all these figures as calculated at the valuation date. But as long as USS have an investment strategy stretching far enough ahead they can apply the basic method outlined above to work out a predicted discount rate at year 1 (i.e. a year after the valuation date) and thereby a technical provisions figure for year 1. Likewise for any other future year. So USS can not only calculate a technical provisions figure at the valuation date but also predict a sequence of TP figures for the coming years (TP1, TP2 etc.) assuming that they stick to the same investment strategy they have adopted. These are the figures which USS predict the value of the scheme’s assets will need to track in order to keep pace with the liabilities progressively accruing over those years. Likewise apart from calculating the self-sufficiency value of liabilities at the valuation date they can predict an SS figure for each coming year (SS1, SS2 etc.) and the value of the assets for each coming year (A1, A2 etc.). Thus they can also predict future figures for both the self-sufficiency deficit and the self-sufficiency technical provisions gap. We shall assume that inflation is allowed for in all these predictions, so that when we refer to, say, £10bn in a future year we shall always mean £10bn in real (CPI-adjusted) terms, benchmarked back to the valuation date.

The fourth concept is that of ‘covenant strength’, which means the degree to which the employer is financially able to support the scheme in case of need. This leads directly to the concept of ‘maximum reliance’, which is an attempt to put a number on the vague notion of covenant strength. USS define maximum reliance as the maximum sum that employers would be able to pay, over a period of years starting from a given date, to support the scheme in case of need. It is a ‘maximum reliance’ in the sense of the largest reliance (in our primary sense of reliance, self-sufficiency deficit) that employers could afford to ‘bridge’ or eliminate through financial contributions. USS also use a number of other terms for this concept, including ‘employer reliance’, ‘supportable reliance’, ‘reliance capacity’, ‘the amount of contributions payable in extremis’, ‘employers’ risk budget’ and ‘employers’ risk appetite’.

Of course all these figures are harder to predict the further ahead in time we look. USS call the period over which they can form a realistic estimate of covenant strength, and therefore of maximum reliance, the ‘covenant horizon’. In ‘An integrated approach to scheme funding’ they estimate the covenant horizon as the 20-year period starting from the valuation date.

4. Test 1

With these concepts in place we can go to the introduction of Guiding Principle 1 and Test 1 in the ‘An integrated approach to scheme funding’ document. Guiding Principle 1 is that during the ‘covenant horizon’ there should be ‘no increase in USS’s reliance on the covenant of the sector’ (USS, Jul 2014, p. 10). Apparently ‘reliance’ here is used in our primary sense to mean self-sufficiency deficit (though it is hard to tell). The announcement of Test 1, which is supposed to ‘support’ Guiding Principle 1, comes a few pages later:

Test 1: Benefit security and additional contribution cover
The difference between the liabilities assessed on a self-sufficiency approach (for this purpose a discount rate of gilts plus 0.5% is used) and the actual technical provisions basis should generally not exceed what we refer to as the amount of contributions payable in extremis, which we will indicatively measure as the difference between (i) the maximum contribution of 18% of salaries stated by the employers as being desirable and (ii) the maximum identified as being affordable by employers […] of 25% of salaries, over a long period such as 15 to 20 years.
The rationale is that, at any given time, the trustee could be required to replace the investment returns assumed in the funding of current benefits with additional contributions from the participating employers, if such a response were needed due to scheme or economic circumstances.
In considering the development over time of the relationship between the liabilities measured on a self-sufficiency basis and on the technical provisions basis, the position at the end of a 20 year horizon will be used. The size of the technical provisions at the end of 20 years will be determined so that the difference between it and the self-sufficiency value of liabilities is maintained broadly constant. This informs the trustee of the size of the technical provisions required, and from that the required investment strategy can be derived.
It’s the gap to the self-sufficiency funding level that is critical, and that is maintained (and not allowed to grow disproportionately) by keeping the technical provisions value at a sufficient level over time. (USS, Jul 2014, p. 14)

So Test 1, as USS later put it succinctly, ‘checks that the difference between self-sufficiency and technical provisions in 20 years time does not become too large for the employers to support’, that is, does not become greater than the ‘maximum reliance’ (USS, Sep 2017, p.58). USS initially took the maximum reliance to be 7% of the total sectoral payroll over 15 or 20 years, but by 2017 they had modified this last figure to 20 years and made it clear that they had in mind years 20 to 40, so we shall use that assumption (USS, Feb 2017, p. 4). In short, Test 1 takes the basic valuation method we described above, predicts the resulting values for SS20 and TP20, and then asks whether:

SS20 – TP20 ≤ 7% of payroll for years 20–40

where ‘≤’ stands for ‘is less than or equal to’.

The most obvious thing to say about this Test is that there seems to be a mismatch between its underlying purpose, or ‘rationale’, and its design. Its underlying purpose is fairly clear from paragraphs 2 and 4 of the above passage. It is to ensure that in future the self-sufficiency deficit does not exceed the employers’ ability, in case of need, to bridge it. In the analogy favoured by USS, it is to ensure that the scheme is always within reach of the ‘safe harbour’ of being quantitatively self-sufficient. Here is how USS described this underlying purpose in 2017:

Test 1 aims to ensure that the scheme’s promised benefits can always be funded, with a high degree of confidence using a low risk investment portfolio from within a level of future contributions which could be credibly paid in extremis from the sector’s operating cash flows. (USS, Feb 2017, p. 10)

This would suggest that the Test should simply require that at year 20 the scheme’s predicted reliance in the primary sense (the predicted self-sufficiency deficit, SS20 minus A20) be no greater than the maximum reliance. For this might plausibly ensure that over the intervening years the self-sufficiency deficit is never much greater than the maximum reliance. Yet what the Test actually requires is that the scheme’s predicted reliance in the secondary sense (the self-sufficiency technical provisions gap, SS20 minus TP20) be no greater than the maximum reliance.

However we shall postpone looking at this mismatch and focus for the moment on how Test 1 drives ‘de-risking’. Here there are two key points.

The first is that the Test does not merely ‘check’ whether the above formula holds true. It requires that TP20 be set so as to make it hold true. So if a preliminary valuation done by the basic method above leads to the prediction that TP20 will fall short of SS20 by more than the maximum reliance then that valuation ‘fails’ the Test and must be revised so that it will satisfy or ‘pass’ it. This means that Test 1 adds an extra layer to the basic funding method described above.

The second is that the way in which a preliminary valuation can fail the Test is by virtue of it assuming the wrong kind of long-term investment strategy. For by requiring that TP20 be set at a certain level Test 1 indirectly requires a certain investment strategy from year 20 onwards. This is because for USS ‘technical provisions at year 20’ does not simply mean ‘the sum we presently predict will be needed at year 20 to meet the liabilities accrued by that time’. It means ‘the sum we so predict will be needed using the method of taking the 67% prudence-adjusted expected rate of return on the portfolio in the long-term investment strategy from year 20 on as the discount rate’. So by requiring that TP20 must be at least SS20 minus 7% of the payroll for years 20 to 40, Test 1 indirectly imposes a requirement on the long-term investment strategy from year 20. As the above passage puts it, the Test ‘informs the trustee of the size of the technical provisions required, and from that the required investment strategy can be derived’ (our emphasis). A report produced at the time by UCU’s actuaries First Actuarial spells this point out clearly:

[I]t seems very clear that the valuation methodology is driving the investment strategy. It seems to us that the trustee has determined that:
• Technical provisions must not be more than a fixed margin less than the self sufficiency value
• The discount rate for technical provisions is set to ensure this is the case.
• The investment strategy is altered to follow the discount rate. (First Actuarial, Nov 2014, p. 19)

In turn a given investment strategy from year 20 onwards will entail what will turn out to be a ‘de-risking’ investment strategy between now and year 20.

5. How Test 1 drives ‘de-risking’

To see exactly how Test 1 drives a ‘de-risking’ investment strategy we shall set out its procedure in steps. We have used the best account of this procedure given by USS that we can find, and Sam Marsh’s clearer exposition in his first submission to the JEP (USS, Nov 2016, pp. 5–6; Marsh, Jul 2018).

We begin by presenting the Test procedure as USS themselves present it, as a ‘stop-and-check’ procedure, in which it is used to check a preliminary long-term investment strategy and to revise it only if necessary. We then present briefly it a second time as a ‘target-setting’ procedure, in which it is used to decide the scheme’s long-term investment strategy from the very start. In the end the results are the same.

1. By whatever means, settle on a preliminary long-term investment strategy from the valuation date. For simplicity, let us assume that this strategy is a ‘maintenance’ or ‘no de-risking’ strategy: one of simply keeping the present portfolio mix indefinitely.

2. Use the basic method above to first predict all the liabilities that will have accrued by year 20 and then, given the above portfolio mix, to work out a 67%-prudent expected rate of return on this portfolio mix from year 20 onwards. Use this as the discount rate at which to discount the liabilities back to year 20 so as to calculate a predicted technical provisions figure for year 20 (TP20).

Note that no accrual of liabilities is considered beyond year 20, so the background scenario seems to be one in which the DB scheme closes at year 20 but is supported by employers for another 20 years until it reaches self-sufficiency. Here we differ from Michael Otsuka, who suggests instead that it is one in which employers start contributing an extra 7% a year at year 20 to enable the scheme to close at year 40 (Otsuka, Jul 2018).

3. Work out the expected rate of return on a hypothetical self-sufficiency-type portfolio held from year 20 onwards and use that as a discount rate to predict the self-sufficiency value of accrued liabilities at year 20 (SS20).

4. Predict 7% of the total sectoral payroll from years 0 to 20. Roll this forward at CPI to give a prediction for 7% of the total sectoral payroll from years 20 to 40, and use this as the figure for maximum reliance at year 20. Arguably USS should have allowed for salary growth above CPI when rolling forward, but we shall leave this point aside. At the 2017 valuation USS provisionally calculated maximum reliance as £13bn, but then reduced this figure to £10bn (USS, Feb 2017, p. 23, Sep 2017, p. 19). The second JEP proposal mentioned above is to go back to £13bn. We shall use the figure of £10bn.

5. Check whether, on the above predictions, SS20 – TP20 ≤ £10bn. If so, the preliminary long-term investment strategy has passed the Test, and needs no revision. If not, then it has failed the Test and we need to move on to step 6.

Since USS currently predicts SS20 at £81bn, and £10bn is a relatively small proportion of this, the portfolio mix needed for a ‘maintenance’ investment strategy to pass Test 1 would have to be relatively similar to a self-sufficiency-type portfolio composed mainly of bonds. If the present USS portfolio were already of that kind then a ‘maintenance’ strategy would likely pass the Test. However in fact the USS portfolio is presently about 60% in equity-like investments (Coughlan, Feb 2018). Therefore a ‘maintenance’ strategy is almost guaranteed to fail it.

6. Using the same data as before, work out what discount rate at year 20 would be low enough to produce a figure for TP20 just high enough to pass the Test, i.e. such that SS20 – TP20 = £10bn.

7. Work out what 67%-prudent expected rate of return on the portfolio for the long-term period following year 20 would be required in order to give that discount rate using the USS’s basic funding method.

Steps 6 and 7 reverse the usual procedure of calculating a TP figure from a planned portfolio of assets that has been decided in advance. They are central, so we quote them in USS’s own words:

[6] Having determined the technical provisions at the end of the covenant horizon, the discount rate at that time can be calculated from the projected pension payments. This discount rate applies only for the time period beyond the covenant horizon.
[7] Knowing the discount rate at the end of the covenant horizon means that we also know the required aggregate expected return on the portfolio of assets at that time (USS, Nov 2016, p.6)

8. Choose an investment strategy for the period from now to year 20 which ‘de-risks’ the scheme’s present portfolio over that period so that by year 20 the 67%-prudent expected rate of return on the portfolio matches the one required for the period following year 20. Adopt this as the revised investment strategy for the period from now to year 20.

In 2014 USS originally proposed a ‘linear’ path for this ‘de-risking’ process, with the 67%-prudent expected rate of return on the portfolio-as-planned, and correspondingly the discount rate, falling more or less in a straight line to the one required for year 20 as the portfolio mix changes. However following the steep fall in bond yields (and concomitant rise in bond prices) from 2014 to 2017, in September 2017 they revised their view and proposed that the scheme keep the current portfolio mix for 10 years and then move progressively over years 11 to 20 to the required final position (USS, Sep 2017, pp. 9–10). This is ‘deferred de-risking’. Then in November 2017 they reverted to a ‘linear de-risking’ strategy (USS, Dec 2017, p. 4). The first JEP proposal above is to return to the ‘deferred de-risking’ strategy.

9. In the light of the revised investment strategies for the period from now to year 20 and for the long-term period from year 20 onwards, use the basic method described above to calculate the discount rate, and so a technical provisions figure, technical provisions deficit (if any), and required deficit recovery and future service contributions at the valuation date.

Since the expected long-term rate of return is now lower than it would have been under the ‘maintenance’ investment strategy, the revised discount rate will be lower and all the other figures will be higher than they would have been under that strategy. For example, Marsh has calculated that with a ‘maintenance’ strategy the £7.5bn technical provisions deficit that USS announced in the November 2017 valuation would have been only £0.4bn (Marsh, 2 Sep 2018). From figures in the new consultation document he has estimated that a ‘maintenance’ strategy would similarly replace the ‘default’ £3.6bn technical provisions deficit USS have calculated for the 2018 valuation by a surplus of perhaps £4bn (personal communication).

Furthermore the ‘de-risking’ strategy does not only raise future service contributions for the present valuation cycle. If the portfolio is progressively de-risked over 20 years then, other things being equal, the discount rate at each three-yearly valuation will be lower than at the previous one, until by year 20 it is low enough to give a figure for TP20 that is within £10bn of SS20. Therefore the level of future service contributions will have to be higher at each valuation. So de-risking forces contributions to rise continually from now to year 20.

Suppose, by contrast to the way we have described it so far, Test 1 is used as a ‘target-setting’ procedure. Here the steps are the same except that we do not bother with steps 1, 2 and 5. Instead we take it for granted from the start that Test 1 will dictate the scheme’s long-term investment strategy. Accordingly we move directly to predict SS20, and then to calculate in turn: the required TP20, the required discount rate at year 20 to produce that TP20 figure, the required 67%-prudent expected rate of return on the portfolio mix from year 20 onwards, the investment strategy for the period from now to year 20 to reach that 67%-prudent expected rate of return by year 20, a discount rate at the valuation date arising from the combination of the above expected rates of return, and so a technical provisions figure and levels of contribution at the valuation date. In our view this is how Test 1 really functions. In reality Test 1 (unlike Tests 2 and 3) is not a test of the status quo that informs whether change is required, but a process for implementing change from the start.

In short, whether it is used as a ‘stop-and-check’ or as a ‘target-setting’ procedure, Test 1 ends up by dictating a policy of ‘de-risking’ over the coming 20 years. When USS first announced a 20-year ‘de-risking’ plan in their 2014 valuation consultation document, they referred only obliquely to Test 1 (USS, Oct 2014, pp. 11–12). But by 2017 they felt able to acknowledge quite openly that the Test was driving this plan:

In order to remain within the bounds of Test 1 (the reliance on the sector should not be greater than the value of the contingent contributions over a period), it is necessary to reduce investment risk over time. (USS, Feb 2017, p. 31)

6. Test 1 as unfit for purpose

Since 2011, USS’s ‘de-risking’ strategy has led to a succession of rising technical provisions deficits, increases in contributions and cuts in benefits which, combined with a fall in bond yields from 2014 to 2017 and some dubious behaviour by UUK, culminated in USS’s November 2017 valuation document determining that contributions would need to rise to 37.4% to maintain existing benefits (USS, Dec 2017). This led UUK to force through a decision to close the DB scheme — an outcome that only the largest strike in British HE history has been able to prevent. So the most obvious criticism to make of Test 1 is that it provides an explicit mechanism to drive the destructive strategy of ‘de-risking’.

Since its launch Test 1 has been subjected to a series of more detailed criticisms, in particular by First Actuarial (First Actuarial, Nov 2014; Salt and Benstead, Apr 2017; Salt and Benstead, Sep 2017). Here we shall not rehearse these but instead focus on a specific criticism, which we believe is unanswerable. This criticism is an elaboration of the point made above, that there is a ‘mismatch’ between the underlying purpose of the Test and its design, into an argument that the Test is unfit for its own purpose. It has been made by Marsh and Otsuka, and given particular urgency by calculations done by Marsh in October 2018. The criticism runs as follows.

The underlying purpose of Test 1 is to ensure that in future the self-sufficiency deficit does not exceed the employers’ ability, in case of need, to bridge it, or in other words that it does not exceed the scheme’s maximum reliance, that:

SS – A ≤ maximum reliance

In our view this purpose is appropriate only in the case of a scheme being made ready for closure. Therefore in the context of the USS scheme we doubt its relevance. However, suppose we take it as given.

If this is the underlying purpose, and if year 20 is to serve as touchstone, then the requirement USS should set in order to achieve that underlying purpose is simply that the self-sufficiency deficit at year 20 be no greater than the maximum reliance, or, if we take maximum reliance to be £10bn, that:

SS20 – A20 ≤ £10bn

By contrast the requirement set by Test 1 is that the scheme’s predicted self-sufficiency technical provisions gap at year 20 be no greater than the maximum reliance, that:

SS20 – TP20 ≤ £10bn

But at any given date the self-sufficiency technical provisions gap is only contingently related to the self-sufficiency deficit. The former depends on the technical provisions figure which in turn is determined entirely by the scheme’s accrued liabilities and planned investment strategy. The latter instead depends on the value of the scheme’s assets. Therefore Test 1 is unfit for its own underlying purpose. Relative to that purpose, it sets the wrong requirement.

In particular it is quite possible to imagine circumstances in which Test 1 gives a ‘false alarm’, that is, in which an investment strategy fails the Test when, given its underlying purpose, that strategy ought to pass it (Otsuka, 13 Oct 2018; Marsh, 15 Oct 2018).

Furthermore using Test 1 has a side effect, for it drives an investment strategy of ‘de-risking’ the scheme over the next 20 years. This strategy reduces the rate of return on the scheme’s investments between now and year 20 and so makes it harder for the scheme to get within maximum reliance of self-sufficiency in the future. So by setting the requirement that it does Test 1 actively undermines its own underlying purpose.

USS has replied to this criticism by saying that, given their legal obligations, by year 20 they will have had to eliminate any technical provisions deficit or surplus, so that we can assume that A20 = TP20. Therefore by requiring that predicted TP20 be within £10bn of SS20 (call this ‘the technical provisions requirement’) the Test automatically requires that A20 be within £10bn of SS20 (call this ‘the assets requirement’). So adopting the former requirement is the means of ensuring that the latter is met. In turn, ensuring the latter is met is a way to achieve the Test’s underlying purpose of making sure that in future the self-sufficiency deficit is no greater than the maximum reliance. Therefore the Test does after all fulfil that underlying purpose. In November they wrote:

Under legislation, assets equalling liabilities beyond the recovery plan period isn’t an erroneous assumption, it’s a given: we know that we will have to adjust our assets (via contributions and investment strategy at every valuation) to equal the Technical Provisions liabilities by that point. Further, the trustee’s Statement of Funding Principles, which are agreed with participating employers, clearly states that the trustee’s aim is to hold assets equal to the Technical Provisions.
Test 1 was designed with this in mind, and consciously sets the Technical Provisions liabilities at 2037 as self-sufficiency less target reliance — not in error but because, as above, the assets will be constructed to equal the Technical Provisions at this time. (Coughlan, 28 Nov 2018; cf USS, Nov 2016, p.5; Sep 2017, pp. 40–41)

However, Otsuka has argued that it is quite incorrect to say that USS are legally required to ensure that at year A20 is equal to TP20. The Pensions Act 2004 states only that ‘Every scheme is subject to a requirement (“the statutory funding objective”) that it must have sufficient and appropriate assets to cover its technical provisions’ (UK Gov, 2004, sec. 222, our emphasis). So it only requires that USS ensure that A20 is at least equal to TP20. In this light it would be arbitrary for USS to stipulate that they must be just equal. But if A20 can exceed TP20 then it is quite possible for A20 to be within £10bn of SS20 while TP20 is not. The statutory funding objective ensures that the ‘technical provisions requirement’ entails the ‘assets requirement’, but not vice versa. So it does not prevent Test 1 from giving ‘false alarms’ relative to its underlying purpose. (Otsuka, 12 Dec 2018).

Instead, the criticism continues, the appropriate way to realise the underlying purpose of Test 1 would be to adopt the ‘assets requirement’, that:

SS20 – A20 ≤ £10bn

Let us call the operationalisation of this requirement using a 67% prudence-adjusted prediction for A20 ‘Test 1a’. Marsh’s October calculations gave force to the present criticism because they showed, using 2017 data supplied by USS, that a ‘maintenance’ (‘no de-risking’) investment strategy applied to the scheme from now indefinitely into the future, with contributions at existing levels, fails Test 1 but passes the more appropriate Test 1a. So this is exactly a case of Test 1 giving a ‘false alarm’ relative to its underlying purpose. Marsh’s figures, which were substantially confirmed by USS, are:

Because of the effects of compound interest, the ‘maintenance’ investment strategy leads to a substantial technical provisions surplus by year 20. So it easily passes Test 1a (the self-sufficiency deficit of £2.8bn is much less than £10bn) but easily fails Test 1 (the self-sufficiency technical provisions gap of £21.8bn is much more than £10bn). Compared to how things would have been with a ‘de-risking’ strategy, the value of assets is high because the fund has been invested in a higher yielding portfolio, while the technical provisions figure is low because the discount rate is high as a result of the plan to remain in this portfolio after year 20 (Marsh, 12 Oct 2018).

It follows that if the more appropriate Test 1a were substituted for Test 1 there would be no need for USS to adopt a strategy of ‘de-risking’ at all.

Although Marsh’s calculations were not available until after the JEP had reported, he made the ‘unfit for purpose’ criticism in his submission to the JEP (Marsh, Jul 2018). The JEP report effectively concurred with the view that Test 1 is badly aligned with its own underlying purpose and drives an investment strategy that does not follow from that purpose:

[W]e believe that Test 1 is given too much weight in determining the valuation and its effects extend beyond its original purpose. Rather than being used as a ‘stop-and-check’ reference point, Test 1 is being used as a constraint on benefit design and driver of investment strategy. The Panel does not consider this helpful. It would be far better if Test 1, were its use to continue, was used as a test that informed other aspects of the valuation and funding strategy rather than acting as its lynchpin. (JEP, 13 Sep 2018, p.8)
Whilst a test of self-sufficiency and employer reliance is a useful principle for the basis of a valuation, Test 1’s formulation, application and implementation is very rigid with the result that Test 1 has led to a valuation that is model-driven rather than model-informed. (p. 24)

The same view is implicit in two paragraphs of a letter from the Director of Supervision at TPR to USS in December:

35. Although some of the details of Test 1 received criticism in the JEP report, we continue to believe that the Trustee should consider the reliance of the Scheme on the sector within their overall valuation approach. Considering a scheme’s overall reliance on its covenant now and in the future is a key part of how trustees should assess the sponsoring covenant. Therefore, we continue to believe that the intent behind Test 1 is relevant and that it should form part of your approach to assessing the Scheme’s funding position. It is only one of a number of factors to consider, so the Trustee should attach the appropriate weight to it.
36. We expect effective monitoring to be supported with pre-agreed actions should the level of reliance of the Scheme on the sector exceed the Trustee’s tolerance levels. We support the retention of the intent of this Test within the Trustees ongoing monitoring approach. (Birch, 11 Dec 2018)

TPR write that they ‘continue to believe that the intent behind Test 1 is relevant and that it should form part of your approach’ but that ‘the Trustee should attach the appropriate weight to it’ (our emphases). The implication is clear: although TPR support the underlying purpose of Test 1 they do not consider that it fulfils that purpose well. Meanwhile they believe Test 1 is being given too much weight in the valuation process, which can only mean too much weight in determining the investment strategy. Thus in these paragraphs TPR tacitly endorse the ‘not fit for purpose’ critique of Test 1, although unsurprisingly TPR are supporters of de-risking as such (Cruickshank, Sep 2017, para. 14b).

7. Test 1’s second purpose

If the underlying purpose of Test 1 is to ensure that in future the self-sufficiency deficit does not exceed the employers’ ability, in case of need, to bridge it, and if Test 1 is unfit for this purpose, why did USS adopt it rather than a more appropriate test like Test 1a?

Marsh and Otsuka have suggested that USS simply failed to notice that A20 might exceed TP20 (Otsuka, 13 Oct 2018; Marsh, 15 Oct 2018). However we believe that a better explanation is that USS designed Test 1 with a second underlying purpose: a purpose that is undeclared, or at least much less openly declared than the first. This purpose is that of bringing the scheme in future into conformity with what we shall call the ‘covenant-risk principle’: the principle that the investment risk taken by the scheme should be proportionate to its covenant strength. This principle was first enunciated, to our knowledge, by TPR in their 2009 statement ‘Scheme funding and the employer covenant’ (TPR, Jun 2009, pp. 2–3). It is clearly stated in their 2014 Code of Practice for DB Schemes:

[A]s a general approach, if trustees accept investment risk, the higher the level of investment risk held within the scheme the greater should be the trustees’ focus on the level of employer covenant available to support it. (TPR, Jul 2014, p. 30)

USS’s ‘An integrated approach to scheme funding’ document, which was published at almost the same time, spells out the principle clearly:

A strong employer covenant typically means employers can tolerate more risk in the scheme as they are believed to have the ability to make up any shortfall if investments do not deliver the anticipated returns. Employers with weaker covenants are unable to carry significant levels of risk as they are considered to be unable to provide additional support if the investments do not deliver the anticipated returns. (USS, Jul 2014, p. 6)

It states almost as clearly that USS consider that the present level of investment risk in the scheme is too high relative to the covenant strength and that accordingly they plan to reduce its investment risk, that is, to ‘de-risk’ the fund:

In broad terms the trustee believes the amount of risk taken should be proportionate to the amount of support available to the scheme from the employers, and specifically that there should be no increase in the reliance placed on the covenant over time. Indeed, it is the trustee’s view that with the right economic conditions, opportunities should be taken to reduce the amount of risk within the scheme and therefore the reliance on the covenant. (p. 4)

Here ‘reliance’ can only be meant in our secondary sense, for otherwise the intertwinement in this passage of ‘reliance’ and ‘risk within the scheme’ (which can only mean investment risk) would make no sense. Reducing the investment risk in the scheme’s portfolio does not reduce its reliance in the primary sense (the self-sufficiency deficit). Rather by reducing the expected rate of return on the scheme’s investments it tends to increase it. But it does reduce its reliance in the secondary sense (the self-sufficiency technical provisions gap), in that it reduces the expected rate of return on investments and so the discount rate (as calculated by the 67%-prudent expected rate of return method) and so raises the technical provisions figure up towards the self-sufficiency value of liabilities. Conversely requiring that the self-sufficiency technical provisions gap be reduced is a way of requiring that the scheme’s investment risk be reduced: that it be ‘de-risked’.

In this light we can see that USS likely adopted Test 1 not only as a means of ensuring that in future employers are always able, in case of need, to bridge the self-sufficiency deficit, even if this was the purpose they declared in unveiling the Test. They also adopted it so as to bring the scheme in future into conformity with the covenant-risk principle. The self-sufficiency technical provisions gap serves as a rough measure of investment risk, and maximum reliance as a rough measure of covenant strength. So Test 1, by requiring that the predicted self-sufficiency technical provisions gap at year 20 be no greater than maximum reliance, is a crude way of yoking investment risk to covenant strength.

Indeed we can even argue that this second purpose must have taken precedence over the first in designing Test 1. For it looks as if USS’s way of combining the ‘technical provisions requirement’ of Test 1, that

SS20 – TP20 ≤ £10bn

with what they take to be the mutually entailed ‘assets requirement’ that

SS20 – A20 ≤ £10bn

has been first to design an investment strategy (both beyond year 20 and up to year 20) that satisfies the first requirement and then to set themselves to ensuring over the intervening period that the value of assets is increased sufficiently so that when the time comes the second requirement will also be satisfied, by raising contributions to whatever level is needed over that period within the constraints of this investment strategy.

We can set the two underlying purposes of Test 1 alongside each other by stating both in terms of covenant strength. The first is to ensure that in future the scheme’s self-sufficiency deficit is proportionate to covenant strength (as measured by maximum reliance). The second is to ensure that in future its investment risk is proportionate to covenant strength.

Since USS has used ‘reliance’ to mean sometimes the self-sufficiency deficit and sometimes the self-sufficiency technical provisions gap, and since they use the self-sufficiency technical provisions gap as an indicator of investment risk, both these purposes can be expressed vaguely in terms of ‘ensuring reliance is proportionate to covenant strength’. This makes it easy to confuse them. However they are quite distinct, and each can be realised without the other. The first purpose refers to the value of the scheme’s assets, whereas the second refers to its portfolio mix. The double purpose of Test 1 is to ensure that by year 20 the fund is both quantitatively and qualitatively similar to the ideal of a self-sufficient fund, with deviations from this ideal allowable only in proportion to covenant strength.

It might be thought that if Test 1 is unfit for its first purpose perhaps at least it is fit for its second one. However we shall argue that is not. We doubt the relevance of this second purpose in the context of the USS scheme even more than that of the first. However again let us take it as given.

A first reason why Test 1 is unfit for this second purpose is that when it is examined carefully, conformity to the letter of the Test does not in fact require ‘a de-risking strategy’. In our reconstruction of the steps of the Test, step 6 derived a discount rate that corresponded to the required figure for TP20. Step 7 then was:

7. Work out what 67%-prudent expected rate of return on the portfolio for the long-term period following year 20 would be required to give that discount rate using the USS’s basic funding method. (our emphasis)

But the italicised words presuppose that adopting a given discount rate at year 20 requires adopting a portfolio mix from year 20 onwards whose 67%-prudent expected rate of return on assets exactly equals that discount rate, and this is not the case. Adopting a portfolio mix with a certain 67%-prudent expected rate of return sets an upper limit to the discount rate, given USS’s method for working out discount rates, but it does not prevent USS from choosing a discount rate lower than this, and there is nothing in law against their doing so. Setting it lower just adds in more prudence. Normally USS would keep the discount rate just at the 67% point so as not to charge excessive contributions, but they do not have to. As Salt and Benstead say:

It is not the case that the actual investment strategy constrains the trustee from setting a discount rate assuming a more cautious investment strategy than the actual strategy in place. The USS is free to lower the discount rate if it wishes but this says nothing about the desirable investment strategy. (Salt and Benstead, 16 Nov 2018, p. 4)

This means that requiring that TP20 be high enough so that the gap from it to SS20 is less than the maximum reliance, and thereby that the discount rate be low enough to be within the corresponding distance of the self-sufficiency discount rate, does not require that the 67%-prudent expected rate of return on the planned portfolio mix be equally low.

Thus pushing TP20 up so as to meet Test 1 pushes the discount rate at year 20 down, but this does not drag the 67%-prudent expected rate of return from year 20 onwards down with it, and so does not prevent the investment strategy from year 20 onwards from being a return-seeking one. Pushing up TP20 so as to push down the 67%-prudent rate of return on the portfolio from year 20 onwards is like pulling on a snapped rope. Since the letter of Test 1 does not require a portfolio mix after year 20 with a 67%-prudent rate of return that is down to any particular figure, it does not require a ‘de-risking’ investment strategy so as to reach such a portfolio mix between now and year 20.

However let us suppose that USS add a hidden clause to Test 1, stating that the discount rate used to calculate TP20 must be neither higher nor lower than the 67%-prudent expected rate of return on the portfolio planned for year 20 onwards. There remains a deeper reason why Test 1 is unfit for its second purpose, as follows.

Even if we agree with USS that the self-sufficiency technical provisions gap is a good measure of investment risk and that maximum reliance is a good measure of covenant strength, this tells us nothing about the ratio that these two measures must bear to each other in order for the investment risk to count as ‘proportionate’ to covenant strength. USS say that the ratio must be less than 1:1, but this ratio is derived from the first purpose of Test 1 and is only relevant in so far as we insist that the very same Test must simultaneously realise both purposes. Once we drop that insistence, the 1:1 ratio becomes arbitrary. USS might as well say that to ensure proportionality of investment risk to covenant strength the ratio of the self-sufficiency technical provisions gap to maximum reliance must be less than 2:1, or less than 1:3. The only thing to underpin the choice of a 1:1 ratio is the judgment of USS’s managers that it is appropriate.

In short, Test 1 emerges as a procedure designed to kill two birds with one stone, but that succeeds in killing neither.

8. Test 1 in the January consultation document

There is no doubt that Test 1 continues to play a central role in calculating the technical provisions figure and the level of future service contributions in the new consultation document. However in this document it has become ‘self-effacing’.

First, the term ‘Test 1’ does not appear once in this document, whereas by contrast it appears 19 times in the September 2017 consultation document (USS, Sep 2017). This suggests that USS are aware how futile it has become to mount a defence of its use.

Second, instead of presenting the ‘technical provisions requirement’ of Test 1 (that predicted TP20 be within £10bn of SS20) as the means of ensuring that the ‘assets requirement’ (that A20 be within £10bn of SS20) is met, USS now present the latter requirement by itself, without any reference to Test 1. They formulate it as a ‘target’:

One of the ways with which the Trustee manages long-term risk is by setting a long-term target level of reliance that is within the risk capacity and the risk appetite of the employers.
In this context the Trustee measures reliance as the difference between:
• The assets required under a low-risk investment strategy to pay all accrued benefits with a high probability without requiring any further contributions (i.e. self-sufficiency); and
• The assets currently held by the DB section of the scheme.
The long-term reliance target adopted for the 2017 valuation was £10bn in real terms in 20 years’ time. In other words, this is the target level of reliance on the employers’ covenant 20 years out from the valuation date. (USS, 2 Jan 2019, pp. 11–12)

The accompanying graph clearly illustrates a plan to achieve the ‘target’ that A20 be within £10bn of SS20.

But this means that there is an incoherence in the consultation document. On the one hand its calculations continue to use Test 1’s ‘technical provisions requirement’, that predicted TP20 be within £10bn of SS20, so as to require and quantify its ‘de-risking’ investment strategy and thereby to calculate the technical provisions figure and levels of contributions. But on the other hand the document does not mention Test 1. It only affirms the ‘assets requirement’, that A20 be within £10bn of SS20, but without referring to the Test which USS had previously presented as the means of ensuring this requirement is met.

When the ‘assets requirement’ is unmoored in this way from Test 1 it invites the obvious question of why any ‘de-risking’ strategy should be needed to meet it. As Marsh’s calculations show, to meet this requirement it would make sense instead to adopt a ‘maintenance’ investment strategy, with its higher rate of return. USS might claim that the statutory funding obligation obliges them to ensure that A20 is just equal to TP20, so that by adopting the ‘assets requirement’ that A20 be within £10bn of SS20 they are necessarily simultaneously adopting the requirement that TP20 be just as high. But as we have shown above they would be on shaky ground if they did so. In the absence of this claim, adopting the requirement that A20 be within £10bn of SS20 does not justify a ‘de-risking’ strategy. Yet at the same time such a strategy continues to be generated by the use of Test 1 in the calculations behind the scenes.

Meanwhile, the burden of justifying ‘de-risking’ is shifted to pointing to the volatility of the self-sufficiency deficit over the last 20 months, and predicting how various possible ‘shocks’ in the next three years might drive it up: a failure of gilt yields to revert to pre-2014 levels as USS have predicted they will, a 10% fall in asset prices, a 0.5% fall in gilt yields and so on (USS, Jan 2019, pp. 12, 16). It is possible that a sober and quantified argument from one of the two underlying purposes of Test 1, together with the possibility of such events, could be made for ‘de-risking’. But vaguely alarming statements such as the claim that

there are credible scenarios that could make the current risk position difficult to recover from — such that the ability to move to a self-sufficiency strategy in the long term moves out of reach (p.12)

cannot substitute for such an argument. And certainly the general spectre of ‘short term reliance risk’ cannot substitute for Test 1 as a means of mandating a 20-year ‘de-risking’ strategy with a specific endpoint.

9. Dropping Test 1

In this USSbrief we have outlined the design of Test 1 and the way it functions to drive a ‘de-risking’ investment strategy that has pushed the USS Defined Benefits scheme almost to extinction. We have criticised the Test as unfit not only for its first, declared underlying purpose, but also for its second, undeclared one.

In our view both of these purposes are anyway based on a false premise. Schemes close when their sponsoring companies go bust, and the whole UK university sector is not going to go bust. Schemes also close because employers choose to no longer support them. But in this case why should members pay the costs? The ‘de-risking’ driven by Test 1 progressively realises in advance some of the costs of closure, and employers, through the JNC and now through rule 76 cost sharing, have been complicit in passing a significant portion of these costs on to members. That said, it is a severe indictment of a Test that it is incompetent to accomplish even the unjustified purposes for which it was designed.

In the four years since the launch of Test 1, USS has consistently ignored the principal criticisms made of it. Meanwhile UUK have criticised it only in the mildest way. Throughout 2017 UUK supported USS’s valuation method. By drumming up unspecified fears of ‘risk’ they precipitated USS’s return to linear ‘de-risking’ in the deeply flawed November 2017 valuation, and then tried to force members to accept a DC scheme at great cost to their retirement benefits. It took the greatest industrial action in the history of the sector to force employers to reconsider their approach and agree to appoint an independent expert panel to review the 2017 valuation.

The JEP report concurred with the critics: Test 1 was flawed and its use in determining the scheme’s investment strategy was unwarranted. TPR have indicated that they too accept these criticisms. Now USS, bizarrely, has published a consultation document which insists on pursuing the ‘de-risking’ investment strategy mandated by Test 1, and consequently on proposing contribution increases almost as steep as those of the 2017 valuation, while disavowing Test 1 in name and presenting it in a way that per se does not support this strategy.

There is a massive crisis of confidence in USS. Many members simply do not trust that USS and employers are managing the scheme in their best interests. Recovering that trust will be difficult. It is not simply a matter of improving communications, though these could undoubtedly be better. It will require positive actions that address the valid concerns at the heart of the crisis. Employers took several steps in the right direction by participating in the JEP and then supporting its recommendations. But USS do not seem to be listening. It is clear to us that the finances of the scheme are sound and that the technical provisions deficit is the product of the unwarranted use of Test 1 in the funding method. Test 1 itself is intellectually in tatters. Even TPR have lost confidence in it. Continuing to use it will progressively drive up contributions so as to finance the misinvestment of billions of pounds of members’ money over the next two decades. Yet USS continue to use it. Meanwhile if USS have an alternative systematic argument for their 20-year de-risking strategy then we have yet to see it. There is no possibility of restoring member confidence in the scheme and its managers while they persist in a course that so clearly flies in the face of reason and justice.

What then should be done? In our view, the answer is obvious: Test 1 should be dropped immediately. As a matter of practical politics, that may not be feasible. But at a minimum USS should accept the two recommendations of the JEP mitigating the effects of Test 1 for the 2018 valuation, without simply imposing compensatory penalties. If USS will not take this step then employers must use the current consultation to force them to.


The authors are grateful to Sam Marsh and Michael Otsuka for comments on an earlier draft. All errors remain our own. Some minor corrections to the text were made on 19 February 2019.

This is a USSbrief that belongs to the OpenUPP (Open USS Pension Panel) series. This paper represents the views of the authors only. The authors believe 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 #USSbriefs68 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.