JEP2’s dual discount rate proposal for USS: an analysis and evaluation
Number 89: #USSbriefs89
Click for printable PDF (available soon)
This is a USSbrief, published on 15 January 2020, that belongs to the OpenUPP (Open USS Pension Panel) series. [Update 18 January 2020: The author has responded to comments on his brief here.]
The remit of the second report of the Joint Expert Panel (JEP) was to make proposals for the governance of the Universities Superannuation Scheme (USS) and for its future valuation methodology or 'funding method' (JEP, 2019a). Scepticism about the valuation methodology, and the deficit and cost of future service it produced, was at the heart of the intellectual case for the 2018 strikes which led to the creation of the JEP. Accordingly, we should see its proposal on valuation methodology as at the heart of the second report, which was published in December 2019 (JEP, 2019b).
In essence the report’s proposal is to replace USS’s present method for valuing the scheme’s liabilities, which is entangled with its notorious Test 1, with a new ‘dual discount rate’ method (see ‘Alternative path 3’, pp. 64–70 of the report). This is to allow USS’s liabilities to be valued in a more optimistic way and its funds invested more productively, while still addressing the issue of pension security which underlies Test 1.
This USSbrief aims to spell out a proper understanding of the proposal and the investment strategy associated with it. Section 1 summarises USS’s present valuation methodology and Test 1 and explains why it is so problematic. Section 2 explains the basics of the ‘dual discount rate’ proposal. Section 3 argues that the proposal implies a parallel ‘dual portfolio’ investment strategy for the scheme. Section 4 argues that the proposal should be understood as implying separate discount rates for past service and future service liabilities. Section 5 looks at the effect on contribution rates of applying the proposal to the 2018 valuation. Section 6 looks at the advantages and disadvantages of the proposal. Section 7 looks at the advantages and disadvantages of the ‘dual portfolio’ investment strategy implied by it. Finally section 8 proposes two modifications to the proposal. The conclusion is that the proposal and the investment strategy it implies represent a significant improvement on the USS status quo, but that they need to be modified in these two ways to provide a positive way forward for the scheme.
1. USS’s present valuation methodology and Test 1
We first need a brief summary of the valuation methodology the proposal is intended to replace. (For a fuller summary see #USSbriefs68 Chitty and Wilson, 2019a, sec. 1). By law, at every three-yearly valuation USS has to choose a prudent ‘discount rate’ at which to discount its accrued liabilities (the pension promises it has made to date), and use this to calculate a present value for those liabilities. This value is called the scheme’s Technical Provisions. If on the valuation date the value of the scheme’s assets is less than its Technical Provisions then the scheme is in ‘deficit’; if it is more then the scheme is in ‘surplus’. If the scheme is in deficit USS has to produce a plan to eliminate the deficit over a certain number of years; if it is in surplus then it may similarly eliminate the surplus, though it is not legally obliged to.
USS must then do another calculation, typically using the same discount rate, to work out what annual level of contributions in the years following the valuation is needed to cover the additional scheme liabilities resulting from members’ service in these years. This is the level of Future Service Contributions. The total annual level of contributions to be paid by employers and members in the years following the valuation is then the level of Future Service Contributions plus whatever annual figure needs to be added to it to progressively eliminate the deficit (Deficit Recovery Contributions), or whatever USS decides to subtract from it to progressively eliminate the surplus (Surplus Elimination Adjustment).
Historically USS has set the discount rate at a prudently expected rate of return on the assets it plans to hold from the valuation date into the future: in the last few years specifically at the rate of return which it calculates there is a 67% probability of the assets equalling or exceeding. Therefore the discount rate depends heavily on USS’s investment strategy: the more heavily it plans to invest its fund in return-seeking assets (equities and property) as opposed to low-risk (bond-like) assets, the higher will be the 67%-probable expected rate of return on the scheme’s assets, and so the discount rate.
Up to 2007, about 90% of the scheme’s fund was invested in return-seeking assets and about 10% in bond-like assets. However, since then USS has progressively ‘de-risked’ the fund, so that the proportions now are about 45% and 35%, with the remainder mainly in ‘private markets’ (Chitty, 2020).
In 2014, USS formalised this de-risking strategy by introducing its ‘Test 1’. The details are complicated, but the effect of the Test is to set a double target for the scheme. First, by year 20 from the valuation date the value of the scheme’s assets must be no more than a certain distance (currently set at £10bn) below the value of the scheme’s liabilities calculated ‘on a self-sufficiency basis’; and second, by that time the fund must be invested in a portfolio that is correspondingly close to a ‘self-sufficiency’ portfolio.
To explain: the value of the scheme’s liabilities on a given date calculated ‘on a self-sufficiency basis’ means the sum of money on that date which, if it were invested from then onwards in a ‘self-sufficiency’ portfolio, would have a 95% probability of meeting all the liabilities accrued by the scheme up to that date, and a ‘self-sufficiency portfolio’ means one consisting mainly of bond-like assets.
The broad idea is that once these two targets have been met the pensions of members will be protected even if the scheme is closed, so that there are no more new contributions, and goes into ‘run-off’.
In order to meet the second target of Test 1, USS argues that it must continue to ‘de-risk’ the fund over the next 20 years. Other things being equal, this will progressively reduce the scheme’s discount rate and drive up the Technical Provisions, the deficit (if there is one), and the level of Future Service Contributions.
Since 2014, Test 1 and its associated de-risking strategy have been extensively criticised. Critics have argued that targets like those set by Test 1 are only appropriate if there is an expectation that the scheme may have to close in 20 years. Or else they have argued that even if the first target set by Test 1 is appropriate, the second is not, and that the best way to meet the first target is by maintaining the present portfolio mix rather than de-risking the fund over the next 20 years. Either way, it is argued, Test 1 does not justify the de-risking strategy. (For fuller accounts see #USSbriefs32 Marsh, 2018a; #USSbriefs58 Marsh, 2018b; Otsuka, 2018; #USSbriefs68 Chitty and Wilson 2019a, secs. 3–5.)
On the other side, USS have argued that pressures from The Pensions Regulator (TPR) obligate it to have something like Test 1 in place to ensure that the scheme will be able to meet its liabilities into the future. However TPR’s announcements on this matter are not clear cut. It has stated that it expects DB schemes to develop a ‘journey plan’ towards being ‘well-funded on a low-risk investment strategy, which places minimal reliance on the employer covenant, in the future’, which would correspond to at least the first target set by Test 1 (TPR, 2019b, p. 47, see also Birch, 2019, sec. 18). But it has also suggested that it is willing to consider ‘a range of solutions’ for open schemes such as USS (Fairs, 2019). This suggestion is in line with the Government’s 2018 white paper on DB schemes, which says that for open schemes the ‘Long Term Objective’ need not be to ‘reach self-sufficiency with low-risk investment strategy and [then] run-off with minimal call on the sponsor employer’, but could instead simply be to ‘run-on with employer support’ (DWP, 2018, sec. 99, quoted in the JEP report, p. 57). While the first of these implies something like Test 1, the second does not.
The JEP report summarises a number of criticisms of Test 1 from respondents (pp. 58–59). Its own conclusion is crushing:
Within the context of an open scheme with strong covenant the Panel could not identify any circumstances in which it believed that Test 1 should be the primary driver of funding and investment strategy. (p. 55)
This sets the scene for the report’s own proposal for a valuation method which aims to avoid the need for automatic de-risking while still ensuring that the scheme will always be able to meet its pension promises: the ‘dual discount rate’ proposal.
2. The dual discount rate proposal
The proposal works like this. Instead of valuing all the scheme’s liabilities on the assumption of a certain investment strategy for the whole fund, USS is to value the liabilities owed to each member of the scheme (the member’s ‘accrued benefits’, or the pension payments that have been promised to the member to date) separately. To value the liabilities owed to a given member, imagine that the money to cover those liabilities is to be invested in a return-seeking portfolio from now until the member retires, and then shifted into a low-risk portfolio for the period of their retirement, and work out how much money would prudently be needed to cover the liabilities on this basis. Thus for the member’s pre-retirement years adopt a discount rate based on the prudently expected rate of return from a return-seeking portfolio, and then for the member’s post-retirement years adopt a discount rate based on the prudently expected rate of return from a low-risk portfolio.
We can illustrate the proposal by using three examples given in the report (p. 65). See Figure 1.
It is assumed that on average a member retires at 66 and has 18 years of retirement.
Member 1 is 25. It is assumed that, starting from the valuation date, they will work for a further 41 years and then be retired for 18 years. The liabilities owed to this member are valued by adopting one (higher) discount rate for a return-seeking portfolio and another (lower) one for a low-risk portfolio, and assuming that the money to cover the liabilities is to be invested in a return-seeking portfolio for the next 41 years and then in a low-risk portfolio while it is gradually drawn down over the following 18 years. This mirrors the standard investment strategy for a stand-alone Defined Contribution (DC) scheme, except that the transition from a return-seeking to a low-risk portfolio would normally be carried out gradually in the years before retirement.
Member 2 is 60. It is assumed that they will work for a further 6 years and then be retired for 18 years. The liabilities owed to this member are valued by assuming that the money to cover them is to be invested in a return-seeking portfolio for the next 6 years and then in a low-risk portfolio while it is gradually drawn down over the following 18 years.
Member 3 is 69 and has been retired for 3 years. The outstanding liabilities owed to this member are valued by assuming that the money to cover them is to be invested in a low-risk portfolio from now on while it is drawn down over the next 15 years.
Finally, the values (as calculated in the above way) of the liabilities owed to all the members of the scheme are added up to get a value for the scheme’s total liabilities, i.e. a figure for its Technical Provisions.
From this, it is possible to calculate what single discount rate would have given the same Technical Provisions figure if it had been applied to the scheme’s total liabilities directly. This discount rate is the Single Effective Discount Rate (SEDR) for the valuation. It is important to see that the SEDR is calculated from the Technical Provisions rather than the other way round. Its purpose is simply to provide a measure that can be used to compare the valuation method with other methods that use a single discount rate.
In response to a request from the JEP, USS have calculated that, if applied to the 2018 valuation, the dual discount rate proposal would have led to about 33% of the total value of the scheme’s liabilities being based on the higher, pre-retirement, discount rate, and 67% on the lower, post-retirement, rate (p. 116). As each member goes through their career, in successive valuations a smaller and smaller proportion of the liabilities owed to that member will be valued at the higher, pre-retirement, discount rate. However (other things being equal), as long as new junior members continue to join the scheme at roughly the same rate as existing members retire, the proportion of the scheme’s liabilities as a whole that is valued at the pre-retirement discount rate will remain stable over time, and therefore so will the SEDR. If new junior members join at a faster rate than existing members retire, then over time the proportion of the liabilities valued at the higher, pre-retirement, discount rate will increase, and the SEDR will gradually rise towards that higher rate. Conversely, if new junior members join at a slower rate than existing members retire, then over time that proportion will decrease, and the SEDR will gradually fall towards the lower post-retirement discount rate.
3. Does the proposal entail a ‘dual portfolio’ investment strategy?
In proposing this method of valuing the scheme’s liabilities, the report tacitly assumes that the scheme’s investment strategy will follow the same pattern as the valuation of the liabilities. That is, it assumes that, if the scheme’s fund is notionally divided between its members in the same proportion as the values of the liabilities owed to each, then the investment strategy for the part of the fund associated with each member will correspond to the method of discounting that member’s liabilities. For example, the investment strategy for the part of the fund associated with Member 1 above will be to invest it in a return-seeking portfolio for the next 41 years and then in a low-risk portfolio for the following 18 years as it is drawn down. We can call this a ‘dual portfolio’ investment strategy. If USS adopts this investment strategy, and if it keeps its method of setting the discount rate on the basis of the prudently expected rate of return on the assets that it plans to hold in future, then the report’s proposal for valuing the scheme’s liabilities follows quite naturally. By contrast if it adopts some other investment strategy then this proposal begins to look arbitrary.
In fact in several places the report explicitly suggests that it recommends a ‘dual portfolio’ investment strategy to parallel the dual discount rate method of valuing the liabilities. For example, at one point it says that under the proposed valuation method:
• the post-retirement years are valued and secured against a low-risk portfolio that will have a very high probability of being able to pay out with no further call on employers while;
• the benefits of those further from retirement are backed by a valuation method that allows for returns over the longer term that a higher risk portfolio can generate. (p. 65, emphasis added)
It is true that in one passage the report states that the scheme’s investment strategy need not follow the same pattern as the discount rate calculation:
Having determined the discount rates, it would then be a separate decision as to how Scheme assets are actually invested. An investment strategy which matches the assumed discount rates would reduce expected valuation volatility, but there may be better ways of guarding against this outcome or minimising its impact while taking advantage of the USS team’s ability to deliver good investment returns. (p. 65, emphasis added)
But the report does not elaborate on this idea. So in what follows it will be assumed that the dual discount rate proposal is to be taken as implying a parallel dual portfolio investment strategy. This would make for a coherent combination.
4. Does the proposal produce different discount rates for past and future service?
It is easy to assume that, because the JEP report recommends using different discount rates for the pre-retirement and post-retirement periods of each member, it must also recommend adopting separate discount rates for liabilities resulting from past service (service up to the valuation date) and future service (service in the years following the valuation date). However, ‘pre-retirement vs. post-retirement’ and ‘past service vs. future service’ are separate distinctions, and the second recommendation does not automatically follow from the first.
The report itself is silent on this second recommendation. It presents its dual discount rate proposal only as a way of valuing the liabilities resulting from past service, and its Single Effective Discount Rate relates only to these liabilities. It says nothing about how to value the additional liabilities resulting from future service, thus how to calculate the rate of Future Service Contributions. Having said that, it seems that the only consistent way to develop the proposal so as to calculate these figures would lead to a separate Single Effective Discount Rate for future service liabilities. Here (in a simplified form) is how the calculation of Future Service Contributions would need to go.
First, work out the additional liabilities owed to each member as a result of their first year’s service after the valuation date. In the case of active members, value these liabilities using the same dual discount rate method as explained above, but value the liabilities at one year after the valuation date rather than at the valuation date itself. In the case of deferred and retired members, there are no additional liabilities because these members are no longer accruing benefits.
So for example the additional liabilities owed to Member 1 above as a result of the first year’s service after the valuation date would be discounted at the pre-retirement rate for 40 years and then at the post-retirement rate for 18 years, so as to value them at one year after the valuation date. The additional liabilities owed to Member 2 would be discounted at the pre-retirement rate for 5 years and then at the post-retirement rate for 18 years. There would be no additional liabilities owed to Member 3 because that member is no longer accruing benefits.
Second, add together the values of all these additional liabilities to get a single value, dated to one year after the valuation date, for the scheme’s total additional liabilities resulting from the first year’s service after the valuation date.
Third, perform the same calculations as above, but for the second and third years after the valuation date. This gives the values of the additional liabilities created in the second and third year, dated to the ends of the second and third years respectively.
Finally, average the results across the three years to get the average value of the additional liabilities created each year, dated to the end of that year, over the three years. Expressed as a percentage of the average expected sectoral payroll over the three years, this is the rate of Future Service Contributions for the three-year period.
It is now possible to calculate what single discount rate, if applied to the scheme’s total additional liabilities created over the three years, would have given the same rate of Future Service Contributions as above. This is the Single Effective Discount Rate for future service liabilities (SEDRf), which can be set alongside the previously calculated Single Effective Discount Rate for past service liabilities (SEDRp).
In this sequence SEDRf is calculated from the level of Future Service Contributions, just as SEDRp is calculated from the Technical Provisions. As with SEDRp, its purpose is just to provide a measure that can be used to compare the valuation method with others that use a single discount rate.
SEDRf will normally be higher than SEDRp, because the calculation of SEDRf refers only to liabilities owed to active members, which are discounted partly at the higher, pre-retirement, discount rate and partly at the lower, post-retirement, rate. By contrast the calculation of SEDRp refers to liabilities owed to active, deferred and retired members, and the part of these liabilities owed to retired members is discounted entirely at the lower post-retirement rate.
However, since the report does not mention the idea of a distinct SEDRf, in what follows we shall assume that when it uses the term ‘Single Effective Discount Rate (SEDR)’ this refers only to SEDRp.
5. The effect of the proposal on contribution rates
The report includes a table, provided by USS, to illustrate the effects the dual discount rate proposal would have had on the Technical Provisions and Future Service Contributions if it had been applied to the valuation as at 31 March 2018, under various different assumptions (p. 116). These effects can be extrapolated to the coming valuation, as at 31 March 2020, though only on the assumption that market conditions on the two dates are roughly the same. See Table 1.
(NB The report includes two versions of this table: the original version supplied by USS, as Table 1 at p. 116, and a second version, with the Normal Pension Age reduced from 66 to 65 but almost all other figures unchanged, as Figure 11 at p. 68. The original version has been used here.)
The 2018 valuation used a discount rate of gilts + 1.33%, meaning 1.33% above the yield on long-dated index-linked gilts. This gave the figures shown in row 1 of the Table for USS’s ‘upper bookend’, later called ‘option 1’ (USS, 2019a, pp. 17–18). In the end, USS moved from option 1 to its ‘option 3’, implemented in October 2019, under which the Deficit Recovery Contributions of 5.0% were reduced to 2.0% for the coming two years, bringing the total contribution rate down from 33.7% to 30.7%. However the rate of Future Service Contributions in both option 1 and option 3 is 28.7% (USS, 2019c, p.5, Mercer, 2019, p.3).
In all cases Table 1 assumes that the post-retirement discount rate is gilts + 0.75%. This is the figure that USS uses in calculating the value of liabilities on a self-sufficiency basis (USS, 2019a, p. 25). The Table suggests three possible figures for a pre-retirement discount rate: gilts + 2.5%, + 3.0% and + 3.5%. Then it calculates the Technical Provisions, the Future Service Contributions and the SEDR for each of these three choices of pre-retirement discount rate.
There is a complication in that, for each of these three possible pre-retirement discount rates, the Table uses two different assumptions about future CPI inflation. (1) ‘TP CPI’ takes RPI as predicted by gilt markets (as the difference between the yields on nominal and RPI-index-linked long-dated gilts, called ‘breakeven inflation’), and subtracts 130 basis points or 1.3% to give a predicted figure for CPI. This 1.3% consists in a predicted excess of 1.0% of RPI over CPI, with an additional 0.3% ‘risk premium’ to reflect future inflation uncertainty. This is the inflation assumption that USS uses when calculating the Technical Provisions. (2) ‘SS CPI’ makes the same assumption about inflation for the pre-retirement years of each member, but then a slightly higher inflation assumption, subtracting only 80 basis points or 0.8% from breakeven inflation, for the post-retirement years. This 0.8% consists in a predicted excess of 0.8% of RPI over CPI, with no additional risk premium. The assumption made for the post-retirement years is the one that USS uses when calculating the value of liabilities on a self-sufficiency basis (USS, 2019a, p. 36).
A second complication is that it is not clear how USS calculated the figures for ‘Future service cost’ (i.e. Future Service Contributions) in the Table. Presumably they used a method like the one outlined in the last section, without going on to calculate a figure for SEDRf. The ‘Single equivalent rates’ calculated in the last column must refer just to SEDRp, as the note to the Table says that they are ‘calculated on past service liability equivalent basis’.
The report does not try to justify its choices of the three possible figures for the pre-retirement discount rate by reference to the prudently expected rate of return on a return-seeking portfolio. Instead it points out two facts (p.69):
First, a pre-retirement discount rate of gilts + 2.5% would give a similar Technical Provisions figure to the one used in the 2018 valuation, and an SEDR (1.34%) that is below the Upper Quartile of the distribution of the Tranche 12 (September 2016 to September 2017) valuations of DB schemes (1.46%) (TPR, 2019a, Table 4.11).
Second, a pre-retirement discount rate of gilts + 3.0% would give an SEDR (about 1.50%) that is only marginally above the Upper Quartile (1.46%) of the distribution of Tranche 12 valuations (TPR, 2019a, Table 4.11).
So, reading between the lines, the report seems to imply that a reasonable figure to adopt for the pre-retirement discount rate would be gilts + 3.0%. However, it should be added that a pre-retirement discount rate of gilts + 3.5% would give an SEDR (about 1.65%) which is still not far above the Upper Quartile (1.46%) and well short of the 95th centile (2.40%) of the distribution (see Table 1 above and TPR, 2019a, Table 4.11). Furthermore, TPR’s analysis of Tranche 12 schemes aggregates schemes which are open, closed to new members, and closed to all new accruals. It is likely that if the analysis were restricted to open schemes such as USS, the SEDR given by a pre-retirement discount rate of gilts + 3.5% would come much lower in the distribution.
Figures from Table 1 can be combined with data from other sources to show what effect adopting a gilts + 3.5% pre-retirement discount rate for the 2018 valuation would have had on members’ and employers’ contribution rates resulting from that valuation. This is done in Table 2.
The decision on how to share the total contributions has to be made not by USS, but by UUK–UCU’s Joint Negotiating Committee (JNC).The coloured figures in rows 4–5 of the Table show the effects of two possible ways of sharing this total.
The first (blue figures) assumes that changes in the Future Service Contributions from the 2014 level are shared in the ratio of 35:65 between members and employers, but that all Deficit Recovery Contributions and Surplus Elimination Adjustments are borne by or credited to employers. This follows the precedent set by the 2014 and previous valuations. It can be justified by the view that employers, rather than members, have the financial capacity to make good on past underpayments, and for the sake of consistency they should likewise benefit from past overpayments. (For a separate argument for this approach, based on intergenerational justice, see #USSbriefs76 Chitty and Wilson 2019b, sec. 8.)
The second (red figures) assumes that all changes in the total contribution rate from the 2014 level are shared in the ratio of 35:65 between members and employers, regardless of whether these result from changes in Future Service Contributions, or in Deficit Recovery Contributions or Surplus Elimination Adjustments. This follows the precedent set by the way contribution sharing for option 3 was decided by the JNC in August 2019 (although against the will of the UCU negotiators and only thanks to a casting vote by the Chair). Here the whole change from the 2014 baseline, including the slight reduction in Deficit Recovery Contributions, was shared in the ratio of 35:65 between members and employers. This way of sharing contributions can be justified by the view that, by default, changes in Future Service Contributions are already shared in the 35:65 ratio, and Deficit Recovery Contributions and Surplus Elimination Adjustments should be regarded as rectifying levels of Future Service Contributions from past valuations that have turned out to be too low or high.
The two tables, taken together, show that the effects of the dual discount proposal on members’ and employers’ contribution rates will vary significantly depending on three separate factors: the figure chosen for the pre-retirement discount rate, the assumption made about inflation, and the decision made about contribution sharing.
6. For and against the dual discount rate proposal
The obvious advantage of the dual discount rate proposal is that it provides a reasoned basis for reducing both members’ and employers’ contributions relative to those derived from USS’s 2018 valuation. To the extent that contributions can be kept relatively low, the scheme can continue to recruit new members and remain in financial health indefinitely. However the proposal has two key disadvantages.
The first is linked to the way that, at least in appearance, the proposal ties both the pre-retirement and post-retirement discount rates, and therefore the calculation of Technical Provisions and Future Service Contributions, to the long-dated gilt yield. In all the examples it gives, the report assumes that the post-retirement discount rate is to be set at gilts + 0.75%, and the pre-retirement rate at gilts plus some set figure (for example 2.5%, 3.0% or 3.5%). The implication is that once the figure of gilts + 0.75% is agreed for the post-retirement discount rate, and a certain ‘gilts plus’ figure for the pre-retirement discount rate, these figures will remain in place through future valuations.
If all the parties could agree to fix the post-retirement and pre-retirement discount rates for future valuations by reference to the gilt yield, it would surely reduce the scope for future conflict, for the long-dated index-linked gilt yield at any time is a matter of public knowledge, so that no-one could claim that the discount rate was being calculated on the basis of secret procedures. Given the current low level of trust in USS on the part of members, this an important point. However, the problem with any ‘gilts plus’ based valuation is that it is likely to lead to ever lower discount rates and so higher contributions in the future. Figure 2 gives a graph of 30-year index-linked gilt yields since 2013.
Figure 3 gives a graph of 10-year index-linked gilt yields since 1990.
It is clear that the trends are both relentlessly downwards. (In general the yields on 10-, 20- and 30-year gilts move in tandem; see the graph in Tapper, 2019.) They are part of a world-wide pattern of declining nominal and index-linked government bond yields since the mid-1980s. If this trend continues, then any valuation method that ties its discount rate (or discount rates) to a fixed margin above the long-dated gilt yield is likely to lead to declining discount rates and so rising contribution rates over the coming valuations.
It might be said that if gilt yields are going to decline over the next 20 years then the rate of return of every other asset class is bound to decline as well, so that setting not just the post-retirement but also the pre-retirement discount rate at a fixed margin above the long-dated gilt yield is realistic. The background assumption here, taken from standard economic theory, is that the return on equities (and similar return-seeking assets) will always remain around a fixed margin above the return on gilts, reflecting the ‘risk premium’, or the disvalue that investors put on the additional risk of holding equities. But this theory has been heavily criticised. For example, Woon Wong calculates that RPI-adjusted UK equity returns declined by only about 1.5% from the 1990s to the 2010s (from 6.3% in 1990–99 to 4.8% in 2008–17, Wong, 2019, p. 9, Table 3). Meanwhile Figure 3 shows that 10-year index-linked gilt yields, which are also measured in terms of RPI, fell by about 4% over the same period.
As mentioned above, USS has in the last few years used a discount rate of gilts + 0.75% to calculate the value of the scheme’s liabilities on a self-sufficiency basis. However, in calculating the value of liabilities for Technical Provisions purposes it has used a discount rate based on first breaking down its actual and planned investments into different types of assets, then predicting a 67%-probable rate of return for each type of asset, and then aggregating the results. This is what it calls its ‘Fundamental Building Blocks’ approach (Gray and Cardinale, 2018). The discrepancy between the two methods of deciding a discount rate is reasonable, for in the case of calculating the value of liabilities on a self-sufficiency basis the background assumption is that the funds to cover the liabilities will be invested in a ‘self-sufficiency’ portfolio consisting mainly of bond-like assets, and the returns on these assets are likely to run in parallel with gilt yields. By contrast, in light of the widening divergence between the returns on return-seeking assets and gilt yields over the last 30 years, to adopt a dual discount rate valuation method in which both the post-retirement and the pre-retirement discount rates are set at a fixed margin above long-dated gilt yields looks excessively pessimistic.
The second disadvantage of the dual discount rate proposal is independent of the way it ties discount rates to the gilt yield. It is that, because the proposal effectively weights both SEDRp and SEDRf by the relative numbers of junior and senior scheme members, it creates the danger of a ‘rising contributions feedback loop’. Suppose that the next valuation for some reason leads to an increase in contribution rates. This is likely to dissuade a certain proportion of junior staff who would otherwise have joined the scheme in the following three years from doing so. This will tip the mix of the scheme’s liabilities in the direction of those discounted at the lower, post-retirement, rate at the following valuation, raising the contribution rate at that valuation and correlatively reducing SEDRp and SEDRf. This in turn may dissuade more junior staff from joining in the following cycle. Over several cycles, the result may be to raise contribution rates to a point at which employers declare that the whole scheme is unsustainable and attempt to close it — as they did at the start of 2018.
7. For and against the dual portfolio investment strategy
As has been said above, the JEP’s report does not explicitly tie its proposal for a valuation method to a parallel investment strategy, under which the funds notionally associated with each member’s liabilities are invested in a return-seeking portfolio until the member retires, and thereafter in a low-risk portfolio. Yet at the same time the proposal looks arbitrary unless it is linked with such a ‘dual portfolio’ investment strategy. So here it will be assumed that the proposal is intended to go together such an investment strategy. This section looks at the advantages and disadvantages of the strategy.
Its greatest advantage is that it provides a way of loosening the noose of automatic de-risking mandated by Test 1, which threatens to strangle the scheme over the next 20 years. In the context of a decades-long secular decline in gilt yields, this prospect becomes even more likely. By contrast, the dual portfolio investment strategy opens up the possibility of halting this de-risking strategy, and with it the resulting increases in contribution rates.
At the same time, this strategy directly addresses the basic rationale that has been given for the de-risking strategy: that once members are in retirement their pensions should be paid from a fund invested in low-risk assets, the returns on which are more or less guaranteed to generate the benefits they are owed.
The greatest disadvantage of the dual portfolio investment strategy is that, despite this loosening, it binds the part of the fund notionally associated with retired members to a portfolio of low-risk investments for the foreseeable future. From Table 1 above we can calculate that, under the various assumptions made, the part of the value of the liabilities which is owed to retired members (‘Technical Provisions pensioner’) is always over 40% of the value of the liabilities as a whole (‘Technical Provisions total’). (To be exact, it is always between 41% and 44%.) So, if implemented immediately, the dual portfolio investment strategy would require that, for the year immediately following the valuation, over 40% of the fund be invested in low-risk portfolio, specifically in a ‘self-sufficiency’ or mainly bond-like portfolio that has a 95%-probable rate of return of gilts + 0.75%. If new junior members enter the scheme faster than older members retire, then this proportion will decline over time. But if we assume that rates of entry and retirement are equal, then the proportion will remain the same indefinitely. Depending on exactly how a self-sufficiency portfolio is defined, continuing to holding 40% of the fund in such a portfolio might not be that different from maintaining the present portfolio mix (which has 35% of the fund in bond-like assets). This is a definite improvement on USS’s plan to automatically de-risk the fund over the next 20 years. However there is surely something absurd in indefinitely holding over 40% of the scheme’s fund in assets that are currently (with 95% probability) expected to return only 0.75% above the rate of long-dated index-linked gilts. Given that the yield for 30-year index-linked gilts is currently about -1.75% (see Figure 2), this currently means a return of only about -1.0% in RPI terms, or 0.0% in CPI terms, with every prospect that these figures will fall further in the coming decade.
There is therefore a danger that the dual portfolio investment strategy will lock the fund into a portfolio that, while not as disastrous as the present strategy of de-risking over 20 years, will still be grossly unproductive and will still result in unnecessarily high, and potentially unsustainable, contribution levels.
8. A gilts-decoupled version of the dual approach
In this final section it will be argued that, if we think of the dual discount rate proposal and the dual portfolio investment strategy as a single package, then some of the main criticisms made here could be met by a version of this package that is ‘decoupled’ from the gilt yield, by making two modifications to it.
The first modification concerns how the portion of the scheme’s fund allocated to a return-seeking portfolio is invested. The basic content of the return-seeking portfolio could be decided for the next, say, 20 years. This content could be used to calculate a pre-retirement discount rate using a transparent version of the Fundamental Building Blocks approach, i.e. one in which the data and calculation method are made public so that in principle anyone could check the calculations. The Technical Provisions and Future Service Contributions could then be calculated on the basis of this pre-retirement discount rate, together with the post-retirement rate of gilts + 0.75%. At one point, the JEP report states: ‘Although the discount rates used have been expressed as Gilts+, the Panel believes that there is considerable benefit in representing them as CPI+ for the pre-retirement years’ (p. 68). This suggests that at least some of the JEP’s members have been thinking along the lines of this modification.
The second modification concerns how much of the scheme’s fund is allocated to a low-risk portfolio, specifically a self-sufficiency portfolio. Advocates of the ‘cash flow’ approach to pension scheme valuation have argued for a number of years that, as long as forecasts show that for the foreseeable future the scheme’s annual incomings will equal or exceed its outgoings, there is no need for it to hold any of its fund in low-risk assets in order to guarantee that it will be able to meet its liabilities as they fall due (First Actuarial, 2014, sec. 6; King and Kay, 2018; #USSbriefs7 Leech, 2018; Wilkinson and Curtiss, 2018). It certainly looks irrational to keep the whole of the portion of the fund that is associated with the liabilities owed to retired members, namely over 40% of the total, tied up in a self-sufficiency portfolio, especially given that some of those liabilities may not fall due for up to 18 years. However USS could instead operate a ‘five year rule’, under which at any given time only that part of the fund associated with liabilities which will fall due within the following 5 years would be invested in a self-sufficiency portfolio, while the rest would remain in a return-seeking portfolio. At the end of each year, the funds associated with those liabilities that fall due between five and six years later would be moved from the return-seeking to the self-sufficiency portfolio. The Technical Provisions and Future Service Contributions could then be calculated on this basis, using a pre-retirement rate determined by the first modification above and a post-retirement rate of gilts + 0.75%.
Untethered in these two ways from the gilts yield, the USS fund could be invested more fully in return-seeking assets. This would enable the scheme’s contribution rates to stay correspondingly lower while still guaranteeing the pensions of retired members. Such a version of the JEP report’s dual discount rate and dual portfolio proposal could genuinely show a way forward for the scheme. It should be on the table in the three-way talks on implementing the JEP report scheduled for the second half of this January (UCU, 2020).
Birch, Michael (2019) Letter to David Eastwood. 6 August 2019. The Pensions Regulator.
Chitty, Andrew (2020) USS portfolio mix 2000–2019. Google Docs, 14 January 2020.
Chitty, Andrew and Tim Wilson (2019a) Why Test 1 must be dropped: a critique of its design and implementation. #USSbriefs68, 7 February.
Chitty, Andrew and Tim Wilson (2019b) Approaches to risk: uncertainties in USS’s Test 1 and stochastic modelling. #USSbriefs76, 28 July.
Coughlan, Guy (2019) Protecting the promises made to members. USS, 28 February.
Department for Work and Pensions (2018) Protecting defined benefit pensions schemes. March.
Dillow, Chris (2019) Low yield warning. Investors Chronicle, 14 March.
Fairs, David (2019) Protecting DB savers: our expectations are clear. The Pensions Regulator, 9 May.
Fixed Income Investor (2020) Chart of UK gilt index-linked Stk 0.25% 22 Mar 2052. January (accessed 14 January 2020).
Gray, Roger and Mirko Cardinale (2018) USS Investment Management’s Fundamental Building Blocks (FBB) approach to expected returns. USS, 22 March.
Joint Expert Panel (2019a) Second call for submissions. 10 May.
Joint Expert Panel (2019b) Report of the Joint Expert Panel. 13 December.
King, Mervyn and John Kay (2018) USS crisis: can the pension system be reformed?. Times Higher Education, 6 September.
Leech, Dennis (2018) Is there really a USS deficit?. #USSbriefs7, 5 April.
Marsh, Sam (2018a) Understanding ‘Test 1’: a submission to the USS Joint Expert Panel. #USSbriefs32, 10 July.
Marsh, Sam (2018b) A flawed valuation: the layperson’s guide to my findings on USS’s ‘Test 1’. #USSbriefs58, 15 October.
Mercer (2019) Scheme funding report of the actuarial valuation, Universities Superannuation Scheme, as at 31 March 2018. 16 September.
Otsuka, Michael (2018) USS’s valuation rests on a large and demonstrable mistake. Medium.com, 13 October.
Tapper, Henry (2019) The PPF could grow up a ©DC lifeboat. Age Wage, 24 December (accessed 10 January 2020).
The Pensions Regulator (2019a) Scheme funding analysis 2019: annex. June.
The Pensions Regulator (2019b) Investment guidance for defined benefit pensions schemes. September.
Universities Superannuation Scheme (2014) 2014 Actuarial Valuation: a consultation on the proposed assumptions for the scheme’s Technical Provisions and recovery plan. October.
Universities Superannuation Scheme (2019a) 2018 Actuarial Valuation: a consultation with Universities UK on the proposed assumptions for the scheme’s Technical Provisions and Statement of Funding Principles. 2 January.
Universities Superannuation Scheme (2019b) The 2017 valuation has been finalised. 31 January.
Universities Superannuation Scheme (2019c) 2018 valuation: trustee’s reply to UUKs feedback and questions on the consultation on the 2018 Technical Provisions. 7 May.
University and College Union (2020) Dates set for talks on future of pension scheme at heart of university strikes. 8 January.
Wilkinson, Tim and Frank Curtiss (2018) Death by discount rate: the fundamental flaws of the accounting approach to pension scheme valuation. Professional Pensions, 29 May.
Wong, Woon (2019) The discount rate debate and its implications for defined benefit pensions. Cardiff Economics Working Papers E2018/12, March.
I am grateful to Felicity Callard, Gail Davies, Jackie Grant, Kevin Wesbroom, Tim Wilson, Woon Wong and especially Michael Otsuka for comments on earlier drafts. All errors remain my own.
Update 18 January 2020
The author has responded to comments on his brief here.
This is a USSbrief, published on 15 January 2020, that belongs to the OpenUPP (Open USS Pension Panel) series. This paper represents the views of the author only. 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 #USSbriefs89 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.