Understanding ‘Test 1’: a submission to the USS Joint Expert Panel
This is a USSbrief that belongs to the OpenUPP (Open USS Pension Panel) series, and has been submitted to the UCU-UUK JEP (Joint Expert Panel).
USS’s Test 1, introduced in the run up to the 2014 valuation, is a risk-management tool described in the glossary of the USS 2017 Actuarial Valuation (consultation document, 1 September 2017) as:
a test designed to measure whether or not the long-term risk in the DB section of the scheme is within the risk appetite agreed between the trustee and sponsoring employers. The test checks that the difference between self-sufficiency and technical provisions in 20-years time does not become too large for the employers to support.
What is strangely absent in any of the documents I have seen is a formal description of how this test is applied. Below is my best attempt to give such a definition and a discussion of its implications.
2.1. Reliance on covenant. To apply Test 1, USS first introduce their reliance on covenant metric, defined as the gap between the liabilities on a self-sufficiency basis (which assumes safe, low return investments) and on a technical provisions basis (this being the standard, prudent valuation of the liabilities). Note that these two methods of valuing the liabilities differ mainly in the rate used to discount future benefits; that is, they use different discount rates.
The standard choice for the discount rate used for the calculation of the technical provisions is the best-estimate of the returns on the investments of the fund, adjusted downwards for prudence. I will refer to this as the standard discount rate later.
Test 1 is designed to ensure that the reliance on covenant metric evaluated at Year 20 does not exceed a pre-determined value. In other words, it aims to make the value of the technical provisions at Year 20 ‘close’ to the self-sufficiency valuation at Year 20 by determining a new discount rate for the latter which achieves the correct gap.
2.2. The algorithm. What follows is my best understanding of how the algorithm is applied. I do not know of any documentation contradicting this interpretation.
(1) Decide on a maximum allowable value, X, for the reliance on covenant.
Previously, USS have calculated this as the present value of 7% of pensionable salaries over Years 1–20. In the USS 2017 Actuarial Valuation (consultation document, 1 September 2017), such a calculation gives a figure of X = £13bn, which the trustee chooses to tighten further to X = £10bn.
(2) Decide on a method of inflation to convert X into a figure relevant at Year 20. Call this X′.
The method chosen in the USS 2017 Actuarial Valuation (consultation document, 1 September 2017) is to revalue X according to CPI, giving the real-terms equivalent value at Year 20. Employers were consulted on other options, including revaluation by predicted salary growth, an option which would have loosened the effects of Test 1 considerably.
(3) Determine the self-sufficiency liabilities at Year 20. Call this S.
The self-sufficiency liabilities at Year 20 are the value of the benefits to be paid in Years 21 onwards, discounted according to the self-sufficiency discount rate. It includes projected benefits accrued in Years 1–20 (see Response to the Valuation Discussion Forum, USS, 22 November 2016, p.12)
(4) Set the required value of the technical provision liabilities at Year 20 as
T = S − X′.
(5) Determine a discount rate for Years 21 onwards which values the Year 20 liabilities at T.
In practical terms, this involves lowering the standard discount rate for Years 21 onwards until the Year 20 value of the liabilities climbs high enough so as to reach T.
(6) Choose a de-risking path over Years 1–20 so that the discount rate falls from the standard discount rate at Year 1 to the rate determined above at Year 21.
Note that this drives a change to the investment portfolio in Years 1–20 to one which will deliver returns low enough to give the required discount rate. This will likely involve a decrease in the holdings of equities and property in favour of long-dated bonds.
(7) Use the newly determined ‘de-risked’ discount rate to value the Year 0 liabilities.
This then determines the funding position (surplus/deficit) and future service costs for the scheme. The overall effect of the de-risking will have been to amplify any deficit and increase the future service costs.
3.1. The scheme’s assets are ignored by Test 1. In the algorithm described above, no attention is paid to the assets held by the scheme until the final step (when the funding position is determined). In other words, whether the fund has a large surplus or deficit and whether certain equities or property are seen as good secure long-term investments are irrelevant: in all cases the test would force a de-risking of investments.
An alternative definition of the reliance on covenant as the self-sufficiency valuation less the projected value of the assets would go some way to addressing this problem. While, on first glance this would seem to be an equivalent definition, it would make a significant difference to how any increased gap would be addressed, allowing, for example, increased contributions to be made without the de-risking of investments. I am happy to clarify this point further if needed.
3.2. Indexation of reliance on covenant. The decision to inflate the target or maximum reliance figure by CPI rather than salary growth is an unnatural choice. If the value of 20 years’ worth of contributions at 7% of pensionable salaries is used to determine the value of the maximum reliance, then at Year 20 the relevant contributions are those in Years 21–40. These would best be approximated not by the contributions in Years 1–20 uprated by CPI, but the contributions in Years 1–20 uprated by salary growth. It is unfortunate that more discussion of this did not occur during the consultations USS ran with Universities UK in 2017.
3.3. The difference between two large numbers. In their excellent 2014 analysis, First Actuarial observe that ‘the reliance on covenant metric is the difference between two similar large numbers, which makes it very sensitive to small changes in either of the large numbers’ (see Report to the USS paper: 2014 Actuarial Valuation, First Actuarial for UCU, November 2014, p.17). They demonstrate how small changes in those numbers have drastic effects on benefits and/or investment decisions. That is, the difference between Test 1 almost being met and Test 1 actually being met can be highly significant.
4. Further information needed
USS have never publicly released any of the following pieces of information.
(1) The expected gap between the self-sufficiency and technical provisions liabilities at Year 20 before Test 1 is applied, and how this compares to the target/maximum reliance inflated with CPI/salary growth.
(2) As in 1, but updated to take account of 2018 data on asset growth;
(3) The funding position and future service costs of the scheme with Test 1 absent;
(4) As above, but under the best-estimate assumptions;
(5) As in 3 and 4, but updated to take account of 2018 data.
Some of these numbers are possible to estimate, but others are close to impossible without USS’s underlying datasets. Some of my findings are below.
(1) To estimate these need reliable data on projected cashflows from USS, which have not been made available.
(2) As above.
(3) While I haven’t been able to reliably determine the future service costs, the funding position on past service with Test 1 absent seems to be approximately break-even (negligible/no deficit). This tallies with what a USS representative told the University of Sheffield USS Working Group verbally, where future service costs were given as 26% of salary (although it wasn’t clear whether this included all benefits or just the DB section). USS should be asked to clarify.
(4) With Test 1 absent, my workings show the fund to have a surplus of close to £13bn on a best-estimate basis. Again, future service costs on this basis are hard to establish.
(5) The assets as at March 2018 are reported to be around £64bn, which will include around £1.9bn of contributions. The ‘on-track’ figures for the position of the fund is around £59.7bn. My estimates show that this results in a surplus with Test 1 absent of a little under £3bn. On a best-estimate basis, the surplus may climb as high as £14bn.
5. Final comments
I do not believe Test 1 is fit for purpose. In fact, its purpose has been hard to determine. Test 1 is likely to be responsible for all of the perceived funding problems in the scheme. Given how poorly documented it is, and how badly understood by many, including those making decisions, it must be subjected to maximum scrutiny by the JEP.
This is a USSbrief that belongs to the OpenUPP (Open USS Pension Panel) series, and has been submitted to the UCU-UUK JEP (Joint Expert Panel). 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 #USSbriefs32 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.