You wait ages for one to come along then two come along at once. Not buses, but reports disrupting established thinking on how we value pension liabilities.
Nikesh Patel, Head of Investment Strategy UK & Maya Beyhan, Investment Strategist at Kempen Capital Management, have sent me their idea for a “Superflat” discount rate while Mercer will be presenting “The future of pension scheme funding – a new approach to pension valuations in a low yield environment” on 13 June 2017 at the PLSA.
So here’s Nikesh and Maya’s paper – which tips its hat to the FAB Index while presenting an alternative approach to measuring long term liabilities.
A billion here, a billion there, and pretty soon you’re talking about real money
Consider the amount £500,000,000,000: enough to increase the total annual investment AND dividends of the UK private sector by 33%, for the next 10 years. Brexit would be but a blip in even the hardest of exits. And that’s only half the story; a similar amount would be released from a swath of poisonous investments guaranteed to lose money into investments which actually have some useful economic utility – and which have some chance of generating returns above zero.
Imagine that. The impact on UK plc would be astounding. All you have to do is fix a broken measure of pension scheme liabilities.
As we keep hearing, the defined benefit (DB) pension system is in crisis. Deficits have ballooned to record levels – somewhere between £400bn and £1tn depending on the measure – straining the balance sheets of corporate and public sponsors, and the wider capital markets, as the £1.5tn pension industry struggles to both find yield and hedge itself against interest rates (the risk driver underlying gilts). This is despite the fact that the underlying pension benefits have scarcely changed, more cash contributed by corporate sponsors than ever before, and investment returns in recent years that would make hedge funds green with envy. So why are things so bad?
Well it depends on how you look at it – and how you measure it. Actuarial valuations have us trapped in the reality provided by their gilt based measures. Those numbers are driving investment behaviour within the schemes as well as the investment behaviour of the corporate sponsor (having to set aside massive amounts of free cashflow to meet their pension deficit recovery schedules). I’m not espousing the end of actuarial valuations (tempting though that might sound) – they are an attempt to provide important and useful data. We must however recognise that they are based on imperfect data and far from perfect assumptions, and we ought to be careful of falling into a circular argument of making decisions simply because they optimise these choice of measures.
It is no measure of health to be well adjusted to a profoundly sick society. ~ Krishnamurti
Mistaking the measure for the thing itself
Valuing the assets of a pension scheme is straightforward. Without getting philosophical, all of the underlying stocks, bonds, properties and other holdings (including large caches of cheese, whisky, wine and art in some cases) have a price at any point in time that can be used to provide a market consistent valuation – the value that is expected to be realised if all of these assets were sold to other market participants at the prevailing price and conditions.
Valuing the liabilities of a pension scheme is not so easy. The liabilities are a series of payments that have been promised by a sponsor to its employees long into the future. The amount of each future payments is determined by the earning of employees whilst they were employed and how long they will live. These are also impacted by inflation – of wages and earnings whilst employed, and upon promised increases to pensions in payment. Projecting these benefit promises into the future is uncertain, but easy enough – the only real uncertainties are whether your assumption about future inflation was correct and if people live longer or shorter than expected.
Now say that you have projected a payment of £100 due in 50yrs time. What is that payment worth today? Does that even matter? If you are completely certain that the sponsoring employer is not going to go bust, and will definitely be willing and able to make that payment, I suppose it doesn’t. But in the real world, that sponsor might not be around and be able to afford that payment when it falls due. And so it has decided to set aside some money that it will invest for those 50yrs, which will then be used to make that payment. This is referred to the “present value”.
A present value therefore needs to have some expectation of what investment return the pension assets will have into the future (this is the “discount rate”). This requires thinking about what the investment portfolio will look like – both now and in the very long term. Assuming that you simply held all the cash under the mattress (a very large mattress), you could assume zero investment returns, and hence the present value is simply the sum of all future benefit payments – a daunting amount indeed.
Most UK pension schemes assume a discount rate which is explicitly linked to gilts, essentially the risk free asset (if you exclude the possibility of a UK default). This assumes that over the long term, these schemes will tend to aim for a portfolio which is largely gilts based in nature. Gilt yields being where they are, this is not much better than assuming a zero return (indeed, allowing for inflation reducing your purchasing power, you are actually guaranteed to earn less than a zero return…). And the valuation of these pension liabilities will change daily, and potentially dramatically.
There has been an astronomical rise in the value of UK pension liabilities over the last decade due to the effect of falling gilt yields – despite no change to the underlying promises made to members. Taking that example of a £100 payment and a gilt yield of 2%, £37 would need to be set aside today for the next 50yrs. Only a few years ago, the yield on gilts was more than 4% – so only £14 would have needed to have been set aside.
At the same time, the sponsors of these schemes are required under accounting rules to assume the sole assets held will be AA-rated UK corporate bonds. This is a less shocking number than using gilts, but hardy ideal, as corporate bond yields are still theoretically linked to gilt yields.
The problem with these measures is that they effectively dictate the assets pension schemes must invest in to reduce the measurement volatility. They are held because the liability measurement methodology is linked to gilts, so holding gilts in the asset portfolio matches (and therefore offsets) this liability risk. Schemes are buying gilts because they hedge this choice of measure – but this is a measure which has no link whatsoever to the underlying liabilities (as noted above, those are only linked to inflation and life expectancy). The current approach, and the consequential ballooning of liability valuations has led to the shouts of DB pensions being unaffordable, and for younger generations led to the demise of any hope of getting these “gold plated” pensions.
This is important. Holding gilts in the size that pension schemes hold them (and lever up) is simply not a rational thing to do for an investor, especially given much of these holdings are guaranteed to lose money under the current environment of negative real yields. If trustees (rationally) chose to invest in assets which are not gilts, they are effectively punished when the values of gilts rise and they create a deficit on this measure (particularly true during economic crises). In other words, the choice of measurement is driving investment decisions, which is clearly a nonsense. We are mistaking measure of the thing for the thing itself.
“Sometimes our fetish with measurement is not useful at all. What we measure affects what we do, and if we aren’t measuring what we think we’re measuring, or if what we measure doesn’t matter, then we might end up improving the kind of performance that we measure, but not the performance that we care about. Insidiously, we sometimes even bring about the opposite of what we want.” ~ Stiglitz, Vanity Fair
Measure what is measurable, and make measurable what is not so.
One might expect the approach would be the other way around – that a rational investor would choose a set of assets and determine the expected return, and use that expected return to arrive at the discounted present value. So if in a miraculous world they could reasonably expect a return of 6%, only £5 would need to be set aside today (the wonders of compounding). But that’s not so easy either. As you move away from the risk free asset, the investment returns become less certain – you expect 6%, but you might get -6% or some other number. And more fundamentally, even the expected return may change; we are in a world of low yields – the expected return on the risk free asset has halved in the last 5 years.
Fundamentally, any approach which is based upon holding a predetermined set of investments of any flavour (be they gilts or otherwise) is subject to this issue: they are not time consistent, because the expected returns of those assets themselves may change over time, even if the underlying projected pension payments are entirely unchanged. We are moving invisible goal posts and attempting to make investment decisions at the same time. Zeno’s paradox.
A more useful, and rational, measure needs to be time consistent. It should also avoid driving the investment decision towards a narrow set of assets and it should treat all schemes the same, so that a measure of liabilities for one scheme is directly comparable to the liabilities of another – the relatively weaker schemes where additional cash or support might be needed can be easily identified, leading to more efficient capital allocation decisions by shareholders. And it would be preferable if the approach was simple, as that is worthwhile in its own right. There is a great deal of complexity (and cost) in the actuarial process, making the valuations process and negotiations incredibly arcane.
A potential solution is to use a flat discount rate of, say, 3.5% pa, over the whole of the future liabilities. Not based on any predetermined set of assets, not linked to gilt yields or any other market force, or changing over time. Simply 3.5%. Why not 3% or 5% or 8%? Well, allowing for a measure of prudence over the discount rate, a 4% investment return is realistically achievable with a modest degree of risk: equity returns tend to 4% over even the most extreme 20yr periods and for most historic periods 4% could have been constructed of solely high quality bond portfolios. It is a level which even in the current environment encourages a diverse mix of assets – it does not require an overly large amount of equities (though encourages more than the current system), nor does it force one into bonds. It would mean schemes would invest dynamically to maintain a portfolio at their target return with the lowest risk possible using a range of assets, including potentially some gilts – but not forced into gilts at current depressed levels. In fact, the ideal end-game strategy is actually one of gilts – if gilt yields ever return to their former levels, the rational investor will naturally seek to hold gilts as the risk free asset (required return at least risk).
A higher fixed discount rate would naturally force more exposure into more risky assets – witness the American model of spiralling unaffordability. A lower fixed rate would be no better than the current situation, concentrating schemes into the relatively limited UK fixed income markets – there simply aren’t enough gilts to go around in any case, and the corporate bond market isn’t sufficiently broad either.
A flat discount rate at this level would also mean a significant capital improvement to UK corporates, approximately reducing the absolute measured size of pension liabilities from c£2.1tn (on a gilts+0.5% discount basis, which is regarded as a proxy for “self-sufficiency”) by £500bn, versus c£1.5tn of assets – essentially closing the existing deficits being measured for the majority of pension schemes. Combine this with the latest Continuous Mortality Investigation (CMI) projections model produced in February and deficits in the UK pensions market all but disappear in all but the most extreme cases (indeed most schemes would be in a modest surplus, and some would be in rude health). The amount of future recovery contributions would be consequently be reduced and the consequential limitations on UK corporates would be removed – on employment and investment (spurring UK growth), as well as substantially higher dividends (which would immediately be of additional benefit to pension schemes who still hold large amounts of UK equities).
Finally such a move would end the dominance of gilts in pension investing. Whilst hedging inflation and longevity will continue to matter, hedging interest rates will cease to be a priority, thus releasing the hundreds of billions of gilts they already hold into the market (potentially revealing the true prevailing yield and cost of government financing than is being reflected by the current short squeeze in gilts), hugely reducing the amount of leverage in pension schemes and the financing costs related to these positions (that last detail could reduce aggregate running costs of pension schemes by £2-5bn per year). At the same time, selling these gilts will release the capital needed to find higher yielding, and more economically useful, assets to achieve the investment targets and meet income generation needs as most schemes turn cashflow negative (paying out more in benefits than they receive in cash contributions). Consider all the infrastructure needs of the UK funded, and more projects still, not to mention putting UK pensions on a field with the biggest global investors.
And this isn’t just pie in the sky thinking. It can be done within the flexibility of the existing rules! That’s the beauty of it. One or two schemes already use a similar approach already because they have no choice (they are very large, and using the conventional approach would be too depressing to bear thinking about); there is no reason why other schemes cannot choose to use this (or similar) approach of their own volition.
Measuring the thing itself
If we consider changing the discount rate in this way, we need a measure of risk which is capable of comparing the old and new measure, and preferably one which looks beyond asset side risks alone and does incorporate the ultimate objective to pay the pension benefits as they fall due.
One approach is to measure the chance that the pension scheme runs out of assets before all the payments are made, with reference to the relative size of the remaining cashflows in that scenario (ruin when there is 50 years of remaining benefits to a large pensioner population is much worse than if there is only 5 years of benefit to a tiny fraction of the surviving pension population).
We’ve shown in the chart (Figure 1 in the linked paper Superflat – full document) this expected shortfall of cash for a 70% funded pension scheme with an investment strategy driven by the current measurement approach (Model 1), compared to an investment strategy invested against a fixed target return of 4% pa allowing for inflation (Model 2), excluding any further cash contributions at all – saving UK plc nearly £20bn per annum.
For both Model 1 and Model 2, our hypothetical schemes begin to run out of assets around the same time. However, the investment strategy of Model 2 has a significantly lower expected shortfall in ruinous scenarios – this is hardly surprising, it has a higher expected rate of return as less of its assets are constrained by holding gilts.
We also explored the impact of including the current cash contribution schedule on the expected shortfall in ruin for both models (labelled as Model 1a and Model 2a), as shown by the second chart (Figure 2 in the linked paper Superflat – full document). The effect is material: the first scenarios where ruin starts to appear are about 5 years later but also the maximum shortfall is reduced to 1.2% from 5.8% for Model 2, and to 2.3% from 9.6% for Model 1. This result can be explained by that when the assets are fed extra cash contributions, there is (a) the simple fact of more assets but also (b) the impact of cash coming in reduces the amount of cash needing to be raised from asset sales, reducing the very real risk of cashflow negativity and the resulting path dependency of investment returns.
Extending the cash contribution schedule beyond the current recovery plans would naturally reduce these ruin scenarios even further, as would open schemes with the effect of new accrual of liabilities (with equivalent contributions) actually reducing ruin risks even more as this cashflow further reduces the path dependency of investment returns versus benefit payments to existing pensioners.
The greater good
Is this approach perfect? Probably not. But it will drive far more rational decision making by pension investors and result in useful economic activity from pension portfolios – there are hundreds of billions of pounds that can be re-allocated within pension funds’ existing holdings and £500bn of cash contributions from UK plc that could otherwise be used for corporate investment or dividends. The obvious counter argument to this approach is that a fixed discount rate mortgages the future of current employees for current pensioners (anecdotally, the people saying this generally have got DB pensions). In fact that is already the case – current DB beneficiaries have (and continue) to take disproportionate benefit relative to the younger workforce (typically DC). A flat discounting approach would redress that today, and provide a sustainable way of funding DB going forward. In an optimistic scenario we could reverse the trend of DB closures, and the future accrual process should greatly diminish the major remaining risk of path dependency of returns during negative cashflows.
All told, a flat discount rate could drive almost £1tn of impact in the UK private sector (and further afield translated through multinationals), not to mention the impact on the public purse through additional tax receipts, as well as the recalculation of the public pensions liability itself as well. It would make the best case cost savings from the pooling of local authority pensions look like chicken feed in comparison, and drive even more market activity than quantitative easing was able to achieve.
It might even make DB affordable again for the rest of us.
It is the greatest good to the greatest number of people which is the measure of right and wrong. ~ Bentham
The views and opinions expressed here are solely those of the authors in their private capacity and do not express the views or opinions of Kempen Capital Management.