CDC at HOC: there are none so blind...
For two days last week the The House of Commons Work and Pensions Select Committee heard evidence on the hybrid pension arrangement called Collective Defined Contribution or CDC. They saw what they needed to see but heard only what they wanted to hear. I hope this isn't how all select committees work.
The prompting for this hearing was the decision by just one DB scheme so far, Royal Mail, to express an intention, supported by its union, to take advantage of the 2015 Pensions Act that allowed in general terms for a 'hybrid' scheme to exist under the law. Previous legislation allowed only for Defined Benefit (DB) or Money Purchase (Defined Contribution or DC) schemes and nothing in between - which might instead be termed Defined Ambition - with characteristics of both. As Chairman of the Committee, Frank Field clearly signalled that he thought some form of CDC hybrid was the best option to replace closing DB schemes and that the Government ought to be prodded into facilitating what Royal mail wanted to do. The Committee gave him the platform to do so.
Any prior bias was less evident amongst the Committee members and we don't know what discussion went on behind closed doors; so watching the televised proceedings does not necessarily reveal what the Committee will in fact conclude, let alone whether the Government will be persuaded by it to seek Parliamentary time if that is what is required to facilitate in practice, as well as in general, any new hybrid scheme. The Minister and his senior civil servant appear a lot more circumspect than the Chair.
Whilst the Committee could have performed a useful role in making the outcome of that listening process open and transparent, without which a negative attitude by the Government to the Royal Mail's plan might otherwise trigger resentment, this is not how it panned out.
Are the benefits worth the costs?
The best starting point is the written evidence submitted to the Committee, which ought to have framed the members' thinking and their questions. This is what I mean by 'they saw what they needed to see', as they were definitely shown it. There were two sets of evidence that covered the ground well. The legal challenges and possible solutions were spelt out in a paper by former Slaughter and May partner, Philip Bennett. This evidence served to point to whether the legal, governance and accounting issues around the new hybrid were trivial or non-trivial. The other important evidence addressed the possible advantages in terms of pension outcomes. A quite readable paper for lay purposes was submitted by the actuarial firm, Willis Towers Watson. This addressed the different sources of potential gain in outcomes, the extent to which they depended on CDC (or could be obtained independently) and the size of the technical challenges associated with those benefits that depended on mathematical solutions. Taken together, the Committee could therefore have chosen to frame its discussion in terms that were useful to Parliament: are the benefits worth the cost? That could also have been used to challenge both sets of witnesses: those who were in favour and those against. This is not what happened.
For a start, there were no witnesses called who had publicly stated the position that the outcome benefits were either ones that could be obtained independently, without CDC, or were ones that could not reliably be delivered technically. This is the position Fowler Drew adopted in 2015, based on our modelling of investment outcomes and our knowledge of other product-based attempts to share or modify investment risks. We posted this blog and there was also an exchange of letters in the FT at that time based on our challenging the modelling done by others, notably the pension consultants Aon Hewitt. We know that we were not alone, either amongst people taking a mathematical approach or amongst those steeped in practical knowledge of the products that had been tried and failed, most notably insurance-company 'with-profits' policies. So opponents could have been called. The nearest to opposition amongst the witnesses called was investment platform provider Hargreaves Lansdown who chose to focus on one perfectly valid aspect of the debate: paternalism versus personal freedom and responsibility. They are big supporters (and also beneficiaries) of the 2015 Act that came down on the side of personal freedom and removed the 'enforced' (effectively if not in fact) purchase of an annuity, as the means of turning pension savings into a pension income stream at retirement.
Pension Freedoms have already done the job
The unsung hero of the Committee session was this Pension Freedoms legislation. Why unsung? The answer is because time after time witnesses explained that the main reason why CDC could deliver better outcomes than DC, as an alternative to DB, was that they replaced the annuity purchase by a form of drawdown. But of course drawdown under the Pension Freedoms itself replaces annuities. Put simply: you turn the pot of savings capital into an income stream by drawing down as much as you need to spend when you want it, subject to whatever constraints you choose, such as not running out of money. You retire, but your capital doesn't: it keeps on working (and earning) in financial markets. Yet time after time the Chairman kept referring to this benefit as one obtainable from CDC, as if drawdown was not the actual source of the benefit. Even when corrected somewhat half-heartedly by Kevin Wesbroom from Aon Hewitt (the source of the original modelling that explained why Frank Field could have been 30% better off had he had a CDC pension) the Chairman persisted in repeating the refrain. He even did it for the benefit of the Pensions Minister, Guy Opperman, when he came into the room later, having missed the exchanges on which the refrain was based.
In the absence of an annuity, the duration of the individual drawdown plan is much longer and comparable then with the duration of a pooled scheme. This is the source of the higher pension. The key task for the Committee was therefore to identify the benefits that were not due to extending the duration. Whilst there was discussion of the benefits of collectivisation, because it was not separated from but rather conflated with the investment risk benefits, it was not easy for the Committee to draw any inference about the scale of the collectivisation gains relative to the cost of the CDC structure.
Collectivisation gains not dependent on risk sharing
The collectivisation gains not dependent on risk sharing are twofold, they heard (rightly). Risk sharing being the tricky area (and I will come back to it), the others should have been easy to deal with.
First, collective pools benefit from economies of scale. But DC schemes, particularly group schemes or auto-enrolment schemes, also benefit from economies of scale and costs are actually falling under the influence of competition. We must therefore be talking about small orders of magnitude of additional gain that depend on say creating multi-industry giants - which is what the Dutch representative was able to explain, but Holland does not have NEST. The case for further scale benefits to weight against the public and private costs of enabling CDC schemes was not made.
The second benefit is longevity risk. This does differentiate the duration of the individual DC drawdown and hence returns. Prudence dictates an assumption the capital must last to say 95, whereas the average life of the post-retirement pool in a DB or CDC scheme might be say 87. That is a material difference in terms of the level of income in each case. However, a number of factors weight against the significance of this difference. Firstly, there is always an option to convert the drawdown pot into an annuity in the late stages of retirement when it does offer significant utility, so the longevity risk is not increased to that extent in every case. Secondly, the scale of the longevity risk is dwarfed by the income uncertainty arising from the investment return uncertainty at that stage. Thirdly, postponing an annuity offers a gain in cost savings. One of the messages the Committee should have heard loud and clear from several witnesses is that annuities are hugely inefficiently priced. (This can be deduced from deconstructing the underlying investment costs and product costs; the net gain may only provide value to an annuitant, relative to holding the same investments directly, beyond about age 95). Finally, the eventual annuity, at a later stage, may not need inflation protection, index linked annuity pricing being particularly inefficient, and may turn out to be bought on an impaired-life basis. I would expect the Committee, had it been properly informed, to conclude (counter intuitively) that pooling longevity risk is not a source of potential gain in utility that they should attach much significance to.
Does investment risk sharing work?
Had the Committee followed this line of reasoning in respect of the possible sources of benefit, as differentiated in the written evidence, it would now be left with just one critical question to focus on: does investment risk sharing work? It might also then have appreciated that if it doesn't work, there is no point spending time on whether the structure might be one in which risk is only shared between different cohorts of members or between the members and the sponsor. As both Slaughter and May and Willis Towers Watson explained, the legal and regulatory challenges, and consequent member trust, associated with each form of risk sharing are vastly different. It seems reasonable to suppose that this might also be reflected in the amount of time Parliament would need to give to each. But the difference is academic if there is no form of risk sharing that can be relied on to deliver 'fairer' outcomes than the market can. It is not enough to establish a framework that seeks to make outcomes fairer: it must not fail, otherwise the sense of unfairness at markets will be directed instead at politicians, trustees, actuaries and employers. This will prove far more corrosive than the lottery of markets, which at least has an 'unseen hand' behind it.
The Committee also needed to consider the time frame over which it thought risk sharing delivered benefits to society. The principle is a general one so this reduces to a simple question: can we predict the time-varying path of markets (or portfolios of different asset classes) over different lengths of time in the future (ranging from say one year to 20 years) with sufficient accuracy to be able to 'smooth' the return path that investors get to enjoy? If we could, then we could observe whether earned returns were above their own average and use the reserves created by withholding some of those returns to boost the payouts in periods of below-average returns.
Had the importance of the Pension Freedoms been properly explained, the Committee would have been able to deduce that the incremental gains available from smoothing short-term return differences once an annuity has been avoided are very modest. Wesbroom's spoken evidence referred to the extreme case of a five year difference in retirement dates (which I call short term) leading to a 50% difference in retirement income. But this exposure to a single window is a function of an annuity purchase, where the resulting income is affected not just by the short-term path of the assets individually but also, critically, by the variance in the relationship (or correlation) between equity prices and bond (or gilt) prices. If drawdown were thought of as a process of gradually retiring capital, over many different market conditions, this timing effect is diminished, exactly as it was when saving over a long period, likewise by averaging many different market conditions.
The logical conclusion is that the only lottery effects worth modifying are medium or long-term ones, affecting the outcomes for different cohorts or generations of retirees. This would have been a challenging insight of the Committee because many proponents of CDC, however informed, hold the intuitive view that the purpose is to cushion within, not between, generations. For instance, Janice Turner, Co-chair of the Association of Member Nominated Trustees, made a statement that intergenerational smoothing was not the aim of CDC. Nobody asked why not; or whether this was a point of moral principle or whether it was a technical issue, for example because it would be difficult to retain trust in the smoothing mechanism if, for long periods, smoothing was all one way, favouring the same generation. This is what successfully taming the longer term lottery requires but for most of the time it will be indistinguishable from an error of judgement about the direction of markets.
These generational differences persist even with drawdown, and even without the timing lottery associacted with annuities, because market returns diverge so much even over periods as long as 15 to 30 years. Deviations from trend themselves trend. Whilst this creates the potential for utility gain from smoothing the variances, the very fact that they arise, in spite of the 'mean reversion' observable over much longer periods, such as 50 to 100 years, should serve as a huge cautionary signal that trying to smooth returns over long periods is an act of defiance against the power of randomness in markets.
The relevant measure of returns when considering the impact on pension income is real total returns. This data incorporates two sources of variance: nominal returns and inflation. The longer the period, the more the real-return uncertainty is affected by inflation because this shows characteristics of unpredictable regime changes, from stable to erratic, from high to low or the reverse. Between changes, the regime can persist for some time. Since returns can be smoothed in any product or type of pension fund by diversifying equity risk between different countries, which also diversifies inflation risk (but via currency movements, which themselves vary), the market portfolio were are measuring real total returns for is a geographically mixed equity portfolio that any fund manager could assemble. We can ignore bond returns for this purpose as the portfolio providing the pension income can also be split between short-duration risk free investments funding the early liabilities (effectively a temporary annuity) and equities funding the long-term liabilities. Since the risk free assets perfectly hedge the liability, we can be indifferent to their variance and focus only on the equity risk that calls to be modified by smoothing.
Let's assume a model for predicting medium-term returns for this diversified risky-asset portfolio that incorporates the mean reversion central to the smoothing logic but also uses realistic history-based values for the error risk. This is the model built by Fowler Drew to plan and manage sustainable draw from personal capital, whether in or out of a pension account, to fund retirement spending. The size of the smoothing task can be represented by the parameter values we adopted for this model, based on observed data histories. The current mean expected real return from risky assets is 6.2% pa. The probability distribution (at two standard deviations) around this mean is as wide as -3.7 to 16.1% pa for a 10-year projection. Even at one standard deviation, which is far too weak for anyone wanting to avoid unfair outcomes resulting from their own smoothing errors, the range is 1.3 to11.1% pa. At 20 years, where 'more complete' mean reversion might be hoped to have validated the smoothing exercise, the range at two standard deviations is as wide as -0.9 to 13.1% pa. The idea that course adjustments to the optimal rate of draw could be confined to 2%, a level of cut to pensions paid which in the Netherlands was already large enough to provoke anger, is questionable if any value is to be obtained form smoothing.
So great is the scale and persistence of the of the medium-term variances that they may even alter the means to which we hope markets will in time revert. Taking an extreme case, the Japanese equity index in yen terms enjoyed a regression trend of 8% pa (real total returns) when fitted to data observed between 1957 and 1987. Fast forward another 20 years and that whole-history trend has halved to 4% pa. We cannot exclude the possibility that the same happens in another market or even that something similar happens in all or most markets at the same time. But we cannot predict that it will or will not. Such predictions are acts of defiance.
The fallacy of 'fair shares'
The last word goes to Wesbroom at Aon Hewitt. He was asked about a technical requirement of the right to transfer out of a CDC scheme. He explained that this could be resolved by actuaries calculating the 'fair share' of the underlying assets. Just to be clear, fair shares is a concept that assumes that the actuary defiantly knows the trend and where we are today in relation to that trend. He gave an example of the US market being widely viewed as 'expensive'. It is indeed widely viewed as such - and has been throughout much of the recent bull market. But relative to a regression trend fitted to a history of the S&P real total return index back to 1871, which has a slope of 6% pa, the current level of the index is just 13% above the trend; and for most of the time that experts were calling it overvalued it was in fact below its trend. Past ratios have reached extremes of 50% over or below trend before eventually starting a period of correction. Yet some corrections have started from levels similar to today. So it is really a guess whether the next 10-year returns will be above, below or close to the trend.
We have been here before, as Aon Hewitt well know. Fair shares was a fundamental principle behind the actuarial calculation of the smoothed asset-value path of an Equitable Life with-profits contract. Risk sharing, reappointing returns between different cohorts of participants according to actuarial assumptions, has been tried before. It failed then, because return paths did not behave predictably over the medium and long term. It was not an unfortunate accident that it failed. It was as predictable as the continuation of variances in returns that are present in a century of records for a number of markets. It does not even matter what the precise reasons were. It only mattered that it was an act of defiance that at some point markets would and did punish.
That failure led to recrimination and mistrust. It is the last sort of risk that employers today should be taking on, either by setting up schemes in which they guarantee bounded incomes (like Dutch CDC schemes) or in which the reappointment of returns is only between different members. And regulators too should be wary of permitting schemes where risk is shared with the employer. Given the data evidence of the risk of error in smoothing returns, this should rightly be seen as posing a nightmare for regulation and accounting. Since the nightmare of accounting for DB schemes is the driver for CDC, there should be no illusions on the part of the Government that the cost in time and effort to devise appropriate law and regulation can possibly be justified.
CDC introduces significant legal, regulatory and governance challenges, as was recognised by the 2015 Act and by the written submissions to the Select Committee. These challenges mainly arise because of the objective of delivering fairness of outcomes, as the only point is to modify the unfairness of markets, or the lottery of age. If the fairness objective cannot reliably be met, acrimony and mistrust that might otherwise be directed at anonymous markets, or perhaps at oneself, will be directed instead at identifiable groups of people and institutions. We have been here before and it is not a good idea.
The benefits for retirees that are available from CDC are generally overstated and that was the case in the Commons committee room last week. The main source of potential gain relative to DC is the increased income arising from not buying an annuity at retirement. This lengthens the duration of the cashflow liability from start to finish and, for an unchanged approach to risk in the scheme, allows a higher exposure to risky assets and therefore a higher expected return. The variance associated with that higher return is to some extent automatically smoothed by drawing at a set rate as markets fluctuate. This is what Pension Freedoms for DC already allows. Cost savings from scale in a collective pool are already available from DC, particularly after auto-enrolment, and competition is finally bringing down pension plan costs. Pooling longevity risk is widely assumed to be a significant source of gain but is trivial until the late stages of retirement relative to the scale of the real investment return uncertainty. Besides, when longevity risk becomes significant relative to investment risk it can still be managed by a partial or full annuitisation to increase expected income. The risk of unfairness or errors in modifying actual market returns to smooth pension payments between different cohorts or generations, which is a sources of variance that persists even with drawdown, is typically understated and this was the case in the Committee. It reflects misunderstandings about the scale of the uncertainty of real investment returns over medium- and long-term time periods, even though mean reversion is observable in the whole-history data and might be relied on as a systematic feature of public markets in capitalist economies.
The realistically-achievable advantages of CDC being so small and the risks of errors so great, it is difficult to justify the intuitive or casual presumption that the benefits must outweigh the costs. This is not the message the Committee heard but it may (hopefully) already be better understood in the Department of Work and Pensions. If the Government is looking to make a difference, it is either in the area of removing the causes of the death of DB or encouraging the industry to develop more robust techniques for advising on drawdown and managing the underlying portfolios to deliver smooth and sustainable incomes in retirement from a DC plan.