This is Stuart Fowler‘s first post on this blog. I am very pleased he is sharing this essay which focuses on the difficult challenge of investing for a pension.
When I met the Chairman of the FCA I asked him if he could define good a good financial advisor, he replied “Stuart Fowler”.
In this fine essay, Stuart argues that if people engage rationally with their experience of the past, technology can help them make the right choices for the future.
Final-salary pension transfers tell us a lot about the wider need for customised investment solutions, goal-based and outcomes-driven, to replace standardised portfolio types.
Failure of conventional solutions
Defined Benefit transfers, where a known, secure real income is given up for a capital sum that may produce a higher income but may also fall short, have exposed a different kind of shortfall: the general lack of an investment solution that is able to quantify the possible difference in outcomes and so support trade-offs individuals might prefer to make. This shows up in the Financial Conduct Authority’s criticism of widespread failures in the transfer advice process.
This is not the first time the standard industry investment approach has failed because it could only handle trade-offs described only by generalised returns and risks, independent of time and not subject to individual constraints based on outcomes and their consequences. An earlier example is the widespread failure of with-profits mortgage endowments to pay off the debt liability in full at maturity.
What Defined Benefit (DB) pension transfer choices and mortgage endowments have in common that were not addressed by general industry approaches were:
- defined outcomes
- at one or many specified dates
- with clearly identifiable personal consequences of either exceeding or falling short of the amounts needed at the time required.
Here’s how one of Europe’s top business schools for finance described the shortfall in industry practice:
“Dealing with a private client usually leads to a detailed analysis of the client’s objectives, constraints and risk-aversion parameters, sometimes on the basis of rather sophisticated approaches.
Yet it is striking that, once this information has been collected, and sometimes formalised, very little is done in terms of customising a portfolio solution to the specific needs of the client.
In general, the approach consists of providing several profiles, expressed in terms of volatility; in some instances a distinction in how the capital will eventually be accessed (annuities or lump-sum payment) is made, but the client’s specific objectives, constraints and associated risk factors are simply not taken into account in the design of the optimal allocation.”
Edhec Risk Institute; Asset-Liability Management in Private Wealth Management; September 2009
I think almost any customer of the retail investment industry would recognise this disconnect between the definition of the problem and the design of the solution.
In their separate survey of industry practice across Europe a little later, Edhec identified that most retail investment firms had as an objective moving to ‘liability-driven’ techniques, mirroring what (in many cases) their own firms had had to do as managers of occupational pension schemes.
Ten years on, we can note that the mass standardisation of ‘providing several profiles, expressed in terms of volatility’ is still holding out against customisation. The reason is the higher cost. Even the advances of digital advice delivery, unhappily known as ‘roboadvice’, have not yet broken the mould, largely because they have sought to replicate standardised solution instead of developing techniques for low-cost mass customisation.
The largest single application of goal-based and defined-outcomes investing that is crying out for a solution is ‘drawdown’: the process of turning capital into a stream of consumable cash payments.
This is the general spending-goal problem of which DB transfers are simply a special case. Both have defined outcomes (real spending) at specified dates (years of spending) and subject to constraints ranging between the general (not running out of capital before a prudent assumed age of death) and the particular (at a very high level of confidence, not falling short of core spending needs or minimum tolerable spending).
Where the DB has a defined outcome (a pension income for life with a calculated present value, in today’s money) and a known capital amount (the cash equivalent transfer value offered by the employer today to enable you to buy out their liability to deliver that income), a general drawdown problem might also have a known resource (today’s total savings and investments available to fund the drawdown plan).
The problem requiring solving is then the risk-dependent outcomes, in today’s money. In the case of a transfer the modelled outcomes can be compared with the gross pension income being given up. But that is only the same as saying, in a general drawdown case, that one of the risk-dependent outcomes is the income produced by a purchased annuity with inflation protection: the risk-free option.
The drawdown problem in many cases, particularly when still saving, will be defined as solving for the resources required, jointly with a specified risk approach, to meet desired or required outcomes. That includes required additional contributions to the plan. It could be to work out when no further contributions are required and retirement is economically optimal.
If both available resources and the minimum required outcome are known (though perhaps now dependent on an assumption about the tax suffered that equates gross ‘income’ to net spending), then the risk approach can itself be solved for, based on maximising spending upside consistent with meeting the required minimum outcome at a desired level of confidence. Meeting core spending needs, and not running out of money by some prudent late age, are likely to be set at as high a level of confidence as is reasonably possible. In terms of formal, mathematical models for solving such a problem, that is arguably as high as 99%.
What we have described here is a balance between the planning variables that a workable or fully-funded plan must show. The variables are four:
- RESOURCES assigned to the plan (today’s capital and any planned new savings)
- TIME horizons from start of draw to the earliest tolerable date of exhaustion
- RISK approach (the approach being constant but the level varying with time)
- OUTCOMES as gross or after-tax cash flows, in real terms, for each year of the plan
Though drawdown is often seen as a problem specific to personal pensions, as a choice created by the Government’s 2015 ‘Pension Freedoms’ legislation, it is in fact by nature holistic, embracing all wealth and income.
Any resources can be assigned to a drawdown plan, both actual (such as ISAs and taxable investment accounts as well as personal pensions) and contingent (such as inheritance, trading down or a lifetime mortgage).
Any plan using these resources is also supported (underpinned, even) by any guaranteed inflation-protected incomes, notably the State Pension but also any retained (not transferred) Defined Benefit pension (although the inflation protection is not in any cases as complete as for the State Pension).
Solving for the right balance for each individual calls for modelling of the mathematical relationships between the planning variables. Only if the modelling is ‘probabilistic’ (technically known as ‘stochastic’) can probabilities other than 50% (think ‘average’) be used to plan the best possible combination of the variables.
Introducing odds makes trade-offs easier to relate to. Setting the required certainty level means the stress tests, or resistance of the plan to extreme risks, are more likely to inspire confidence in the robustness of the plan.
None of this removes the need for what conventional portfolio approaches rely on: can you live with the expected volatility of the portfolio?
If the plan that maximises satisfaction with the possible spending outcomes (meeting the minimum but providing scope for higher spending or gifting that the individual also values) introduces a high level of equity exposure, then these outcomes will only be obtained if the plan owner can stick with it when the equity component is performing badly.
One difference is that the portfolio solution may be structured so that equity exposure reduces as the remaining plan duration shortens, so that the annual volatility and peak to trough ‘paper losses’ will tend to be higher initially than later in the plan. Acceptance of volatility may itself be greater initially than at an older age.
Volatility tolerance should never be tested in abstract. How risk is experienced is likely to be different as a function of the presence of a plan because of greater clarity about expectations and the focus on a vision of the outcome motivating the risk taking.
How would a drawdown solution have helped in the past?
In the case of a mortgage endowment, an outcomes-driven modelled approach would have calculated the monthly savings or contribution rate (described in this case as a ‘premium’) that would meet in full the nominal mortgage liability maturing in year x at a high level of confidence.
At any level of premium below that required, it would allow the potential shortfall at different confidence levels to be estimated. These are not numbers the life insurance companies running with-profits funds wanted to tell us, because they knew full well that funding the liability at greater than 50% confidence would have cost more in monthly contributions than the payment to service a conventional repayment mortgage.
It needed a lower outlay to sell it. So what the modelling would have done is prevent a bad product reaching the market.
In the case of Defined Benefit transfers, the use of a stochastic cash-flow model would have allowed (and in our case did allow) the potential gain in individual welfare to be tested and demonstrated by a record of the dialogue-based interaction with the model.
Models support iteration, or trial and error, to arrive at the optimal choice based on maximising satisfaction.
The FCA prescribes the transfer advice process to a level that is unheard of in other areas of advice. These detailed rules have been developed by them to address failings in typical transfer advice that we attribute to standardised investment solutions: it isn’t holistic in reach; it doesn’t consider how the client’s best interests could be met by other changes than just a transfer, such as insurance or altering the risk approach for the free assets; it doesn’t consider the personal consequences of both shortfall and surplus; it doesn’t address the dependence on time; it relies on mean expected (that 50% probability again) rather than stress-tested possible outcomes.
The FCA did not get here in one go: it has had several attempts at prescribing the advice approach, hampered by not first defining the problem in terms of accepted investment theory and not understanding, or wanting to play down, the economic origins of the potential gain in welfare. It started by assuming that what economists call a ‘safety-first’ utility function does and should apply to retirement provision: don’t take any risk until your minimum needs have been fully funded with total security.
That ignored the inconvenient fact that without taking risk we may never be able or willing to fully fund our needs. And that was before real interest rates were negative! The idea that the right solution to retirement provision is a deferred annuity, risk free at every stage of accumulation and in retirement, went out in the 1950s and 1960s and is not now evidenced by a single workplace-pension default option.
The FCA’s assumed definition of utility effectively denied advisers the opportunity to make trade-offs, such as increasing the probability of achieving more than the minimum required but introducing a smaller probability of shortfall. In its latest ‘guidance’ paper, setting out how to interpret the rules, it seems finally to accept that the transfer choice is by nature one involving trade-offs, but subject to idiosyncratic constraints that must be identified in every case.
It also wants the client to be able to recognise themselves in the report they are given, which they would if it emphasised the idiosyncratic rather than the general, and if it was dialogue-based rather than relying on a template. If the process and report follow the rules and guidance, any knowledgeable reader of the report would be able to describe the actual utility function, even if the report doesn’t do so in so many words, based on what the client has revealed about themselves in the advice process.
The future for drawdown
An outcomes-driven solution should serve a number of different purposes:
- Specific cases like a DB transfer
- Any retirement plan that defers or avoids an annuity
- Other spending goals requiring the same answers, such as living off a divorce settlement or injury settlement
- Investment mandates involving draw (including ‘natural yield’) but subject to constraints of sustainability or equity, such as life-interest trusts, charitable trusts and endowments.
Of these, the mass-market requirement is clearly for retirement spending: delivering the ‘freedom and choice’ envisioned by the 2015 legislation by making it easy for individuals to decide how much risk to take and how much to draw.
This requires quantification of spending outcomes and quoting odds, so that people can make choices informed by something they can relate to: the consequences of higher or lower spending, based on their pinch points and their motivation.
It requires the ability to visualise versions of your future, perhaps even to imagine yourself in the future looking back at versions of your past to see what satisfaction and regret look and feel like.
Even if it requires no special knowledge of maths or finance, success does depend on encouraging people to engage with the realities, because of the economic constraints on what is possible in financial markets. In our experience, elements of gaming, such as making a plan balance or maximising a key number, can help that. That itself requires some kind of iterative interaction with an engine via a user interface, with or without human guidance.
The investment solutions are the least important to the customer. They could even be wrapped up inside a product rather than a service. The cost matters but they do not need to know what is going on under the bonnet to engineer the defined outcomes.
All goal-based investment is a journey but with drawdown, where the draw and spending rate is a constant reminder of what this is all about, it is critical the investment solution supports a fairly frequent re-optimisation of decisions by returning to a familiar framework.
Clearly then, mass-market success for a drawdown solution needs a combination of elements.
- Investment technology
- Interface design
- Delivery technology
- Branding and marketing
Focusing as we do on high-end clients, we are not the future. But with so much experience of making technology the basis of people’s relationships with money, we are a pointer to the future for the mass market.