The post-retirement sector is large and growing rapidly, and financial advice processes need to be more sophisticated to deal with potential problems such as sequencing risk and longevity risk. What I will call ‘Post-Retirement Optimised Portfolios (PROPs), based on stochastic modelling, should be part of the solution.
In the March 8, 2013 issue of Cuffelinks, I wrote, “Growth in the post-retirement population over the next 20 years is mind-boggling, with different five year age groups (e.g. 65-70, 70-75, etc.) growing between 60% and 100% between 2010 and 2030. There will be major growth in the age 65-70 segment over the next five years as the first wave of ‘boomers’ reaches retirement. We need much greater sophistication in post-retirement advice and investment solutions, including the use of investment scenarios or stochastic modelling in building optimised portfolios. This poses particular challenges for product and advice providers.”
Based on the latest Deloitte Superannuation Report, Australia’s post-retirement sector is forecast to grow from $333 billion as at 30 June 2013 to $1.4 trillion in 2033. Over $800 billion of the 2033 assets is forecast to be in SMSFs as the baby boom bulge transition into post-retirement.
Changing the post-retirement emphasis
Let’s start by recognising the post-retirement issues we want to address:
- sequencing and longevity risks are acknowledged, but without solutions
- debate is too product-orientated
- deterministic rather than stochastic planning.
There is much debate about conservative versus growth-oriented asset allocations for retirees in the post- retirement phase. This often rests on assumptions regarding the equity risk premium, and the arguments put forward tend to be too simplistic.
There is now much greater discussion regarding sequencing risk and longevity risk, which is encouraging. However, we still lack sufficient advice solutions and portfolio construction tools to deal with these risks.
Too often the debate about post-retirement solutions revolves around product – for example, the relative merits and demerits of annuities and deferred annuities – rather than a focus on advice solutions and portfolio construction, with products only being considered following that process.
Retirees with less than say $250,000 in financial assets (including super) at retirement will have a significant reliance on the age pension. This largely addresses sequencing and longevity risks for this group. Some retirees will also rely on scoped advice or simple retirement calculators, with all their shortcomings.
People with over $3 million in financial assets, including super, at retirement should also be relatively well-placed to weather sequencing and longevity risks, providing of course they do not practice an overly lavish lifestyle.
This article is primarily directed at those with financial assets at retirement in the $250,00 to $3 million range, and those receiving comprehensive financial advice. A common approach to post- retirement financial advice for these people is risk profiling, cash flow modelling, high level asset allocation, portfolio construction, and product selection.
My concern is that existing processes for cash flow modelling tend to be based on deterministic approaches. That is, determining a fixed expectation of investment returns corresponding to the asset allocation, and conducting future cash flow modelling based on this expected return.
Volatility of returns and the impact of the order of returns (sequencing risk) tend not to be well incorporated. Longevity risk tends to be dealt with in a rudimentary way by assuming living to average life expectancy, sometimes complemented by basic scenario analysis where we assess ‘how long the money will last’ at various ages of death.
The problem is that deterministic models only consider an expected outcome and do not consider the range of possible outcomes that retirees may experience. This is valuable information for the planner to consider, to communicate with their client and to be used as a basis for plan design.
On average, deterministic predictions will have approximately a 50% success rate, meaning that there is a 50% likelihood that a retiree would have experienced their expected planned retirement outcome, with some left over as an estate, and a 50% likelihood that the retiree would exhaust their retirement accumulation and have to accept a lower level of consumption in retirement. Is a ‘50% right’ plan good enough? Even with cash or contingency reserves of up to twice the targeted annual income draw to deal with sequencing risks, the percentage likelihood of a deterministic based plan being ‘right’ will only be marginally improved.
A proposed approach
We need a stochastic approach rather than a deterministic approach.
A stochastic approach simply considers variability in outcomes by allowing for variation in the inputs which go into the forecasting model. This can be done using mathematical techniques where one derives a distribution of outcomes. The other, more common approach, uses simulation techniques, producing many simulations which are aggregated into a picture of the range of outcomes.
A stochastic approach uses many different sources of variability which affect outcomes. Examples include returns, risks and correlations between asset classes, and factors which affect mortality outcomes (systemic piece (longevity risk) and the idiosyncratic piece (that individuals may experience a different outcome to the population expectation). The amount of detail can be significant when one incorporates products (allocated pensions, life annuities, variable annuities, reverse mortgages), direct investments in the various key asset categories, and the age pension (with means testing and indexation) and various taxes.
The broad approach to the construction of a Post-Retirement Optimised Portfolio (PROP) would be as follows:
- Retiree client to force rank their retirement objectives (see below).
- Input key information into the modelling engine, including current assets and allocation, targeted income and capital cash flow requirements, risk profile, age, force rankings, etc.
- Perform stochastic runs to produce the many simulations (typically thousands), using random number generators to model many different outcomes, and difference sequences of returns.
- Review the results of the stochastic model, including the likelihood of meeting the various force ranked post retirement objectives.
- Adjust inputs to better match achievement of the highest ranked post retirement objectives.
The three key post-retirement objectives are:
- a capital or estate goal – i.e. to have a certain amount of capital preserved at various future ages, either to cover aged care costs, health costs, or targeted estates for beneficiaries
- income goals
- goals regarding tolerance for variability in income.
Clients will need to rank what is most important to them.
A key output from the simulations is the construction of an optimised portfolio with asset allocations (including direct investments) which seek to fulfil the ranked objectives. A simpler version is the consideration of say 4-10 alternative portfolio constructions, and using the simulations to assess which constitutes a ‘best fit’ with the ranked objectives.
The outputs from the stochastic modelling will show the full range and extremities of future financial asset holdings and income levels, allowing for different sequences of returns and the risk of outliving the assets.
An optimised portfolio would show the probability of fulfilling the various objectives. For example, it may show that a retiree has an 85% chance of achieving the capital or estate goal at the age of 87 based on an income of $55,000 per annum, indexed to inflation from retirement.
If this client has chosen the capital goal as their most important objective, they may find this outcome unacceptable. They can then recalibrate the inputs to produce a result which more closely fits their force ranked objectives. For example, this may require some increase in their appetite for variability of income. Or a small reduction in targeted income requirements, from $55,000 to $50,000, can significantly increase the likelihood of achieving the capital goal.
This approach enables well informed trade-off decisions. Typically this would be presented in a visual format to increase client understanding.
It is not unusual for such optimised portfolios to have a ‘bedrock’ source of income, which may generate 20-40% of post-retirement income. This could include social security, annuities or an equity release product. Such sources of income also have a low degree of correlation with returns from other financial investments. This PROP approach would also be able to demonstrate where deferred annuities may be appropriate.
Current developments and an appeal for action
Some entities are developing this capability, including Deloitte US, Mercer and Count Financial, and investment research house Lonsec and global actuarial firm Milliman. This paper is a high-level overview and I would love to hear from you if you have been pursuing similar endeavours. It can be refined for factors such as the different phases of post-retirement (active, passive and impaired) and funding future health and aged care costs.
The wealth industry needs to deliver superior post-retirement solutions, including:
- more research on this topic, such as the work by the Actuaries Institute’s Retirement Incomes Research Group
- greater innovation and commercial initiative, as mentioned above
- large wealth management institutions and advice firms embracing these more sophisticated post retirement approaches
- advisers encouraging their licensee to pursue this more sophisticated approach.
Hopefully, as these approaches are adopted, it will be possible to integrate such modelling with mainstream financial planning software platforms.
Andrew Gale is co-owner and Executive Director at Chase Corporate Advisory and a board director for the SMSF Professionals Association of Australia (SPAA). The views expressed in this article are personal views and are not made on behalf of either Chase Corporate Advisory or SPAA.