In the UK in 2014, The Independent Review of Retirement Income (IRRI) was commissioned to look at retirement incomes. Two recommendations from IRRI were:

“The use of deterministic projections of the returns on products should be banned.”

“They should be replaced with stochastic projections that take into account important real-world issues, such as sequence-of-returns risk (and) inflation.”

Quite a bit to digest. There is a broader discussion of these issues in a previous article written by David Bell.

**What is deterministic forecasting?**

In retirement projections, deterministic forecasting is a set of fixed assumptions around investment returns and inflation to produce one scenario to establish whether a retiree has sufficient financial capital.

An example of a deterministic forecast is the superannuation balances required to achieve a comfortable retirement as calculated by The Association of Superannuation Funds of Australia (ASFA). The ASFA Retirement Standard was developed to objectively outline the annual budget needed by the average Australian to fund a lifestyle in their post-work years, providing benchmarks for both a comfortable and modest standard of living.

ASFA details that:

* “a comfortable retirement lifestyle enables an older, healthy retiree to be involved in a broad range of leisure and recreational activities and to have a good standard of living through the purchase of such things as: household goods, private health insurance, a reasonable car, good clothes, a range of electronic equipment, and domestic and occasionally international holiday travel”.*

As of March 2018, for a retired couple, the budget for a comfortable lifestyle was $60,264 per annum. Based on a rate of return of 6.0% per annum and inflation of 2.75% per annum, ASFA has determined that a sum of **$640,000** is required for retirement. This method draws down capital over the period of withdrawal so that nothing is left at the end of an average life expectancy.

**Stochastic modelling introduces variability and stress testing**

A stochastic model considers different outcomes by allowing for variation in the inputs within the forecast.

For example, the Accurium Retirement Healthcheck is a projection tool that allows the assessment of retirement sustainability. It stress tests a retiree’s plans through 2,000 possible future scenarios. The investment return assumptions are provided by Willis Towers Watson and are generated using their Global Asset Model. This Model produces ‘random’ future sequences of possible investment returns for each asset class. These are generated so that ‘as a whole’ the simulations represent a full distribution for how ‘real world’ markets could perform in the future.

Stochastic modelling assists in making a scientific assessment of sequencing risk, which is the risk of experiencing poor investment returns at the wrong time. Stochastic modelling can’t predict the nature of ‘Black Swan’ events. Is the chance of a significant crash 1% or 10%? These models can’t estimate these odds, so the best that can be said is that these events don’t happen very often.

**Modelling a ‘comfortable’ retirement**

It is an interesting exercise to model the ASFA example in the Healthcheck software for a two 66-year-olds in a couple couple who are eligible for the age pension.

Let’s assume $640,000 in superannuation in a ‘balanced’ portfolio using the asset allocation constructed to align with the Morningstar Multisector Balanced Market Index (22% Australian shares, 25% international shares, 5% listed property, 33% fixed interest and 15% cash). Fees of 0.9% per annum are assumed. The default long-term asset class return assumptions within the Healthcheck produce a return of 5.3% per annum based on the asset allocation after the deduction of fees. Gross returns of 3.5% for ‘cash’ and 4.5% for fixed interest are assumed, which could be considered optimistic in the current environment. Nevertheless, the 5.3% per annum is lower than the ASFA assumption of 6.0% per annum.

The Healthcheck also allows settings around the lifestyle of a retiree. ASFA produces budgets for those around 65-years-old and 85-years-old. For those around 85-years-old, the comfortable lifestyle budget is $56,295 per annum, which is approximately 7% lower than that for a 65-year-old. So in the Healthcheck, it has been assumed that expenses reduce to this level at age 85. ASFA also produces budgets for singles and the comfortable lifestyle budget is $42,764 for a single, which is approximately 29% lower than for a couple. Again, in the Healthcheck, it is assumed that expenses reduce to this level at the first death. Personal effects of $25,000 have been assumed for age pension calculations.

The chart below illustrates the misleading nature of a deterministic forecast. It shows that the retirees’ money in their account-based pension (the blue bars) should not run out until they are 98-years-old. The dotted line shows where there is a 50% chance that at least one member of the couple will still be alive, and that’s in 26 years’ time.

**What if a greater range of outcomes is considered?**

However, the stochastic modelling produced via the Healthcheck shows that in 64% of the 2,000 scenarios tested, the retirement lifestyle is sustainable. This means that the couple has a 36% chance of **outliving** their savings.

I’m not sure what readers think about a 36% failure rate, but I wouldn’t feel comfortable crossing a bridge if it had a 36% chance of collapsing!

The chart below details where variability has been allowed for. There is an 80% chance that the retiree’s future savings will fall within the blue shaded area. The bottom of the blue range represents a ‘worst case’ outcome at each age. There is a 10% chance of running out of money after 22 years. The top of the blue range represents a ‘best case’ outcome at each age. The green line represents the median of 2,000 scenarios.

When presenting the deterministic forecast, I don’t think anyone could reasonably imagine that there was a 10% chance of running of money after 22 years. The deterministic forecast illustrates that the retirees should be fine until they are aged 98, which is 32 years away.

**So how much does a retiree need to be ‘safe and comfortable’ with 95% confidence?**

Let’s assume that we want 95% confidence that the retiree would not run out of money. I am risk averse, so I still wouldn’t walk across a bridge if it had a 5% chance of collapsing. Nonetheless, it is better than walking across a bridge that has a 36% chance of collapsing.

By reverse engineering within the Healthcheck software, based on the previously detailed assumptions, it is estimated that a retiree couple would need **$1,036,000** to have only a 5% chance of running out of money. This is 62% greater than the amount estimated by ASFA. However, it is lower than that implied by some safe withdrawal rate articles that suggest drawdowns of not more than 4% ($1,506,600). This is due to the benefits of the age pension, but if social security laws change significantly in the future, as they have in recent times, then this would impact the results produced by the model.

The chart below is based on $1,036,000 in superannuation at retirement. There is a 10% chance of savings being approximately $40,000 in 29 years’ time and a 10% chance of being roughly $460,000.

There is no doubt that a deterministic forecast is easier to explain, easier to understand and still has its place. However, stochastic modelling, while more complicated, considers a vast range of possible scenarios and estimates the most significant financial risk for retirees, which is the likelihood of running out of money.

As David Bell succinctly notes in his article mentioned earlier:

“Many other industries develop complex products which are explained effectively to consumers … Too hard … cannot be an excuse.”

*Patrick Malcolm is a Senior Partner and Financial Planner at GFM Wealth Advisory. He has attained the Certified Financial Planner ^{®} designation, completed a Master of Applied Finance and is also a SMSF Specialist Advisor™. This article is general information that does not consider the circumstances of any individual.*

David, I came to this site looking to see if anyone else thought the AFSA “comfortable” wasn’t all that comfortable. I’ve never thought of myself as a spendthrift, but to maintain my current lifestyle, I will need way more than the $43K that ASFA says will cover me as a single. The AFSA model leaves a lot of stuff out e.g. major renovations like the need to replace a bathroom or a kitchen; the need to replace a car; pets are not accounted for (and many of we oldies love our pets); donations to churches or charities (60% of Australians identify with a religion per census figures); gifts to family and friends are not mentioned; the amount allocated to overseas travel is a pittance – perhaps they mean across the seas to Tassie. All up with so many things left out, the AFSA guidelines are not all that helpful.

The diagrams above assume that spending levels are static (or at least reduce by a fixed percentage when we get to 85). There are other withdrawal techniques which modify spending levels based on Super performance. E.g. if your funds decline faster than expected then reduce your spending a little. Using these techniques, you never run out of money.

David or Patrick, to your knowledge, has anyone done a calculation on how much is required in Australian dollars to retire “comfortably” (as in with local health insurance etc.) in Thailand or Bali or the Phillipines? Would the age pension be enough?

Assumptions are always a concern when creating future scenarios. Projecting future returns on growth assets over defined periods is some form of science or art that never quite seems to work out. However, projecting returns on cash and fixed interest particularly when it comes to individual clients scenario should be a far more predictable task. There is no chance of getting 3.5% p.a. on cash, now or any time soon. The chances of getting 4.5% p.a. on fixed interest with high security arn’t likely either. In fact both of these assumptions are about 50% out, currently and over the past 5 years.

Is it prudent to project interest rate returns based on long-term historic returns, when the interest rates declined from record highs to current, record lows?

We know what happens to fixed interest when rates start to rise.

A lot of small business owners will tell you that even if 100% of your after tax income is invested in retirement savings you still won’t have enough money to retire – let alone meet the incomes shown in the ASFA retirement standards.

The ASFA deterministic approach might be restated as you probably need $640,000, but could fall short. Patrick’s approach is that you’ll need more than $1m to get only a 5% chance of falling short.

Neither are helpful when you flip them:

– you have a 50% chance of falling short with $640,000 – (ASFA)

– you have a 95% chance of selling yourself short with Patrick – you’re most likely grossly underspending in your active retirement years.

(I’m particularly uncomfortable with the bridge analogy for this reason, life and wealth planning is not like that at all. We humans are perfectly able to plan logically then respond and adapt as things unfold. I’m just as uncomfortable with planning for the 95% confidence outcome without acknowledging the much more likely outcomes)

So the guidance industry practitioners give should lie in middle.

Accepting that all the fancy stochastic modelling never really anticipates real world outcomes (I couldn’t agree more with you Ashley), the likes of Tim Farrelly and Michel Kitces suggest planning to an “eyes wide open” confidence level of perhaps 70%, revisiting every few years and whenever there’s a big change with the client or the market.

So for me, I’d like the tools and planning conversations to acknowledge risk but not go overboard. Practitioners can set a spending budget to a confidence level of around 70%, and acknowledge it’s subject to adjustment at the next review.

Come review time there’s a realistic expectation the adjustment is likely quite small and perhaps twice as likely to be pleasant rather than unpleasant. This lets the client get on with their life with some very realistic confidence. Does that sound like a plan?

This shows how ineffective it is for individuals to deal with the investment risk and longevity risk in an account-based pension.

– If you are risk adverse and you have a target income you want to drawdown, you need to save a lot more than what the deterministic projection shows.

– Otherwise, you have more chance of not having enough or having too much than having “just enough”.

This is where insurance-based/collective pooling-based solutions may help to form part of retirement income solutions. The value of protection from adverse event needs to be recognised and solutions should not always be compared in an “investment” lens.

There are three kinds of lies: lies, damned lies, and statistics.”

Forecasts are exactly that – a forecast – or more accurately, an opinion about the future – whether it is based on data or just “the vibe”! We need to remember that whatever they are, they are not self fulfilling prophesies! Plan accordingly!

The ability by experts to complicate a simple problem is mind numbing. What you “need” in retirement is directly related to what you personally spend not some academic number – it is personal, very personal. Take your expenses, multiply by 20 and that is the capital you need to retire around 60 with a wrap around the arm!! As robust as all the nonsense churned out in the models

I agree. How much you need for a retirement depends on how much you want to spend each year to deliver your lifestyle expectation over 40 years from age 60 to 100. Spending will be higher in the years 60 to say 75 and will hopefully and likely be less after 75/80 as we travel less and personal needs are less demanding.

The secret to self-sufficient or partially self-sufficient retirement planning then is knowing how much you spend on what each year over 5-10 years prior to retirement. It really gets serious about knowing this information around age 45-50. If we can teach our population to track what they spend on what over time, individuals will be in an informed position to engage in planning requirements for retirement. They will develop an innate understanding of discretionary and non- discretionary spending, saving, investing, income and expenditure.

On this theme it is heartening to see banks are now at last raising the bar regarding loans and requirements for applicants to know and understand and provide evidence of their expenses and income.

Ashley makes some timely observations about the limits of risk analysis models in the above comments, but still think this is a great article that opens up the broader conversation around retirement saving adequacy. Suspect if this aspect of financial management was more of a focus in the community, attitudes to risk would be more realistic and the value of wise financial counsel would become implicit. More importantly, less photos of couples their 60’s walking down a beach would be required! (I jest…) Thanks again Patrick.

…… so what is the answer then Ashley ?

Short term falls get the attention but the long term lulls don’t.

International shares July 2000 to April 2014 – 0.1% return.

http://insights.vanguard.com.au/VolatilityIndexChart/ui/retail.html

SO perhaps the 4% “rule” remains a decent heuristic, with a bit of margin if safety embedded.

‘Stochastic’ means ‘random’ and the approach assumes all of the parameters are static, stable and constant. But nothing in the world is static, stable or constant and nothing in the financial world is normally distributed.

So we should not have a false sense of security in stochastic models.

Stochastic models grossly underestimate downside risk. Eg the All Ords fall on 20 October 1987 was a 34 standard deviation event, which according to the stochastic theory was only supposed to happen once in every 10 to the power of 243 ‘universes’, (where 1 ‘universe’ is 14 billion years).

The flash crash in the 12 minutes from 2:32 and 2:45pm on 6 May 2010 was supposed to be even rarer, but it happened. As robots take on more and more functions, they will probably be more frequent.

But even a monthly fall of say 15% (a 4 standard deviation event) is supposed to occur only once in every 2,600 years, but we’ve had 4 of those in the past 50 years.

Or an annual fall of 43% (a 4 standard deviation event) is supposed to occur only once in every 31,000 years, but we’ve had 4 of those in the past 100 years.

Etc, etc

I’ve long told anyone I know personally, who’s interested in retirement finances, to be very wary about the comfortable / modest retirement standards, and the assumed lump sums attached, for two reasons: (1) I don’t think the “comfortable” standard is particularly comfortable and (2) they ignore the huge regulatory risk of changes to the nature of the OAP in the future.

Having recently dived into all the various SWR calculations and theories, out of personal interest, I’m not surprised to see the conclusions above about the potential failure rate.