Sequencing Risk and Asset Allocation

The comparisons use both the Austmod historical data and simulated scenarios. 

What is sequencing risk?

Sequencing risk is the risk that your accumulated superannuation savings receive the worst returns at the worst possible time – the years around your retirement and, in particular, the years just after retirement when account balances are highest and cash flow is usually negative. 

Investment returns can vary widely from year-to-year and the timing of this volatility can make a big difference to how much you get to enjoy your retirement.

Many financial planners and retirement income experts and commentators suggest that an optimum retirement strategy should be based on a combination of:

(a) an account-based pension;

(b) a lifetime investment linked insured annuity;

(c) a lifetime “group self annuity” (GSA); or

(d) a guaranteed lifetime or term annuity; and if applicable

(e) the means-tested Government provided Age Pension.

However, a concern for many is that (a), (b) and (c) are subject to sequencing risk. Also, the cost of sequencing risk is built into the price of (d). Fortunately, the Age Pension provides some offset for part pensioners. The question we will now examine is, how do these sequencing risks vary by asset allocation?

Underlying data used

The tables below use the Austmod historical investment performance data described in Sections 3 and 4 of Australian Investment Performance 1959 to 2022 (and Investment Assumptions for Stochastic Models) – which I presented to the 2023 International Congress of Actuaries (ICA) in Sydney. 

This data is similar to the APRA Heatmap benchmarks but extends back to 30 June 1959 at quarterly intervals. Table 1 (and Table 6 in part) is based on a typical balanced portfolio with 70% growth assets. All results are before investment tax, after approximate imputation credits (for the full period) and after investment fees, and therefore should be reasonably realistic for pension products. 

What is “cumulative decline”?

With investment performance, “cumulative decline” is the percentage fall in the relevant unit price over a given period. It’s similar to the health “cumulative decline” which is what many, myself included, experience when they turn age 80. 

In the tables below:

• the “number of occurrences” is the number of times the unit price has fallen over  one or more integral years 
• the “average cumulative decline” is the average percentage fall in unit prices for all occurrences (since falls result from negative investment returns the “average cumulative decline” is recorded with a negative sign)

For example, in Table 1 there is one occurrence when balanced portfolio investment returns during the 65 years ending 31 December 2024 suffered a cumulative decline over six years. This was between 31 December 1968 and 31 December 1974 when the unit price fell 9.2% from 16.333 to 14.827.

The calculations do not allow for cash flow (and also the variability of unit prices during quarters may slightly increase the number of occurrences) but nevertheless the number and average of cumulative declines are realistic and reasonable indicators of relative sequencing risk outcomes for comparisons over different periods or with different asset allocations.

Table 1: Balanced portfolios

Table 1 considers periods ending 31 March, 30 June, 30 September and 31 December 2024.

In this table, 81% of the occurrences of yearly declines occur over one, two or three years. The highest average cumulative decline (of -14.0%) occurs over two years. Though not recorded in the table, the maximum cumulative decline (of -31.6%) also occurs over two years.

The average investment return over each of the four periods is almost identical. The September quarter-end is notorious for high volatility and had the highest standard deviation (of 12.4%).

Retirees would be pleased to note that only once did a cumulative decline occur over seven years and this was just -0.5%. Cumulative declines never occurred over longer than seven years.

Table 2: Australian shares

Table 2 considers the same periods as for Table 1.

In this table, 89% of the occurrences of yearly declines occur over one, two or three years. The highest average cumulative decline (of -24.0%) occurs over four years. Though not recorded in the table, the maximum cumulative decline (of -44.5%) occurs over two years, as for the balanced portfolio.

Again, the average investment return over each of the four periods is almost identical and the September quarter-end had the highest standard deviation (of 22.1%).

Retirees would probably be very pleased to note that with 100% Australian shares only twice did a cumulative decline occur over six years (shorter than the seven years for a balanced portfolio). Cumulative declines never occurred over longer than six years (again, shorter than for a balanced portfolio). 

These favourable results are mainly due to the negative auto-correlations for lags of both one and two years recorded in the last two lines of Table 2.

The auto-correlations in Tables 1 and 2 are determined directly from the Austmod database of annual forces at quarterly intervals. This is why the results in Tables 1 and 2 are so stable over time.

The prevalence of negative auto-correlations for Australian shares over lags of up to 5.5 years was established in my various ICA Papers on Australian investment performance. The significance of this was explained and explored in three recent articles by Cary Helenius and myself.

In the six tables in this article, auto-correlations have been recorded for only one and two year lags to reduce the complexity of the tables. Nevertheless, the full impact of Australian share negative auto-correlations appears to be a feature for lags up to 5.5 years. For lags of six years and beyond the auto-correlations are often significantly positive.

Table 3: Australian and international shares and fixed interest

In this table, all results vary considerably between the four investment classes. As expected, all the results for hedged international shares are closest to those for Australian shares. But it should be noted that even for hedged international shares, cumulative declines occur over up to nine years (compared with up to only six years for Australian shares). 

Also, with only one exception for each, both the number of occurrences and the average cumulative declines for hedged international shares exceed those for Australian shares.

Retirees should take note that for unhedged international shares, cumulative declines occurred over up to 15 years. Also, because of long-term currency cycles, with only one exception, both the number of occurrences and the average cumulative declines for unhedged international shares significantly exceeded those for both hedged international shares and Australian shares.

In Table 3, for Australian fixed interest, the average investment returns, standard deviations, number of occurrences and average cumulative declines are all significantly less than those for the three share asset classes. 

However, it is interesting to note that the number of occurrences of cumulative declines extends to six years, equal to that for Australian shares. 

Consistent with the two comparisons immediately above, it should be noted that positive auto-correlations are evident for both unhedged international shares and Australian fixed interest.

Table 4: Australian “All Ords Accumulation” over 125 Years

Table 4 spans 125 years, with 65-year periods commencing 1900, 1920, 1940 and 1960.

To further explore whether the negative auto-correlations and the above sequencing results are simply a feature of the past 65 years or whether they have been a feature of the Australian share market for longer periods of time, we sourced and updated a database produced by Marketindex.com.au. This database includes data for 122 years of December year-end Australian share market returns from December 1900. It was then updated to produce a 125-year database to the end of December 2024.

The data is based on the All Ordinaries accumulation index from 1980 and prior to this date, it used a number of different approaches and sources to compile the database. In comparison, the Austmod “share” sector is primarily based on National Mutual EFG unit prices until 2009 and then the ASX 300 Accumulation Index.

In Table 4, 85% of the occurrences of yearly declines occur over one, two or three years. The highest average cumulative decline (of -33.8%) occurs over five years (this relates to the five years ending 31 December 1974). For the four overlapping 65-year periods ending 1964, 1984, 2004 and 2024 respectively:

• the cumulative declines never exceeded six years, which is the same as in Table 2;
• auto-correlations at a lag of one year became increasingly negative (i.e. .074, 0.000, -.152 then -.202 for the 65 years ending 2024); and
• auto-correlations at a lag of two years were all negative (i.e. -.138, -.313. -.205 and -.158).

As expected, all the Table 4 results for the 65-year period ending 2024 are similar to Tables 2, 3, 5 and 6 Austmod Australian shares results for the 65-year period ending 31 December 2024.

Table 5: Australian shares simulated results

The second section in Table 5 is simulated Austmod Australian shares based on Section 18 of Australian Investment Performance 1959 to 2022 (and Investment Assumptions for Stochastic Models) except that:

• average investment returns and standard deviations have been set equal to the 65-year historical results to aid comparison with the first section; and
• the technique for determining ‘S sector’ auto-correlations has been improved.

The third section in Table 5 is the same as the second section except that auto-correlation inputs have been set to zero. The very small residual auto-correlations in the simulated output are mainly due to cross-correlation effects.

The fourth section in Table 5 shows the ratios of third section cumulative declines to those from the second section.

The number of occurrences of cumulative declines more than doubles when auto-correlations are not allowed for and there is a small increase in average cumulative declines. These results are solely due to auto-correlation. This clearly shows that simulation modelling which does not allow for auto-correlations significantly overstates sequencing risk from Australian shares.

Table 6: Combining Australian shares and cash

Table 3 above shows that the international share portfolios have suffered higher sequencing risk than Australian shares.

Also, a close examination of Tables 1 and 2 shows that, apart from increased one-year cumulative declines (both in number and quantum), a portfolio of solely Australian shares has lower sequencing risk and better overall long-term investment returns than a typical balanced portfolio.

These features suggest that a combination of just Australian shares and cash, such as in Table 6 below, might have merit.

Table 6 shows historical (not simulated) results. The four sections comprise 100%, 75%, 50% and 32% allocations to Australian shares. The 50%/50% combination is consistent with two recent articles^[2] by Cary Helenius which also illustrate results from a dynamic combination of shares and cash.

Apart from increased one-year cumulative declines (in number but not quantum) the 75%/25% combination experienced less sequencing risk yet had higher investment returns than a typical balanced portfolio. 

This extraordinary result demonstrates how the benefits that flow from negative auto-correlations at lags up to about 5.5 years can offset the adverse impact of higher volatility (i.e., higher standard deviations).

Similarly, the 50%/50% combination experienced less sequencing risk and had a lower standard deviation than a typical balanced portfolio, with an investment return marginally greater than a typical balanced portfolio.

Further indirect benefits of these Australian shares and cash combinations are that:

• they are highly liquid;
• they are easily accessible to individuals and Self-Managed Super Funds; and
• the combinations between Australian shares and cash can be varied over time or between individuals to suit their risk profiles.

More work

The historical results in Table 3 could be extended by examination of the impact on sequencing risk, returns and volatility of other investment classes (e.g., property trust, direct property, loans/credit, inflation-linked bonds, international fixed interest or semi-government securities). The historical results in Table 6 could usefully be supplemented with investigation of historical experience over say 10 or 20 year periods rather than the 65 years adopted here, and with forward-looking simulations to test the current conclusions.

It might also be informative, for both retirees and their advisors, to include specified withdrawals at regular intervals under the same (or some of) the scenarios in this article.

References

CPD: Actuaries Institute Members can claim two CPD points for every hour of reading articles on Actuaries Digital.