Is sequence risk real, or overhyped? Investment consultant and industry stalwart Graham Bentley attempts to instil some order in the debate...
Industry commentators wailing about the horrors of sequence risk recently got a soft slap from FinaMetrica's Paul Resnik, when he opined their "reasoning was not supported by the evidence", based on his research. Given advisers are beginning to design shiny new ‘Centralised Retirement Propositions', what should we make of this spat?
As I'm sure you're aware, sequence risk is the chance that a withdrawal and capital preservation strategy in retirement could be ruined by a series of negative returns. Put starkly, during a valuation fall of 40%, a 5% withdrawal taken monthly rises to over 8%, accelerating the fall. Meanwhile, the 40% fall requires a 66% capital recovery - plus the withdrawal - to return to par.
Fear of sequence risk has been used to justify a number of portfolio strategies that eschew equities in favour of bond-dominated, lower-risk portfolios. From simple asset allocation rules of thumb (e.g. percentage equity exposure should be 100 minus your age) to complex ‘multi-strategy' approaches (e.g. Standard Life's GARS), a common theme is the apparent need to avoid the volatility that equity prices bring to bear on a portfolio.
Sequence risk can be easily illustrated; a number of such sales aids are doing the rounds, stringing artificial but mathematically rational price sequences together to make a point. If I select the ten worst years on the market since 1962 I can string them together and run out of money in six years with a withdrawal rate of only 3% of initial capital. Clearly if you're flogging a bond-dominated product to this innately cautious income-hunting market, it's probably in your interests to collectively wring hands over the dangers of equities, particularly if you have a cohort within the adviser community for whom the term ‘risk' is most likely to be associated with the word ‘regulator'.
Related reading: Understanding the silent portfolio killer of sequence-of-return risk
On the other hand, advocates of equity weight legitimately point to the phenomenon of ‘reversion to the mean', where extreme events are followed by less extreme ones, and consequently prices and the average price converge. Falls, we are led to believe, are generally recovered before too much damage is done. Data can be gathered to support this belief, and an adviser's received wisdom may be to confidently present this to clients as fact.
However, risk is subjective. It can be described as probability multiplied by impact; how likely is the bad thing to happen, and what is its impact on me if it does? Analysis generally focuses on likelihood (i.e. what is experienced by all investors) but rarely is it concerned with impact (what is felt by individuals). That's a subject for another day so, for the time being, let's look at likelihood. In analysing this, sequence risk deniers follow two approaches.
Good versus bad
The first uses randomised price movements ironically described as ‘Monte Carlo' simulations to produce thousands of portfolio histories, and presents the distribution of outcomes - how many ‘good' versus ‘bad', but with a focus on the average (I won't quibble about mean versus median here). There is a sort of ‘white coat' syndrome about this: it's maths, therefore it's science, therefore it has meaning. Well frankly it doesn't. The average result is the one you won't get. Would you confidently let your non-swimming child wade across a river with an ‘average' depth of three feet? Retirees may hope for the best, but they should plan for the worst. The most meaningful data should account for the personal impact of bad luck. How would the client feel if, despite the statistics, they experience a set of never-seen-before conditions?
The second approach uses neat parcels of asset price history, generally in round numbers (ten years), and rolling. Daily rolling ten year periods are biased in that each period studied contains most of the information used in the last period, thus virtually replicating the result. Ten-year periods can overlook the ramifications of ‘compounded sequences,' where an impactful negative sequence in year 11 onwards applies to the fresh-start portfolio, not the most recently tested period. The portfolios themselves are usually composed of proprietary indices so the effect of charges is ignored. This significantly overstates returns by up to 100bps per annum and is wholly at odds with investor experience.
Most importantly, both approaches are biased - they choose to compare equities to bonds, and the portfolios are so constructed (e.g. 40% equity, 60% bond). This makes for simplified analysis, but is profoundly flawed and positively (or is that negatively) misleading, since they avoid the impact of ‘surprise' assets - where asset price behaviour has further risk-dampening impact (reducing the falls) and yet has contributed significant excess returns. Property is an example of this, and as we'll see below, this may better explain ‘real life' relative portfolio performance than the equity/bond weight.
Whatever the approach to analysing the data, it is important to understand that there are more things that can happen, than have happened. We should be careful using ten year rolling periods from the last 45 years. That period coincided with significant (and unusual in a 200 year context) contributors to portfolio returns, including...
The demographics of western economies, and the productivity of baby-boomers who are now decumulating and consuming, and no longer accumulating and producing, as they were for the bulk of the period.
Astonishing rises in equity values: Six of the highest ever annual returns on the UK market were experienced in sequence in the ten years between 1976 and 1985.
The rise of emerging markets - close to 10% per annum compounded over 40 years.
UK inflation falling from 24% - and interest rates from 17% - with the concomitant re-rating of bond prices.
Commercial property returning almost 10% per annum between 1990 and 2007 - and delivering over 6% per annum to date despite a 50% fall in 2007.
To assume that the next 40 years can be as beneficial may be the triumph of hope over reason - these are extreme events. Markets that have delivered exceptional returns are, by definition, the exception. Given the anti-sequence risk cohort's fondness for the reversion-to-the-mean argument, there may be some confirmation bias going on here. Even relative performances are under threat. It is difficult to reason that bonds should benefit investors to the extent they have since the ‘Great Moderation' began in 1982. In fact, one could argue that sequence risk might be more applicable to a series of annual bond returns rather than equities, if (and when) interest rates revert to the mean. This would have major implications for those who have been persuaded to believe a high weighting in bonds is a cautious and beneficial strategy.
What we found
That said, people do like to reference history, if not to rely on it. We (gbi2) were asked to conduct research relating to the impact of the financial crisis. Intuitively, this seems like a good example of a potential sequence risk issue, particularly for a retiree starting withdrawals at the market top in October 2007. However, we decided to look at a longer-term and more realistic experience that might be considered a worst-case scenario.
Our hypothetical investors were born on 30 June 1930, and every month-end thereafter until 1945. They retired on their 65th birthdays with a portfolio worth £500,000. We wanted to see the effect of a series of downturns rather than a single event. This 20-year period contains four significant market downturns ('98, '00, '07 and '11) through four business cycles.
We used three portfolios that we knew had been modelled in 2000 to reflect three risk levels: two (12% Equity), five (47%) and eight (80%), and applied withdrawal rates of 3%, 5% and 7% paid monthly. We used Investment Association sector averages to represent the asset classes rather than no-charge indices. We then extracted 181 price series for each portfolio to analyse every retirement period in that dataset from 20 years (240 months) to five years (60 months).
We produced a significant amount of data (see the graph and *note below), and hope to publish our detailed results presently. However, in summary:
Risk level two was essentially high risk in disguise - withdrawal rates needed to be below annuity rates to preserve capital. Most data sets produced a constantly falling capital balance. The only relative benefit was realised where retirement started at the outset of a major downturn, e.g. September 2000. Even then, risk five results were comparable given their greater propensity to recover. Protection in downturns had little ultimate payoff.
Sequence risk was clearly evidenced, and was more punitive as withdrawal rates rose. It was most in evidence where multiple market falls were in the dataset.
Risk levels' five and eight returns were positively skewed but, counter-intuitively, level five more so than level eight. Risk eight had a median return of 4% per annum, a low of less than 0.5% per annum, and a high of more than 11.5% per annum. Risk level five had a more normal distribution of returns, but more closely gathered around a median not too dissimilar to level eight at just under 4%. This better ‘bang for your buck' was attributed to an 18% holding in property in risk level five.
As a rule of thumb, dividing the annual withdrawal rate by the proportion of initial capital that remains is a good indicator of portfolio stress; if it is more than 10%, recovery is significantly more difficult to achieve.
Retirees all suffered varying degrees of drawdown pain, where their balances fell below their initial capital. Some 97% of our sample would have experienced valuation falls below their initial capital - median maximum drawdown was over 30% across the 181 periods, on both risk levels five and eight.
Equity came into its own in recovery phases, with risk level eight returns accelerating as markets reverted to their mean prices. It was this phenomenon (extreme rebounds) that boosted level eight's overall median return. Being brave enough to increase equity exposure after a severe fall produces spectacular returns.
A sensible withdrawal rate - if inflation linked - appears to be 4%.
In summary, sequence risk is real and can play havoc with portfolios, but unless one is prescient enough to market-time (e.g. holding risk level two when markets are about to collapse and level eight when they rebound), then that risk has to be tolerated, or a compromise (risk level five) reached. Heavy bond portfolios with withdrawals beyond annuity rates are probably the worst choice.
The only long-term meaningful returns over time come from significant equity exposure, but some painful drawdowns will almost certainly have to be experienced.
*Note: The chart assumes 7% per annum monthly withdrawals (i.e. the absolute worst case scenario).
Graham Bentley is managing director at gbi2
More than 4,500 retail investors affected
$17trn of debt is now ‘paying’ a negative yield
‘Widespread problem’ researchers claim
Equity market can be ‘scary place’
Retail savers pull almost £5bn from equities in Q3 2019