A common theme in this blog’s last few years has been the search for the perfect retirement portfolio. If you are familiar with the ideal goal, you know the challenges to achieve it: longevity and sequence of returns risks, declining cognitive capacity to manage complex assets, various left-tail events, inflation and so on.
The simplest solution I thought of so far is a combination of two elements:
- an annuity that grows with inflation for non-discretionary spending, covering the premium of medical insurance for that left tail risk.
- a risky portfolio for discretionary consumption. The discretionary consumption budget depends on the value of the portfolio and (T – t), where T is your life expectancy and t is today. T is not constant: if you use the formula to calculate how much you can spend each month, (T – t) decreases by less than a month because the longer you live, the longer your life expectancy.
Even this strategy is not bulletproof. If you are renting, your particular rent might grow more than general inflation; also, the impact of moving when you are 80 is not negligible because you are not that flexible anymore. If you own the place where you live, the budget to implement the plan becomes untenable for most people, and the wealth locked into housing is too much.
While targeted only at discretionary spending, that income can be very volatile and not many people would like it. Based on the feedback we received from listeners of our podcast, there is also a group that hates the concept of annuities (mainly because they do not understand the concept of longevity risk). Even if there might be a theoretical ideal solution to the problem, some might reject it for personal preferences ignorance.
In this journey, I stumbled upon an interesting paper from Amundi (link here).
It covers three decumulation strategies:
- the classic Safe Withdrawal Rate
- maximising the expected discounted utility of lifetime consumption
- maximising the expected utility of the ruin date (RDUM)
If you are reading this blog, you are familiar with the first one. The second is basically the same strategy I described above for discretionary income, which is not a viable option because consumption is too volatile. The third one sets a certain level of consumption, like the 4% rule in the SWR, but allows the composition of the portfolio to vary through time: instead of having a constant 60/40, the “risky” and “risk-free” parts of the portfolio change based on life expectancy and the portfolio size.
In an ideal world, I should not have to say it but I know you, I see where you click, so go read the bloody paper, don’t be lazy. Everything would be clearer 😉
The RDUM strategy has a take-profit behaviour: when you target constant consumption, if the value of the portfolio grows, you want to de-risk in order to decrease the probability of ruin (deplete the portfolio before death) and vice versa. There is also a point where the portfolio is big enough that even investing at the risk-free rate covers all future consumption.
This passage is really insightful:
This is the main issue of high-risk portfolios in retirement: when they work, they are too good (unless you care more about the wealth of your grandsons than what you consume); but they still have a relatively high probability of not working at all. The retiree would either be alive and broke or a dead multimillionaire.
Whereas:
The RDUM strategy performs better than SWR because it increases what matters most: certainty. In the above table, 50, 60 and 70 represent the MAXIMUM allocation to the risky asset allowed; the higher the max risk, the higher the stats: a sign we are going in the right direction.
Problem solved?
The RDUM strategy has a potential behavioural issue. If run “at home” (assuming the person can solve the equations shown in the paper), it requires the retiree to take more risk when their wealth is down; but this can be solved if, for example, this strategy would be implemented inside a suite of Target Date ETFs. Targeting the date of your death is a bit darker than targeting the date of your retirement but well…
Unfortunately, even the RDUM strategy is not immune to longevity risk. There are more than 20k lads ladies (80% are women) older than 100 in Italy. That’s the main issue with adopting a take-profit strategy: you lower the volatility/sequence of return risk while increasing the longevity one. A positive bequest amount is an insurance against longevity risk.
The second issue related to what we have shown so far is the methodology’s assumptions: so far, the returns on the risky asset follow a normal distribution with constant expected returns and volatility over time.
The paper tries to solve this by simulating a risky asset with a skew normal distribution:
The probability of success sensibly decreases and the RDUM strategy starts to behave like the SWR one (the average wealth goes up, bringing us again in that space where if things go well, you are too happy). This is why, at least intuitively, adding positive-skew assets to the portfolio (trend following and long-vol) should increase the probability of success.
The paper ends with Amundi employing a “very Europoor Risk Parity” portfolio as the risky asset (average return = 3.5%, volatility = 6.8%) for the SWR and RDUM methods:
The only value I see in this exercise is to prove that by reducing volatility and max drawdown, i.e. sacrificing returns but maximising the Sharpe ratio, we get a better probability of success. Should ring some bells to this blog readers (and also believers in the confirmation bias) 😉
The third issue with these simulations is, as always, inflation. You can ‘cheat’ and use real rates instead of nominal rates, but then you should not have anymore an always positive risk-free rate. Over the long term, it is correct to assume a positive real risk-free rate (see the UBS Yearbook); unfortunately, this is not true for the short term, meaning the volatility of the real risk-free rate introduces a sequence of return risk. Also, while the LT looks good for many countries, it hasn’t been true for all countries. The retiree has to be in the right place at the right time to enjoy a positive real risk-free rate.
The RDUM and SWR methods, as modelled here, imply the retiree can always store their portfolio in positive yielding, inflation protected investments: I wish we could live in that world.
My naive conclusion reading the paper is that the “V” shaped glide path is still, so far, the best retirement solution for retail investors. As we know, the SWR method leads to ruin when we have bad years at the beginning of retirement and this precaution erases a big portion of the issue. The beauty of the SWR strategy is its ability to deal with longevity and inflation risks.
Higher-Sharpe portfolios reduce the risks around the “V” shaped glide path: they create fatter margins of error around the timing of the “V” (in some cases you can even argue the V is not needed anymore) and the yearly consumption amounts, if you want to stick to the classic 4%. They have obvious perils as well: past correlations might fail, increased model risk for whatever is included in the “alternative” bucket and higher reliance on a sharp brain to maintain it (there is the potential benefit that the constant mental challenge would keep your brain brighter for longer).
What I am reading now:
Follow me on Twitter @nprotasoni