Hi Guys,
There has been much MFO hot ink spilled over Paul Merriman’s purportedly controversial proposal for an all equity diversified portfolio during the accumulation phase prior to
retirement.
This is not an entirely new idea or strategy. It has been long advocated by folks with a cast iron constitution who do not panic when adversity erodes their portfolios. These folks tend to not monitor their portfolios too closely and have a deep rooted faith in the equity market’s resiliency. They trust that volatility over the long haul is not worrisome, and that the superior returns of an all equity position outdistances untimely and unpredictable late-period disappointments. Is that trust warranted?
The MFO commentary over Merriman’s proposition has been mostly sensible, but contaminated just a little with nonsensical assertions and opinions.
Surely there is no shortage of assertions and opinions. Most are made of solid stuff, but some are composed of mirrors, smoke, misrepresentations, misunderstandings, and misinterpretations. But something has been completely missing from the exchange. What story does the data tell? Where are the analyses?
Here is my attempt to fill that void.
I ran a few Monte Carlo simulations using the Flexible
Retirement Planner’s version of that useful tool. I surely do not claim that my analysis is exhaustive. It definitely is not. The entire sequence of calculations that I completed took about 15 minutes. That’s one of Monte Carlo’s most attractive powers. It enables a host of what-if scenarios and sensitivity analysis that permits the user to explore various aspects of the problem quickly. The code runs thousands of random simulations per case.
I hypothesized two alternate portfolios: an all equity portfolio, and a 60 % equity, 40 % fixed income mixed portfolio. For the all equity holdings, I postulated a reasonable 10 % annual average return with an 18 % standard deviation. For the mixed portfolio, I postulated a 7.5 % annual average return with a calculated 12 % standard deviation. I say “calculated” because I needed to assume a 0.4 correlation coefficient between equities and fixed income for the study period.
On the personal data level, I presumed a starting age 0f 35, a planned
retirement age of 65, and a life expectancy (combined husband and wife) of 95. So there is a net wealth accumulation period of 30 years and a balancing net
retirement phase of 30 years. For the
retirement phase, I postulated a portfolio of only fixed income positions with a 4 % average annual return and a 6 % standard deviation. These are reasonable inputs and easily changed.
Initially, I assumed an annual accumulation phase savings rate (in non-taxable accounts) of $15,000 (15K) per year. You’ll see that that saving level is totally inadequate given that I parametrically explored portfolio survival likelihoods for 30K, 40K, and 50K yearly withdrawals during
retirement. The Monte Carlo code adjusts these inputs internally for inflation. For the purposes of this analysis, I assumed an average inflation rate of 3 % with a 1 % standard deviation.
For the all equity portfolio, the survival probabilities were 55 %, 40 %, and 28 % for the 30K, 40K and 50K drawdowns respectively. None of these are acceptable and speak volumes for plan revision.
For the 60/40 mixed portfolio, the survival rates were only 22 %, 14 %, and 6 %, respectively for the same drawdowns. This finding is even less acceptable than the all equity portfolio.
With its enhanced net worth at the
retirement date, the all equity portfolio delivered superior survival prospects during
retirement when the retiree was assumed to convert to fixed income holdings.
The Monte Carlo simulator is a great tool to parametrically examine the impact of optional pathways.
For example, what happens to survival likelihoods if the savings rate is increased from the initial 15K to 20K per year? For the all equity portfolio survival rates increase to 71 %, 56 %, and 43 % for the 30K, 40K, and 50K withdrawal levels, respectively. That’s an improvement, but still far to risky for a comfortable
retirement.
At this juncture, I said “forget about the 40K and 50K
retirement withdrawal rates; they’re not realistic”. Focusing now on the 30K annual
retirement spending requirement, I asked the following question: “How much annual pre-
retirement savings is needed to reach a comforting 90 %
retirement survival probability?” The
retirement planner yielded a 31K per year reply.
You might not like the analysis or the resultant outcome numbers, but those are the hard numbers. They do not include any assertions or opinions. They are simply the output from thousands of calculations that did require a set of guesstimate inputs. Nothing new here when forecasting future outcomes.
The inputs can be easily changed to accommodate personal preferences and beliefs. Their influence on the robustness of the analytical findings becomes self-evident with additional calculations. Parametric analyses of this sort allow the pre-retiree to explore his options, and provide guidance for needed directional shifts in terms of time scale, savings level, and likely post-
retirement spending limitations.
The Flexible
Retirement Planner code does not model Fat Tails or Black Swan events, so it is slightly optimistic in its projections. I say “slightly” because I considered such limitations when I generated my own Monte Carlo simulator about 1990. I did include some Fat Tail modeling in my version of that tool to test sensitivity to that modeling shortfall.
The necessary program changes are relatively modest, only requiring a few additional lines of code. Basically, instead of using a single log-normal distribution for returns, two distributions are programmed with a break point at the extremes of what the user believes is the usefulness of the conventional log-normal distribution.
I tested break points at the 1.5 and 2.0 standard deviation levels. One obvious problem is that the choice of the second distribution is rather arbitrary given the paucity of real world data in this regime. I did examine a few Fat Tail distributions that I selected based on historical Black Swan events and outlier returns data. These did change the projected survival rates somewhat, but NOT in any decisive, decision reversing way, especially in the context of the overarching uncertainties of future market returns.
The Black Swans are rare. outlier events, and are statistically overwhelmed by the more common happenings. Waiting for Black Swans is a loser’s game. Nassim Nicholas Taleb discovered that when his earlier barbell investment style suffered weekly losses while waiting and waiting and waiting for the hammer to drop. His coworkers said it was like dying from a thousand small cuts.
It is my conclusion that Monte Carlo tools, even with their admitted shortcomings, are a superior way to address long term highly uncertain happenings. Many academic, military, and industry wizards who are tasked to predict the uncertain future would support me in that assessment.
I hope this post is helpful It is meant to be. I invested about 5 times more effort in writing this summary than I invested in actually doing the Monte Carlo number work.
I’ve posted this Link previously, but if you have an interest in pursuing the Monte Carlo tool, here is a Link to the simulator that I referenced and used:
http://www.flexibleretirementplanner.com/wp/planner-launch-page/Good luck. This
retirement planning tool will reduce the need for luck, but you will still need a healthy dose of it.
Best Wishes and Merry Christmas.