A "Ripcord" Simulation

Since Monte Carlo simulation is so irritatingly passive when it comes to retirement failure and so blasé about the constraint of only having one life to live, I asked myself this random question the other day:

What if, in a simulated context, I wanted to: 1) try to eliminate longevity risk by using annuity risk-pooling, 2) only make a risk-pool purchase at some future date and only if it's necessary, 3) use some minimal two-notches-above-indigent threshold level of lifestyle as a safety-net floor assumption, 4) be able take action while my portfolio is still viable and can still afford to make the purchase, and 5) make the purchase at an age when it is supposedly efficient to do so? Assuming that this engineers a way to mostly eliminate longevity risk (while still retaining a ton of other risk), how much more could I spend today? Or conversely, how much of a margin of error might I have in my current spending?

That's a great question. It's too bad I am not equipped to answer it. Or at least answer it with any serious rigor or credibility. That means, like I've done before, I'll try a back-of-the-napkin amateur hack to see what I can conjure up while also not imagining I am the first to either ask or attempt answer this question. I'm not going to present data or tables or a lot of charts because it's my own private data and also I weary of charting as summer approaches. This, by the way, is a much shorter post than I had planned mostly because the results seemed underwhelming and it saves me a ton of effort on that rigor thing. But, in turn, I guess you'll just have to assume I am being honest with myself…and you.

Let's start with a couple baseline assumptions. First, let's assume that "safe" spending for an early retiree with an expectation for a long retirement, for many many reasons, can be well under 4% in the first couple of years. This shoot-from-the-hip assertion can be confirmed, if you wish, by checking almost any credible and honest article or paper that addresses this issue of retirements lasting more than 30 years[1]. I've certainly commented on and linked to the idea on this blog before. It's probably under 3% for a 58 year old but let's say 3% just for fun which happens to also roughly comport with my own comfort level at 58. Second, let's say that an optimal age to annuitize, if one were so inclined to do that, is somewhere around age 75[2] depending on the assumptions.

Then, to do this back-of-napkin thing I:

- Determined a minimum-threshold income in current dollars that is above destitution (ignoring social security for a moment) but lower than a basic, fully comfortable lifestyle. This is the amount I could accept as a bare minimum if I had to rely on it forever without SS. (amount not shown of course),


- Inflated that minimum income out to age 75 (at 3%) and then created a spreadsheet version of an inflation adjusted minimum-threshold income stream from age 75 to 95 (my current terminal-planning age),


- Took the average of that stream. The average was a way to try to partially hedge inflation in the annuity without getting into COLA riders; I assumed I'd be able to successfully invest the underspend in early years in order to fund the overspend in later years but that's a pretty sketchy assumption I think,


- Determined, using inaccurate but free internet tools, the premium for an immediate annuity for a 75 year old (and unjustly assumed that that calc and the underwriting assumptions will miraculously hold true for a whole bunch of years) that delivers an income stream the same as the result of the last step. I checked this against my own PV calcs for reasonableness and to see if there was a reasonable discount received for the pooling, which there was,


- I simulated a base case using a Monte Carlo simulator using some idiosyncratic and sometimes personal assumptions [3] that were tweaked here and there. The goal was to see if the spend rate was anywhere near an initial 3% spend rate with a 95% chance of success (portfolio run to zero, ignoring SS though) when assuming a terminal age of 95, among other things. It was. (actually it ended up being a little under 3%),


- I then re-ran the simulation in trial and error mode (changing only spending [4]) with the terminal age reset to the annuitization age of 75. The threshold I set for the portfolio was that 95% or more of the time I wanted the end value to be greater than the premium for the age-75-annuity that was determined above. There had to be a reasonable chance I could pull the ripcord and buy an lifetime income stream at an age that made sense while I still had money to do it. That right there is the attempt to be non-passive and less blasé, something at which traditional simulation is not very good.



The result? Well, I had been hoping for more drama. I wanted to be able to swim in a giant pool of new money that had been liberated by my sheer retirement-finance-genius. In practice, however, I could only nudge spending up to about 4% or a little more[5]. The percentage change is honorable of course (let's call it a 33% change from .03 to .04 … actually it was a little over 40% using my real results) but I'm thinking that a 4% spend rate is back to where most people think they are starting in the first place. It's certainly not 7 or 8 or 10%. I guess at least I know I now have a little margin of error in my spending budget this year. Also, I think I can now presume to better gauge, whenever the weight of longevity risk gets too heavy for me to bear, the rationale and cost and timing of opting out of a self-managed "risk pool of one" and into a more reliable and certain pool of millions. If I have to.







-------------------------------------------------


[1] Too many sources to cite. As shorthand, think of Inglis's divide age by 20 rule that I profiled on this blog or Waring and Seigel's ARVA using the Excel PMT function, or any recent article by Wade Pfau on withdrawal rates.

[2] I used Milevsky's formula from Annuitization and Asset Allocation, Moshe Milevsky and Virginia Young, 2007 to get a thumbnail estimate for an optimal annuitization age. Ignoring the option of scaling into annuities over time, I've charted it out thus for a couple risk-aversion coefficients while varying expected returns on the x axis:



[3] not shown but it has things like age 58, a declining spend path in three stages, a 50/50 allocation though that is not really mine, some return suppression for 10 years, social security at 70, terminal age 95, along with some other stuff.

[4] Remember that I have three stage spend plan; I only changed the first stage which happens to end at 68.

[5] Yes, but I am managing to worst case here.  After 75 there should still be simulation scenarios that are wildly successful in addition to the annuity. I am ignoring these and they could interact with the decision on how much I might want to spend now. This is a big hole in the analysis but I think the worst-case thinking is useful.