My First Whack At Simulating A Floor-and-Upside

This is necessarily very imperfect because: a) my simulator is not nearly sophisticated enough or well enough tuned to do what I am trying to do, and b) my first "whack" used some idiosyncratic "software-testing lab" assumptions more than anything remotely generalizable.  Mostly I just wanted to see what would happen.

The basic idea was this, for someone with the asset capacity to be able to execute a floor-and-upside strategy: 1) conceptually commit some portion of  assets to an income floor, 2) for simulation purposes zero out that portion of assets and convert it into an (at risk) income stream that grows at some rate[3], 3) allocate any "excess" assets 100% to an upside risk portfolio as a long term upside option (this is a kind of Kelly criterion bet because the hold-time-frame is long and the equity wager has an edge that often says go all-in in cases like this), and then 4) run a sim and see what happens.  Step 5 might be to add back any residual value of the income producing assets at the end to the terminal wealth created by the upside-option but that'd just be icing on the cake with legacy planning implications.  

Step 1 - Figure out the floor

For step 1 I used Michael Zwecher's formula for %assets committed to a floor that he proposes in Retirement Portfolios. It looks like this:

where

A% = flooring allocation

L = lifestyle

W = wealth

r = constant rate

i = expected inflation

N = number of years payments need to cover

M = lag before payments start

Using a hodgepodge of assumptions not documented here I came up with about 70%. General sim assumptions are described in note 1.  The upside portion would then be 30% of assets.

Step 2 - Figure out the income stream and the scope of upside assets

I multiplied what would otherwise be the starting endowment by 30% and used that number for the starting endowment in the simulation.  

For income I calculated what might plausibly be produced from the 70% by a mix of income sources from bonds to dividends, to REITS/MLPs/alts etc. In this case it worked out to be ~3.4% yield. Tax stuff is ignored. I discounted a bit to be more conservative and I plugged it into a sim placeholder for income streams.   Another placeholder for income has a modest assumption for SS starting at age 70.

Step 3 - Allocate the upside portion

I allocated the "upside assets" determined in step 2 100% to the risk asset. In this case, for better or worse, that is the S&P. If inflation-adj-spending outruns income this upside portfolio gets tapped but the "seed corn" in the floor portfolio doesn't.

Step 4 - Let 'er rip

The base case, using some combo of idiosyncratic and standard assumptions^[1], had an already low fail rate estimate of 7-8% with a median fail duration of 8 years.  This was not a longevity -varying run; all sim paths led to age 95.  Switching assumptions to floor-and-upside, the fail rate estimates came out to between .5 and 1% and the mean fail duration was 10 years.  Letting longevity vary with a SS Life-table re-sampled for a 58 year old basically rounded the fail rate towards, but not quite, zero with a median fail duration of 5.5 years for those paths that didn't work out.  In my experience, this looks like it effectively took almost all of the risk off the table, at least in fake sim-world. The upside portion of the portfolio shows strong upside optionality using this chart of terminal wealth vs terminal age in the longevity varying run.  

Caveats

- unsophisticated simulator with overly simplistic approach

- dividend compression and div fails during recessions not modeled

- fail rates are already working from a starting point of small numbers

- income growth is constant, not stochastic...or correlated to anything
- even during fail, income still continues like an annuity which wouldn't really happen[2]

- terminal value of income assets ignored but could be significant

- eating the "seed corn" implied in the floor portion of assets is not modeled
- lifestyle opportunity cost or what we could have spent but didn't not modeled
- the floor over funds lifestyle here which means tax and reinvestment considerations

- I'm sure there are plenty of other holes in this "first whack" analysis...

Conclusions

The floor and upside approach in the sim, while not a traditional total-return-accumulation style portfolio choice, seems to look pretty good for at least one unique set of retirement assumptions where there might be capacity to try the strategy. For these fake assumptions it feels like risk went to zero.  My guess is that for lower net worth and/or higher spending rate assumptions and for situations where the Zwecher A% is > 1 the results would be less fun.  If these were to ever be my results it would be a clear call to stop fussing with retirement math and go do something else.  It'd be interesting to extend this a bit but I think I'd want to enhance the simulator first, something I am not inclined to do anytime soon, and I believe others affiliated with RIIA have been down this path before with convincing results.  

Notes ------------------------------------------------

[1] The sim was using a hodge-podge of mostly ideosyncratic lab assumptions I sometimes use to shake out the sim software but could be characterized as med-high net worth with conservative spending around 3%, 50/50 allocation in the base case, slight downward trend in spending over time, return suppression for 10 years, to age 95 unless otherwise specified, plus some other...

[2] this is part of the naivete of the software.  In real life the upside portfolio would be depleted, which it is in the software, and then the floor portfolio would be consumed, which doesn't happen, and then one would start reducing consumption, monetizing residential real estate, etc.  I take the upside depletion as a fail proxy here.  No number from a simulator really means too much especially after about the first year of the sim.  It's really just a risk assessment tool that can be used in conjunction with other tools and in conjunction with an ongoing process of evaluation and adaptation.  The only point here is to think about committing a large chunk of an estate to a floor.

[3] Basically this looks and acts like an inflation adjusted immediate annuity inside the sim.  If I were a better programmer I would figure out how to start cannibalizing the floor when circumstances warranted...or even better adapt spending with better rules.