What Would A 4% Fixed Withdrawal Actually Have Felt Like In 2009?

Let's be clear, I am a committed partisan of adaptive retirement plans.  Sticking to a fixed, deterministic plan in the face of an often aggressively changing universe strikes me as flirting with one of the many definitions of insanity.  Yet the literature of retirement finance has a seemingly inexhaustible supply of commentators that continue to say that 4% or even 5% (or more) constant inflation adjusted spending will "probably" be ok based on, well, whatever.  But these commentators are usually men and women, more often than not academics or institutional practitioners, that still get a paycheck. Their research is math-rich, vaguely abstracted, and cool to the touch. That's why I generally trust the retired people to write about retirement finance, people like Darrow Kirkpatrick or Dirk Cotton or KenSteiner. They have skin in the game.  It is also why I wanted to run this little sketch on how constant spending might actually feel to a retired guy during a really bad time i.e., let's let the plan NOT adapt and see what happens scrolling forward one month at a time through something like 2008 and 2009.

The basic idea for this post then is to say: as a thought experiment, let me retire as a 57 year old in 2004 with $1M invested in a low cost 60/40 fund. I'll withdraw or spend a fixed $40k (inflated at ambient CPI) and see what happens to the effective withdrawal rate (the inflated 40k divided by the beginning endowment value in any month, an endowment that has been adjusted for spending, dividends, returns etc).  I'll also look at the forward-looking fail rate estimates using simulators or calculators using the fixed spend but also using all the new inputs (e.g. endowment, age, longevity) that would have been available at any of the new times T (we'll do this annually).  So the "game" here, to summarize, is to answer these questions:

1. What would happen to a fixed withdrawal rate recalculated as % of a changing endowment at each time T during a market crash,

2. What would happen to an instantaneous or forward fail rate estimate at any given time (e.g., as a proxy, calc an estimate at the beginning of each new year) during the same time frame, and

3. How would it have felt as a real live retired person and what would I have done?

The assumptions are critical of course.  Some of the main ones* include:

- Initial endowment at T(0) is $1M

- The time frame is a relatively arbitrary 1/2004 to 4/2016

- Spend rate is a fixed 4% calculated at inception and cpi adjusted thereafter

- Spending is taken out at the beginning of a period

- A low cost total market 60/40 mutual fund was used as portfolio

- A custom built Monte Carlo simulator was used to look forward at any T(n)

- An aftcast historical simulator was used to contextualize MC simulator

- Model scrolled forward each month/year to recalc the spend and fail rates

- Only spending and allocation are kept constant (cpi adjusted)

- Age, duration/longevity, and endowment vary

- I farcically ignored taxes and Social Security

When one does this kind of process using these kinds of assumptions it looks -- if I have not erred in the model, which I may have -- like this with the effective spend rate on the left (blue line) and the fail rate on the right (red and green lines):

Terminal Portfolio:                  $1,149,659

Terminal Return/CAGR:           14.97% / 1.03%

Min/Max Spend Rate:               3.6% / 6.3%
Ending Spend Rate:                  4.5%

Max Fail Rate - MC:                29.4%

Max Fail Rate - aftcast:            51.3%

Max Portf Drawdown             -27.52%

Portf Highwater:                       23.84%

Now let's go back and look at the questions. 

1) What would happen to a fixed withdrawal rate recalculated as % of a changing endowment at each time T during a market crash?

Well…it skyrockets to over 6% in 2009.  If I knew what I know now about sustainable spending rates I would have gotten very uncomfortable if I had calculated the effective spend rate at any time during this period and if I had had a dyspeptic view of the future at the time (which I did). The fact that it ends well does not change the feeling one would have had in March of 2009.

2) What would happen to an instantaneous or forward fail rate estimate at any given time during the same time frame?

Again, these calculations explode up.  Between the two simulators, fail-rate estimates ranged from 30 to 50% in early 2009.  This was mostly due to the drawdown of the endowment from spending and market losses up to that point (there is a more formal point to be made on "sequence of returns risk" of course but I want to focus on the emotion of 2009 here). Technically speaking I was not exactly retired in 2009 but I was also not formally employed by anyone other than myself and I remember Q1 well.  Everyone talks about the crash of '08 and maybe a lot of people think it was over by the end of that year but in March of 2009 -- being in the midst of an unwanted move, a divorce, and a market crash all at the same time -- I remember the last final downdraft as a nauseating flush into the abyss.  It's a good thing I was not calculating fail rates or risk of ruin at the time.  This is a point well made by Moshe Milevsky here. 

3)  How would it have felt as a real live retired person and what would I have done?

I would have lost my mind. A greater than 6% spend rate and up to 50% risk of failure (actually it would have been a 63% fail estimate at its monthly peak)?! If I had done that math in 2009 it would have made things feel worse than they were; I probably would have thrown up (OK, I exaggerate but only to make a point). If I had seen the left half of this chart I probably would have ended up being a good behavioral finance case study, though.  The temptation to sell or reallocate or whatever would have been overwhelming.  The chart -- without the benefit of lines extending out to the right of '09 -- looks like the end of the world...or at least pretty bad.  In real life in '09 I did nothing (or at least with respect to spending and strategic allocations.  In fact, I used the opportunity to lean into high yield at historically wide spreads, what later turned out to be my trade of a lifetime). That's probably because: 1) I was not perseverating on retirement finance at the time (a good lesson all by itself), and 2) I also had a dependable floor of income (see Pfau on the concept of floors plus upside in retirement).  The ex-post crystal clarity we have in seeing the right side of the model/chart and the providence of having been on the receiving end of one of the bigger bull markets in history, 2009 through today, distorts our view of what it was like in the moment.  On the other hand, here we are at more or less historic highs in equity and credit markets. What are the chances that the now 69 year old (in the model anyway) is going to see another spike in the chart? What is he going to do then? For that matter what is he going to do now? Those are good questions. It is also why I am a partisan of continuously adaptive retirement methods. 

Next: What Would It Have Felt Like in 2009 With a Fixed Withdrawal- Part 2 (3% spend)

*Other assumptions:

- dividends are reinvested

- Monte Carlo Sim…

   - terminal age not fixed; varies within Gompertz distribution

   - longevity distribution is capped at age 100

   - inflation and returns are independent

   - inflation and returns vary within historical distributions

   - returns are capped a bit to try to mimic modern return expectations

   - equity and bond returns are dependent and vary within historical distrib

   - run age is current age at any time T(n)

   - run done at beginning of year using end of prior year data

   - assumes 60/40 allocation forward expectation; 70/30 TBill/bond

   - .5% fees and a factor for taxes on capital appreciation

   - 500 sim runs (short, to speed up the process)

   - spend rate 4% inflated over time

- Aftcast simulation

   - uses FireCalc

   - uses 95 as terminal age so 95-current age at any run

   - CPI inflation

   - constant spending

   - total market and 60% equities for allocation

   - spend is 40k adj by shiller CPI at any time T(n)

   - I acknowledge that using long time frames means fewer aftcast cycles

- CPI is from Shiller data, MF data from Yahoo finance

- Why 57? That's my age but also I mis-coded Sim so that <57 doesn’t work

- No spending shocks but that would be fun to try

Reading...
It’s Time to Retire Ruin (Probabilities), Milevskey 2016 
39 Modern Retirement Income Planning Techniques, Pfau 2016 
Can I Retire Yet, Darrow Kirkpatrick 
The Retirement Café, Dirk Cotton 

How Much Can I Afford To Spend In Retirement, Ken Steiner  
Retirement Researcher, Wade Pfau

The Only Spending Rule Article You’ll Ever Need, Waring and Siegel 2014. 
Annually Recalculated Virtual Annuity: A Test Drive, by Me.