Thursday, January 30, 2014

Retirement Spending Experiment

I decided to run an experiment to test a retirement spending strategy (described here) on actual stock returns over the past 100 years. The goal of the experiment is to see how well the spending strategy balances the need for stable income against the need to adapt spending to portfolio gains and losses.

Along with the retirement spending strategy, I gave a spreadsheet to calculate the yearly spending amounts. My experiments used the default values in the spreadsheet (a 4% real return on an all-stock portfolio, low investment costs, target longevity of 100, and 5 years of spending kept in safe investments, among other assumptions).

I used inflation-adjusted stock returns in the U.S. from 1913 to the present to simulate seventy 30-year portfolios for an investor retiring at age 60. The spreadsheet calculations set yearly retirement spending for a 60-year old at 4.17% of total retirement savings. This percentage rises to 9.74% by age 89.

I chose a starting portfolio value so that monthly spending starts at $5000. After that, the execution of the spending strategy and investment returns dictate how spending changes over time. In addition to the spreadsheet percentages, I used some filtering. Each year, the spending level changes by only 30% of the change dictated by the spreadsheet percentages. So, if spending is scheduled to drop from $5000 to $4900, I only drop it to $4970. This smoothes out the short-term bumps.

You can think of this experiment having 70 different people all retiring at age 60, but in different years from 1913 to 1982. Each retiree has a different experience due to different investment returns. Here are the experimental results:

U.S. stocks have averaged more than 4% returns above inflation, which is the reason for the general upward trend in the monthly spending. The worst case scenario had spending drop to $2766 per month. I thought this might be related to the 1929 stock crash and the depression, but it isn’t. The worst case scenario began with the 60-year old retiree in 1966. By the time this retiree hit age 76 in 1982, spending had dropped by 45%.

It turns out that the 5 years of spending held in safe investments is what saved the investor who retired just before the 1929 crash. This retiree weathered the storm reasonably well before the huge stock rebounds in 1933 and 1935.

The best results came for the 1942 retiree. By age 90, this retiree saw spending more than quadruple. It would probably be more desirable for this retiree to spend more earlier on, but we can’t know when the stock market will give such great riches.

Overall, I’m pleased with these results. There seems to be a good balance between safety and the desire to be able to spend as much as possible. I can’t say that I’m prepared to follow this strategy myself yet; I doubt that I’ll make a final decision until the time comes to actually retire.


  1. Hi Michael,

    What's your opinion about this:

    1. @Anonymous: The "Variable Percentage Withdrawal" method described at that link resembles my strategy (and also resembles RRIF minimum withdrawal percentages). The main difference with my proposal is the portfolio make-up. I propose to keep asset allocation the same before and after retirement, except that some number of years of spending (e.g., 5 years) are kept in a completely safe investment such as a high-interest savings account. This reduces the portfolio's expected real return, but it significantly reduces the sequence-of-return risk; retirees can better ride out rough patches in market returns. Another difference is that I set a lower limit on the expected remaining life so that you never get to the point of spending 100% of savings in a last year. I also question whether 30 years is long enough, but maybe I'm just too optimistic about how long I'll live :-)

  2. It's hard to pick it out from the graph, but I'd be interested in what percentage of hypothetical retirees (of the 70 total) saw their spending drop at age 65, 70, 75, 80, 85, 90. Looks like everybody got back to the original $5000 by 90.

    1. @Greg: Good question. Here are the results:

      65: 28/70
      70: 22/70
      75: 18/70
      80: 14/70
      85: 13/70
      90: 0/70

      If we focus on only larger drops in income, here are the numbers for incomes below $4500/month:

      65: 20/70
      70: 16/70
      75: 15/70
      80: 13/70
      85: 7/70
      90: 0/70

      For those below $4000/month:

      65: 6/70
      70: 13/70
      75: 11/70
      80: 9/70
      85: 0/70
      90: 0/70

      These numbers make it tempting to add more safety margin. But it's hard to talk people into believing they need more saved by the time they retire.