Tuesday, December 15, 2009

An Experiment with the Modified 4% Rule

Recently we discussed a modified 4% rule where we base the size of monthly withdrawals from retirement savings on the current portfolio size rather than the portfolio size at the beginning of retirement. Let’s follow this up by looking at actual market data to see how monthly retirement income is affected.

Because U.S. data is more readily available, I’ve gathered returns on the S&P 500 (including dividends) since 1988, adjusted for inflation. Let’s consider the case of Rita who retired in 1988 with a portfolio worth $750,000 in today’s dollars. The actual figure in 1988 was a little over $400,000.

Rita has 100% of her money in the S&P 500 index. This is obviously a more aggressive portfolio than most retirees would want, but let’s see what happens to Rita’s income.

Using the modified 4% rule, Rita would start drawing 1/12 of 4% ($2500 in today’s dollars) per month. But this amount gets adjusted each month depending on whether the portfolio grows or shrinks. Assuming that Rita managed to stay alive for nearly 32 years, here is how her income changed over the years:

Rita had only one month where her inflation-adjusted income dipped below $2500 to $2479. By the year 2000, the buying power of her income had tripled. However, this was followed by one big drop from 2001 to 2003, and then another big drop in early 2009 where it reached $2711.

It seems that Rita fared rather well over the years considering that her income started at $2500 per month. However, if her lifestyle had expanded to consume her income when it was over $7500, she may have found the subsequent drop painful.

Things would have been far different for Ray who retired in 2001. Just two years later his retirement income was chopped in half. This may seem to be an argument against modifying the 4% rule, but if Ray had stuck to making monthly withdrawals based on his 2001 portfolio size, he would run out of money fast. Ray’s situation isn’t so much an indictment of the modified 4% rule as it makes it clear that a portfolio 100% in stocks is risky.

One thing that this experiment has left out is that it makes sense for retirees to increase the percentage of their portfolio that they spend as they get older. If Rita turned 65 in 1988, then she is 96 now with a portfolio worth over $1.1 million as of the end of October. She can probably afford to spend more than 4% each year now. How much more is an interesting question.

[Update: Reader Blitzer68 pointed me to an excellent article by William Bernstein on this subject that I highly recommend.]


  1. The Stingy Investor has a nifty calculator with lots of data.


  2. Or you can decide not to retire, or force your children to pay out to you like Larry MacDonald pointed out!

  3. Hazy: That's an interesting site. I'd love to have the raw data behind their calculator so that I could do my own analyses.

    Big Cajun Man: Not retiring doesn't sound good to me. Sponging off my kids sounds much better.

  4. CC: Thanks for the pointers. Schiller's monthly S&P 500 data look very useful. I'd love to have similar data for Canada, and for bonds in both countries. Even better would be daily data, but that seems unlikely to get.

    I can't imagine how one could conclude that asset allocation in retirement doesn't matter much. In the extreme, you'd think that 100% bonds and 100% stocks would give different expected results and different risks.

    I agree that 4% withdrawals are risky if the percentage is based on the starting portfolio size, but it should be less risky if based on current portfolio size.

    Basing income on yield is what many people seem to do. This should fluctuate less than portfolio size, but can still fluctuate. This idea might go out of favour in periods of very low dividend yield, like 2000-2001. However, if dividend yield is predictive of the direction of stock prices, yield may be the best choice to base withdrawals on.

    1. The comment above is a reply to Canadian Capitalist's comment:

      @Michael: The Stingy Investor calculator is based on this spreadsheet:


      I don't know where you obtained the US data but Robert Schiller's website has monthly S&P 500 data going back, well, a long time.


      I haven't put much thought into portfolio withdrawal but I read somewhere (don't remember where) that in the withdrawal stage, asset allocation doesn't matter much. A poor sequence of returns could decimate a stock portfolio and low returns decimate bond portfolios. What does matter is the withdrawal rate and 4% might be too risky under some sequence of poor returns.

      Personally, I'm thinking a withdrawal rate equal to the portfolio yield (dividend plus interest), provided the stock portion doesn't reach for yield should be close to bullet proof. I haven't studied this in any detail though.

  5. This whole topic was covered quite well by William J. Bernstein a while back in his "Retirement Calculator from Hell" series of articles. I would highly recommend it. From Google, search for "site:efficientfrontier.com hell".

    I would be interested, Michael, if you have anything to add to his treatment on this topic.

  6. Blitzer68: I hadn't read that piece by Bernstein before. It is very good. I've decided to add a reference to it in the main body of my post. I don't have anything useful to add yet, but I'm still thinking about it.