Tuesday, May 15, 2012

Valuations and Withdrawal Rates


Having researched about the connection between market valuations and retirement withdrawal rates, I’ve long been aware of the incongruities of a long and acrimonious dispute about the subject spanning across a variety of personal finance Internet discussion boards. A fellow named hocus claims to have discovered the connection between valuations and “safe withdrawal rates” just over 10 years ago. But as with many other sources of dispute in this drama, I assumed that the relevant original discussions have been lost and cannot be checked for possible misinterpretation (as the Economist recently warned, much of our recent digital history risks being lost with the passage of time). It turns out that the discussion board thread which started this is still alive and well at the Motley Fool.

Much to my surprise, the “debate” that has supposedly been raging for the past 10 years was actually resolved just 84 minutes after it began. A fellow named intercst replied:

I'm not aware of anyone who has been successful in timing when stock prices are high or low over the long-term. If you possess this unique ability that others lack, perhaps the easiest thing to do is to adjust the safe withdrawal rate by the amount that you consider stocks to be higher than they have ever been. If stocks are 25% higher, drop the 4% withdrawal to 3%. If you feel stocks are 50% higher than they have ever been, then drop the 4% down to a 2% withdrawal.

Yes. If you want to keep using 4% it is up to you, but what was ultimately shown by people applying Campbell and Shiller’s regression methodology to this question is precisely this outcome suggested by intercst. This is the end of the debate. So what have people been discussing for the past 10 years? I think the answer lies in the misstep hocus took in his follow-up reply later that afternoon.  The first part of what he wrote sounds fine to me (except that 4% is not the average outcome):

Just to be clear on the nature of my concern, it is not that stock prices are higher than they have ever been before. My impression is that stock prices today are not much higher than what they were in the highest valuation years examined in the Safe Withdrawal Rate Study (years like 1969). Thus, I believe that the experience of investors who bought in those sorts of years provides clues--but only clues--as to what is likely to happen to investors investing at the prices prevailing today.

Analyzing what happened in any one year of high valuations does not provides much in the way of practical insight. If I believed that 1969 were the perfect analogy fo today, I wouldn't even consider purchasing a share of stock. The Safe Withdrawal Rate study tells me that, If I had invested $800,000 in stocks in 1969 and took out $32,000 to live on each year since, I would now have a portfolio value of zero. That's a wipe-out. The deficiency (in my view) of the existing studies is that they tell you what happened over the entire time-period from 1870 to the present (a 4 percent withdrawal rate applied "on average") and they tell you what happened to investments made in any individual year (like 1969), but they do not tell you what happened in a sufficienctly large number of years with characteristics similar to those that prevailed in years like 1969 so as to allow you to draw useful conclusions as to what the safe withdrawal rate might be for an investment made today.

1969 is just one year. So I am not willing to use the nunbers that applied for stock purchases made in 1969 to persuade me to rule out altogether stock purchases at times of high valuation. However, I view it as equally unrealistic to expect that an "average" safe withdrawal rate gathered from years in which stock prices were generally not comparable to those that prevail today can tell me the safe withdrawal rate for stock investments made now. To make informed investment decisions, I need a number that is drawn from a large enough set of data to provide meaningful information but that also provides an average of results from time-periods somewhat similar to the one I am now living through.

That’s the whole idea of lifecycle finance. Current conditions matter more than historical averages, stocks are risky even over the long term, and markets are unpredictable. Paul Samuelson grew so weary of explaining this concept over and over that he wrote an article in 1979 to explain this point using only one syllable words. It is why financial economists find it so perplexing to discuss the concept of a safe withdrawal rate. There isn’t one. The U.S., even since 1871, represents a rather unique period in world history. The worst-case that showed up during this time can hardly be expected to be representative of the future. Recently, I reviewed the international evidence on the 4% rule. My discussion included this table:
Table 3.2
Maximum Sustainable Withdrawal Rates, for Retirees 1900-1981
Asset Allocation: 50% Stocks & 50% Bills



Withdrawal Rate = 4%
Withdrawal Rate = 5%

SAFEMAX
SAFEMAX Year
10th Percentile
# Years in Worst Case
% Failures Within 30 Years
# Years in Worst Case
% Failures Within 30 Years
United States
3.96
1937
4.42
29
1.2%
19
41.5%
Canada
3.96
1937
4.58
29
1.2%
19
42.7%
New Zealand
3.78
1935
4.06
27
8.5%
18
42.7%
Denmark
3.65
1937
4.17
25
7.3%
19
62.2%
World
3.58
1937
4.09
23
3.7%
17
57.3%
Australia
3.5
1970
3.78
21
13.4%
15
39.0%
United Kingdom
3.36
1937
3.8
22
24.4%
16
48.8%
Sweden
3.22
1914
3.83
20
12.2%
14
41.5%
World Ex-US
3.21
1934
3.71
20
22.0%
13
52.4%
Switzerland
3.13
1915
3.42
19
34.2%
13
62.2%
South Africa
3.01
1937
3.81
21
15.9%
17
30.5%
Norway
2.98
1939
3.09
17
47.6%
12
69.5%
Netherlands
2.82
1941
3.76
17
22.0%
13
65.9%
Ireland
2.79
1937
3.15
19
36.6%
14
64.6%
Spain
2.16
1936
2.61
12
53.7%
10
87.8%
Belgium
1.58
1914
2.02
11
56.1%
9
73.2%
Finland
1.32
1917
1.8
6
43.9%
5
56.1%
Germany
1.01
1914
1.22
3
56.1%
3
70.7%
France
0.82
1943
1.31
7
75.6%
6
86.6%
Italy
0.8
1940
1
5
80.5%
4
86.6%
Japan
0.26
1937
0.29
3
37.8%
3
43.9%
Note: Assumptions include a 30-year retirement duration, no administrative fees, constant inflation-adjusted withdrawal amounts, and annual rebalancing.
Source: Own calculations from Dimson, Marsh, and Staunton (1900 - 2010) data.

The US was a special case. Also a special case, but the 4% rule failed in 80% of the historical episodes for Italy. Why should we expect any safe withdrawal rate? It is the unyielding commitment to what hocus wrote next which seems to have fueled his efforts for the past decade:

My proposal is to break the historical data into three categories: high valuation years; medium valuation years; and low valuation years. That way, when you determine the safe withdrawal rate that applies for a time-period like the one we are living in today, you are looking at a data set one-third as large as the data set used for the original study. I don't know what result looking at that data set would produce, but I am almost certain that it would show a safe withdrawal rate of less than 4 percent.

No, he just finished explaining why we don’t have enough data to have a good understanding about the full range of potential outcomes, and then he wants to torture the data more by trying to define different safe withdrawal rates for different valuation levels based on even more limited subsets of data? Though creative, it doesn't work. 

It is a subtle distinction to be sure, but in my research I never spoke of predicting a “safe” withdrawal rate (I may have gotten lazy at some point about terminology in my blog, but not in the research papers). I was instead focused on predicting as well as possible what the actual withdrawal rate would be. That could be perilous when market valuations are low, but my overwhelming emphasis was on recent years when withdrawal rates are more likely to fall below 4% than above 4%. I do think valuations may help gain insights about what the withdrawal rate will be, but this certainly does not make the estimates safer.

Hocus desperately wants someone besides him to say that the Trinity study needs to be "corrected," but I've explained that this isn't how research works. Rather, new studies with new methodologies come to replace old studies. This is a case, however, in which old studies were already available to suggest that we shouldn't project the findings of the Trinity study forward to future retirees.

At any rate, already a number of leading financial economists (including Zvi Bodie, Laurence Kotlikoff, Moshe Milevsky, William Sharpe, and others) have condemned the 4% rule. Financial economists do not respect the Trinity study for a variety of reasons such as: (1) current conditions are what matters, not historical averages, (2) the short historical period for one country used in the study gives us very little basis for being confident about 4% as safe, (3) strategies withdrawing fixed amounts (either in nominal or real terms) until wealth is depleted are not even a close approximation to optimal retirement spending strategies, and (4) the 4% rule ignores the role of other income sources outside the portfolio and ignores the flexibility and willingness for retirees to spend more now and potentially cut back later. 

William Bengen already made the very important and essential point that due to sequence of returns risk, the safe withdrawal rate is much lower than the 6 or 7% being bandied about when he wrote in 1994. That was really all we needed to see. After illustrating this point, the emphasis should have shifted almost exclusively to variable withdrawal strategies. That would have been the way to better match the planner's perspective about research with what finance academics had already figured out long before.

When viewed in this context, valuations are an interesting issue for retirement income, but they may not really be of central importance when building an overall retirement income strategy. That is because just as historical withdrawal rate outcomes don’t guarantee future safety, the relationship between valuations and withdrawal rates can change as well. We are still prone to black swans. Past is not necessarily prologue. Don’t treat the predictions coming from a model incorporating valuations as a guaranteed annuitization rate from your portfolio, especially when trying to extrapolate outcomes when valuations reach previously unseen levels. Trying to estimate safe withdrawal rates after incorporating valuations does not "correct" anything. End of story.

And don’t forget to account for fees, taxes, and whether your investment returns actually match the underlying indices used in research studies.

And so what retirees should do is to focus on building an income floor to ensure that basic needs will be met by construction no matter what happens (with Social Security, other pensions, bond ladders, and fixed annuities), and then to spend as they wish for discretionary expenses with whatever remains from their portfolio after building the floor.

Probably the two best ways to help middle class Americans to have a successful retirement are to make sure they don't leave money on the table by making suboptimal claiming decisions with Social Security, and by helping them to get their money to last as long as possible by shifting their funds into a low-fee diversified portfolio (we can debate about what the default asset allocation should be for people who don't care to customize, but maybe just using a fixed 50/50 throughout retirement is fine since this is supposed to be for discretionary expenses) and then to leave it alone except to withdraw and to rebalance. Update: The issue of best default advice for people who want to retire, but do not want to plan for it, probably deserves more attention than a throw-away paragraph at the end of an unrelated topic. I will return to this again. In the mean time, Steve Vernon recently provided his own take on these matters. I link to each of his "Just tell me what to do" series here.


12 comments:

  1. Outstanding article! Pulls together the ongoing discussion of withdrawal rates with what really matters: using the funds for retirement AFTER assuring basic needs are accounted for.

    ReplyDelete
    Replies
    1. Thank you. I'm now serving as curriculum director for the Retirement Management Analyst designation, and according to the curriculum, the fundamental goal of retirement planning is: First build a floor, then expose to upside!

      Delete
  2. So to help not make a suboptimal social security claiming decision could you clarify this from an earlier post?
    people should spend their financial assets down first and hold off Social Security until age 70 [this is the Series strategy]. Don’t just tap all your assets at the same time at retirement (if retiring early) [this is the Parallel strategy].
    What are the Series and Parrallel strategies that you refer to? I was unable to find them on Boglehead site.
    THanks
    Matt

    ReplyDelete
    Replies
    1. Matt,

      The basic idea here, which is another reason to ignore the 4% rule, is that while not 100% guaranteed, there is a very high probability that you will be better off by waiting to start Social Security and in the mean time spending more of your financial assets until Social Security begins (which allows you to spend less afterward). That is the parallel strategy.

      I still am not comfortable providing too specific of advice about Social Security. I need to master all of those rules better. The Center for Retirement Research at Boston College also has a lot of good information. You might like to start here:

      http://crr.bc.edu/special-projects/books/the-social-security-claiming-guide/

      Delete
    2. Matt,

      Sorry, I didn't mean to confuse you. In the comment I was describing the series strategy.

      I just put up a brief new blog post to illustrate the idea:
      http://wpfau.blogspot.jp/2012/05/extending-your-financial-wealth-by.html

      Delete
  3. Wade,

    I don't like this article. Can't you for once write an article which promises that I will be able to live off my wealth indefinitely with no problems whatsoever?

    Thanks.

    Interested reader

    ReplyDelete
    Replies
    1. Yes, funny!

      I wish I could! I could probably start a good business if I could provide such advice!

      Delete
  4. Wade --

    Working from your last paragraph, addressed to middle class Americans, could you add the inflation-adjusted SPIA into the arena?

    If SPIA income does not meet what the middle class American considers "needs," how does she choose allocation between the SPIA and the 50-50 portfolio?

    Dick Purcell

    ReplyDelete
    Replies
    1. Dick, good point. I updated the above to say more work is needed.

      Maybe we really need two sets of default advice: one for when Social Security provides enough to meet basic needs, and one for when it doesn't. I was thinking in terms of the first case.

      Delete
  5. Well, Wade, I guess the best thing I can say about your extensive quoting of the first Motley Fool guy... is that your blog traffic should go through the roof. But I appreciate the benefits of having the debate on your turf.

    I like Steve Vernon's articles too. Perfect for the military servicemember who's stuck in the desert or underwater.

    ReplyDelete
    Replies
    1. Thanks for the comment, Doug. I hadn't thought about how there would be a few extra links to this post to account for the nostalgia factor. I don't think there is much more to be seen here now though. As I already mentioned above, the debate ended 84 minutes after it started, and I'm not willing to re-engage with the usual talking points any further.

      Steve Vernon is definitely one of the good guys!

      Delete