Wade Pfau On Problems With Applying The 4% Withdrawal Rate Rule To Economies Beyond The U.S.

Sunday, May 06, 2012 09:19
edit
Wade Pfau On Problems With Applying The 4% Withdrawal Rate Rule To Economies Beyond The U.S.

Tags: retirement | retirement income | retirement planning

From an international perspective, a 4% real withdrawal rate is surprisingly risky. Even with some overly optimistic assumptions, it would have only provided "safety" in four of 17 developed countries studied, and a fixed asset allocation split evenly between stocks and bonds would have failed at some point in all 17 countries.

   

This Website Is For Financial Professionals Only


An argument in support of the 4% rule is that, despite what I have argued before, is that the U.S. post-1926 did suffer a number of calamitous market events, such as the Great Depression and the Great Stagnation of the 1970s. It's hard to imagine an even future retirees ever facing more difficult times. Thus, the historical success of the 4% rule suggests that we can reasonably plan for its continued success in the future.

 

My first foray into researching about personal retirement planning was a study about the sustainability of the 4% rule in other developed market countries. Here I will provide further perspective about the results described in my article, An International Perspective on Safe Withdrawal Rates: The Demise of the 4% Rule? in the December 2010 Journal of Financial Planning.  

 

From an international perspective, a 4% withdrawal rate has been problematic. In the original article, to avoid claims of bias that I choose an asset allocation which exaggerates the risk of the 4% rule, I used a rather generous and unrealistic “perfect-foresight assumption.” For each country and in each retirement year, I used the asset allocation (between stocks, bonds, and bills) which supported the highest withdrawal rate. Though not always, in many cases this meant using 100% stocks. With the perfect-foresight assumption and an updated dataset which includes two more countries, the calculated SAFEMAX (this is the maximum sustainable withdrawal rate over 30-years in the worst-case scenario from a country's history) exceeds 4% in Canada (4.4%), New Zealand (4.1%), Sweden (4.1%), Denmark (4.1%), the GDP-weighted world portfolio (4.1%), and the US (4%).   The perfect foresight assumption is quite unrealistic, though, and here I will focus on a more plausible 50/50 retirement asset allocation. In terms of the SAFEMAX, 50% stocks and 50% bills generally outperforms 50% stocks and 50% bonds. For this reason, I will further consider results for the stocks and bills case. Though not shown here, I do also note that stock allocations below 50% support lower SAFEMAXs in all countries except Switzerland and Sweden. The results for 50/50 are in Table 3.2.

 

 

 

 

 
 
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.

 
 

With a 50/50 asset allocation, the 4% rule did not survive in any country, though it came close in the U.S. and Canada, with support for 29 years of expenditures and a SAFEMAX just below 4%. Even allowing for a 10% failure rate, 4% made the cut only in Canada, the US, New Zealand, Denmark, and the GDP-weighted world portfolio. In 10 of 19 countries, the SAFEMAX fell below 3%. World War II era Japan, in particular, faced the sort of crisis suggested by William Bernstein, as the SAFEMAX was only 0.26% for 1937 retirees. The 4% rule would have supported expenditures for only 3 years. Shockingly, the 4% rule would have failed more than half of the time for countries including Spain, Belgium, Germany, France, and Italy. Italians attempting to use the 4% rule would have actually faced failure in 80.5% of cases. The table also shows results for a 5% withdrawal rate, and failures rates are substantially higher than for 4%. Even if the 4% rule could somehow be deemed as safe, there is clearly not much room for error when seeing how quickly failure rates rise for 5% withdrawals. I do not use hyperbole when suggesting that the results in this table do not portray the 4% rule in a positive light.

 

To expand the previous table further, Figure 3.2 shows a boxplot of the distribution of withdrawal rates for each country, ranked by order of their SAFEMAX from smallest to largest. For each country, the red central marker is the median, the edges of the box are the 25th and 75th percentiles, and the whiskers extend to the most extreme datapoints not defined as outliers. Outliers are plotted individually. This figure does provide a broader view about the range of outcomes for each country.

 

 

 

 

 

Figure 3.3, meanwhile, shows the distribution of withdrawal rates across countries for each retirement year. From this figure, we can see the contemporaneous correlation of withdrawal rates across countries and also observe general historic trends. In recent years, those who have challenged the 4% rule are sometimes accused of falling victim to recency bias, placing too much weight on the poor financial market conditions of the recent past. However, Figure 3.3 suggests that recency bias may instead be responsible for too much faith in 4%. Across countries, the median sustainable withdrawal rate was under 4% in many cases prior to about 1945. Since World War II, however, the median outcome never fell below 4%, and in only a few cases did the 25th percentile fall below 4%. Indeed, in recent years the performance of the 4% rule across countries was much stronger than the earlier part of the historical period. While some of the worst outcomes can be connected with World Wars I and II, given the broader view of history suggested by William Bernstein, is it appropriate to ignore those cases? Whether there are important structural changes which might explain why the earlier period is less relevant to today’s environment is an exercise to be left for the reader’s interpretation.

 

 

 

 

Keep in mind also that I am only looking at the 19 developed market countries in the dataset, with data going back to 1900. Travel back in time to 1900, though, and ask people to put together a list of 19 developed market countries for the 20th century, and you would probably find frequent mention of countries like Argentina, Russia, and China, among others. As those countries never made the dataset, even the results I describe here include survivorship bias.

 

From the perspective of a U.S. retiree, the issue is whether the future will provide the same asset return patterns as in the past, or whether Americans should expect mean reversion that would lower asset returns to levels more in line with what many other countries have experienced. And for international readers, do keep in mind that the 4% rule is based on U.S. historical data, and mileage may vary quite dramatically in other countries.

 

It may be tempting to hope that asset returns in the 21st century United States will continue to be as spectacular as in the last century, but John Bogle cautioned his readers in his 2009 book Enough, “Please, please, please: Don’t count on it.

 

This speaks to the RIIA's fundemental goal of retirement planning to first build a floor and then expose to upside. 

 

 

Wade Pfau, Ph.D., CFA, is an associate professor of economics at the National Graduate Institute for Policy Studies (GRIPS) in Tokyo, Japan, and the curriculum director for the Retirement Management AnalystSM designation program. He maintains a blog about retirement planning research at

wpfau.blogspot.com

 

 

 

 

 


Comments (5)

...
brentb843
Wade - lots of neat charts; however, your conclusion, as with Monte Carlo, etc is only valid IF returns become symmetrical and repeat on the basis of 12 or 365 (or whatever number of actual trading days).

Research has shown this is not valid and we need a new manner to answer the retirement income success rate. I would recommend you consider Omega combined with correct bootstrapping methodology to answer the real question -

based on all the returns known (either monthly or daily) what mix of assets has the greatest probability of achieving the distribution at the lowest risk (downside).

Right now, the problem is framed incorrectly asking what random mix of historical calendar year returns will maintain 4%?

Our research shows framed correctly, a 7% distribution rate can be achieved with a nearly 10% annualized rate of return.
brentb843 , May 09, 2012
...
RIIA
Thanks for commenting. Do you mean Scott Burns' Omega strategy? I have been meaning to simulate that soon.

And what do you consider to be the correct way to do bootstrapping?

I'm interested to understand more about what you are describing. Thanks, Wade
RIIA , May 12, 2012
...
brentb843
No. Omega is a mathematical formula...see Shadwick and Keating.

Effron has the only prescribed and validated bootstrap model for the task you describe.

You don't need to model Burns he used a discreet interval to suggest it will always work.



brentb843 , May 14, 2012
...
RIIA

Thanks. Though I didn't know the name Effron, after looking it up, it does seem that this is what I think of as normal bootstrapping.

Actually, I did recently write about the point I think you are making, which is that historical simulations are not really very ideal for these questions.

http://wpfau.blogspot.jp/2012/05/monte-carlo-simulations-vs-historical.html

Thanks, Wade
RIIA , May 15, 2012
...
RIIA
Brent,

I've been reading up a bit on what you are describing. I'm not sure I follow all your steps, but is the basic idea to choose the asset allocation that will support the target rate of return with the lowest downside risk?

It seems that this sort of approach may be too closely connected with modern portfolio theory to be helpful with retirement planning.

On the accumulation side, people do not need to focus on a target rate of return, because they do not need to focus on meeting a wealth accumulation target. People care about income for life, not about wealth. And the income supported by a given amount of wealth will vary dramatically as interest rates (and possibly valuations) change.

As for retirement spending, retirees don't need to worry about a target rate of return. They want to spend sustainably from their savings, and the probability of this happening is closely related to best combination of returns and volatility, which must be addressed jointly.

If you think I am missing your point, I would appreciate further elaboration.

Bootstrapping may be better than Monte Carlo with a lognormal distribution, since it preserves more of the underlying characteristics of the data and doesn't require any parametric assumptions, but in practice I've found that the results using each are almost indistinguishable.

And how does your approach deal with the issue that their may not be a stationary underlying distribution anyway? If not, then bootstrapping will still be garbage in, garbage out.

My point with this article is not to identify safe withdrawals for different countries, but to just show how little we know about safe withdrawal rates. This implies that retirees should focus on building a floor and then exposing to upside, as is RIIA's mantra.

Thanks, Wade
RIIA , May 15, 2012

Write comment

You must be logged in to post a comment. Please register if you do not have an account yet.

busy