SWRs versus Returns (Updated 7/14/08)This article was inspired by a discussion on the Vanguard Diehards forum at Morningstar.com. Click HERE. The good stuff starts at around post #338.
SWRs and 30-Year Returns Fortunately, I already have spreadsheets to look at the historical data. Stocks are the S&P 500 (large stocks) and bonds are 5-year Treasuries. Data is from Ibbotson. Most of it can be found at Gummy's site HERE.
I'll start with this chart:
Maximum "Safe" Initial Withdrawal Rates for years 1928-2000 
This chart was created by Gummy's Max Rate Withdrawal Spreadsheet (xls)
Then, I looked at returns.
30-Year Rolling Returns 1926-2005 (Nomimal Values) 
This chart was created by my Allocation Spreadsheet (xls)
30-Year Rolling Returns 1926-2005 (Real Values) 
This chart was created by my Allocation Spreadsheet (xls)
And finally, here is the raw data for the charts above.
Start
YearMax
SWRNominal
ReturnsReal
Returns1926 - 7.4% 6.0% 1927 - 7.2% 5.7% 1928 6.3% 6.5% 4.8% 1929 5.3% 6.4% 4.7% 1930 5.6% 6.7% 4.9% 1931 6.1% 7.2% 5.1% 1932 7.6% 8.6% 6.2% 1933 7.2% 8.6% 5.7% 1934 5.9% 8.1% 5.2% 1935 6.1% 8.3% 5.4% 1936 5.1% 7.7% 4.8% 1937 4.5% 7.0% 4.1% 1938 5.8% 8.0% 5.1% 1939 4.9% 7.7% 4.5% 1940 5.0% 7.5% 4.1% 1941 5.4% 7.9% 4.4% 1942 6.6% 8.5% 5.2% 1943 6.8% 8.6% 5.4% 1944 6.4% 7.9% 4.6% 1945 6.2% 7.1% 3.5% 1946 5.5% 7.2% 3.4% 1947 7.0% 8.0% 4.6% 1948 7.8% 7.7% 4.4% 1949 8.2% 7.8% 4.3% 1950 7.7% 7.8% 3.8% 1951 7.3% 7.9% 3.6% 1952 7.2% 7.5% 3.2% 1953 6.9% 8.0% 3.6% 1954 7.2% 8.5% 3.9% 1955 5.8% 7.9% 3.3% 1956 5.2% 8.3% 3.5% 1957 5.4% 8.7% 4.0% 1958 5.8% 8.9% 4.1% 1959 5.1% 8.6% 3.7% 1960 5.0% 9.1% 4.2% 1961 5.0% 9.0% 3.9% 1962 4.5% 9.3% 4.1% 1963 4.8% 9.6% 4.4% 1964 4.5% 9.6% 4.3% 1965 4.2% 9.1% 3.8% 1966 4.2% 9.8% 4.4% 1967 4.6% 10.3% 4.9% 1968 4.3% 10.6% 5.3% 1969 4.3% 10.9% 5.7% 1970 5.0% 11.4% 6.3% 1971 - 11.1% 6.1% 1972 - 10.6% 5.6% 1973 - 9.9% 5.0% 1974 - 10.7% 5.9% 1975 - 11.3% 6.9% 1976 - 10.7% 6.4% And here are some more charts created from the raw data:
SWRs and P/E10 Later in that same Diehards conversation at post #356, a poster brought up the subject of P/E10 and SWRs. Click HERE. This subject has come up numerous times and on many message boards over the years. There have been a few studies examining this phenomenon. ( See note 1 )
Let's take a closer look. Gummy has P/E10 data in the form of E10/P which is contained in his spreadsheet ( E10-P.xls ). I have combined Gummy's E10/P data with Max SWR data in the following table:
Start
YearMax
SWRE10/P
(January)1928 6.3% 5.3% 1929 5.3% 3.7% 1930 5.6% 4.5% 1931 6.1% 6.0% 1932 7.6% 10.7% 1933 7.2% 11.5% 1934 5.9% 7.7% 1935 6.1% 8.7% 1936 5.1% 5.9% 1937 4.5% 4.6% 1938 5.8% 7.4% 1939 4.9% 6.4% 1940 5.0% 6.1% 1941 5.4% 7.2% 1942 6.6% 9.9% 1943 6.8% 9.9% 1944 6.4% 9.0% 1945 6.2% 8.4% 1946 5.5% 6.4% 1947 7.0% 8.7% 1948 7.8% 9.6% 1949 8.2% 9.8% 1950 7.7% 9.3% 1951 7.3% 8.4% 1952 7.2% 8.0% 1953 6.9% 7.7% 1954 7.2% 8.3% 1955 5.8% 6.3% 1956 5.2% 5.5% 1957 5.4% 6.0% 1958 5.8% 7.3% 1959 5.1% 5.6% 1960 5.0% 5.5% 1961 5.0% 5.4% 1962 4.5% 4.7% 1963 4.8% 5.2% 1964 4.5% 4.6% 1965 4.2% 4.3% 1966 4.2% 4.2% 1967 4.6% 4.9% 1968 4.3% 4.6% 1969 4.3% 4.7% 1970 5.0% 5.9%
From this data, I created this chart:
10-Year Treasuries and P/E10 While doing research for this article, I came across another article written by Gummy. It is called P/E Ratios and Bonds Rates. This chart from that article also shows a correlation to E10/P:
There is no question that the discovery of P/E10 and SWRs has generated considerable debate on several discussion boards. However, I have noticed that those who promote the P/E10 view also adopt the notion that one should use P/E10 as the correct way of valuing the stock market and specifically the S&P 500. Their contention is that today, the stock market is at P/E10 of 26. P/E10 data is maintained at Robert Shiller's website HERE. (Click on the "Stock Data" tab at the bottom of the page.)
Adopting P/E10 as the correct way of valuing the stock market ignores several other considerations.
Conclusion
- We can acknowledge that valuations for the S&P 500 were indeed at extreme levels during the tech bubble and crash. Some investors also point to Robert Shiller's pessimistic 1996 prediction that stocks might have returned nothing for 10 years. However, Shiller did provide some caution:
The conclusion of this paper that the stock market is expected to decline over the next ten years and to earn a total return of just about nothing has to be interpreted with great caution. Our search over economic relations that us to study the price divided by 30-year moving average of earnings may have stumbled upon a chance relation with no significance. In other words, the relation studied here might be a spurious relation, the result of data mining. Neither the statistical tests nor the monte carlo experiments take account of the search over other possible relations.
It is also dangerous to assume that historical relations are necessarily applicable to the future. There could be fundamental structural changes occurring now that mean that the past of the stock market is no longer a guide to the future.
- Despite that dire prediction, the S&P 500 actually returned 9.1% (annualized) during the 10-year period of 1996 through 2005. Small stocks returned a similar amount at 9.3%. If someone abandoned the stock market and instead put all their money in "safe" investments, they would have earned 5.1% in intermediate government bonds and 3.8% in 90-day Treasury bills. Data is from my spreadsheet HERE.
- One could argue that in order for an investor to have reaped that nice 9 percent 10-year return, he would have had to suffer a brutal 2000-2002 bear market. That is very true.
However in hindsight, we can see that there were other areas of the market such as value stocks, small cap stocks and REITs that were not caught up in the bubble. For an investor holding a portion of their money in these other asset classes, their portfolio did not suffer as much volatility. For a more detailed discussion of "slice and dice" in a retirement portfolio, you can read an article I wrote HERE.
- With the possibility of including different asset classes in a retirement portfolio, the notion of analyzing SWRs with P/E10 may need rethinking. For example, what if instead of using the S&P 500, we used value stocks? How would having value stocks change Max SWRs?
Maximum "Safe" Initial Withdrawal Rates for years 1928-2000 using Large Value
This chart was created by Gummy's Max Rate Withdrawal Spreadsheet (xls)
From the chart, we see that using Large Value instead of the S&P 500, Max SWRs were higher. In fact, the "Safe" Withdrawal Rate became 5.3% using Large Value. This in contrast with the "Safe" Withdrawal Rate of 4.2% when using the S&P 500.
- There is one final consideration when using P/E10 as the basis of valuing the market. And that is whether it makes sense. I find it hard to believe that E10 -- the average of real earnings of the previous 10 years -- makes for a credible metric. It just seems strange. So out of curiosity, I created this chart:
Data for this chart is from Robert Shiller's Database.
(Click on the "Stock Data" tab at the bottom of the page.)
A more common approach to value the market is just the regular P/E ratio. Most investors look at either the prior 12 month's earnings or an estimate of future 12 month earnings. The P/E based on prior 12-month earnings can be easily found on financial websites such as Yahoo Finance. Click HERE. The P/E based on future 12-month earnings estimates can be found at a website like Standard and Poors. Click HERE.
And here is a chart displaying the P/E ratio based on 12-month trailing earnings. Both "as reported" earnings and "operating" earnings are shown.
Source: Vanguard.comI doubt that this article will change minds. I acknowledge that there is a deep feeling in two investing philosophies. The first is embodied by John Bogle's view that the Total Stock Market (TSM) is the best way for investors. From this viewpoint, a "slice and dice" portfolio offers no advantage.
The second investing philosophy is that investors should be varying their equity-fixed allocation based on how high or low the stock market appears to be valued. This is the reason why followers of P/E10 have caused so much debate. After all, with P/E10 at supposedly high levels, the logical action is to lower one's stock allocation until P/E10 returns to "moderate" levels. From this mentality, an investor is constantly bracing himself for a bear market lurking just around the corner.
The sad part of this mentality is that based on P/E10, stocks have been "overvalued" and primed for a fall for the past 10 years. Stocks did fall in 2000 and then rise again in a new bull market in 2003. Unfortunately, an investor looking only at P/E10 would have missed that entry point. Who knows whether P/E10 will return to "moderate" levels at the end of the next bear market. One can only wonder how patient an investor needs to be while sitting on a lowered equity allocation.
And finally, a word to retirees. Be flexible. While a 4% initial withdrawal rate is a good guideline, nothing is set in stone. For more ideas on being flexible, read my article Variable Withdrawals in Retirement.
SWRs and Initial Returns Later in that same Diehards conversation at post #370, a poster wondered if Max SWRs depend on the sequence of returns. Click HERE. Sequence of returns may have had some effect. In fact, I wrote an article about this topic called Reverse Dollar Cost Averaging. The cure for this situation could be a buckets strategy. This is where funds are withdrawn from the less-risky part of the retirement portfolio first. Then after stocks recover, the portfolio is rebalanced.
But this question lead me to think that instead of confronting a poor initial sequence of returns, a retiree may have faced a just plain poor initial returns. This could have caused the Max SWRs to be lower than if returns started strongly.
So I went back and looked at initial returns. I started with the first 5-year returns (using real values). I kept the same parameters: 50% S&P 500 and 50% 5-year Treasuries. Here is a chart:
So we see that there is some correlation with poor initial returns. It is certainly better correlation than when we looked at the full 30-year period before.So out of curiosity, I went back and lengthened the initial returns. Here is a table with the results:
Number of
Initial YearsR-Squared 5 0.42 6 0.49 7 0.57 8 0.64 9 0.70 10 0.68 11 0.74 12 0.79 13 0.85 14 0.84 15 0.80 16 0.76 17 0.73 18 0.71 19 0.67 20 0.60 So it looks like Max SWRs do correlate to real returns after all. It seems to be strongest for the first 13 years. Here is the chart:
You may also be interested in seeing a chart of 13-year rolling returns.
13-Year Rolling Returns 1926-2005 (Real Values) 
This chart was created by my Allocation Spreadsheet (xls)
Now lest you think we've discovered an important insight, I will be the first to admit that the preceding was an exercise in data mining. I do not want the reader to come away with the conclusion that 13-year initial real returns can predict Maximum SWRs.
What we likely have seen is that a combination of factors, such as stock and bond returns along with inflation, all had a hand in producing the results we have just seen.
Real Returns and Inflation The topic of SWRs came up again on the Vanguard Diehards forum at Morningstar.com. Click HERE.
The point was made that the high inflation of the 1970's was a major factor for producing the poor returns (in real terms) for large stocks following the 1966 peak where P/E10 reached a high of 24.
10-Year Rolling Returns 1926-2005 (Real Values) 
This chart was created by my Rolling Returns Spreadsheet (xls)
As you can see from the chart above, there was a "black hole" during the 1970's period of high inflation. You will notice a similar "black hole" in the 13-year chart shown in the previous section. So even though maximum SWRs correlated well with initial returns, a major reason was because high inflation pulled down real returns.
Why did this happen? Here is an edited version of what I wrote on that Diehards thread:
In 1966, P/E10 was 24. By 1982, P/E10 had fallen to 7. However, the reason stocks fell so far out of favor was not merely because valuations were "high" at the start of the period. It was also because inflation ran at a high rate. This, in turn, forced interest rates higher. So with the relative attractiveness of fixed income enticing investor dollars during that period, the attraction of stocks waned. Furthermore, the Shiller stock database did not operate in a stable environment. One cannot lump peak P/E10 years of 1901, 1929 and 1966 AND TODAY IN 2006 together as if the before and after environments were similar. They were not.
If it wasn't for the high inflation period of the 1970's, P/E10 level of 24 could have been the norm for the previous 40 years. That is my point.
In summary, it would be fair to describe stocks as two different asset classes. Stocks purchased during a period of low inflation perform very differently from stocks purchased during a period of high inflation.
By the way, the same can be said for bond yields as shown in the chart below.
Click on image to enlarge.
~~~ Later in that thread, there was a discussion about changes in P/E10. What does a change from P/E10 of 24 in 1966 to P/E10 of 7 in 1982 mean? "P" represents price only without reinvesting dividends. If an investor had reinvested dividends, then the Real Total Return would have been flat.
To show what this looked like, below is a chart starting with $1000 in 1966.
Growth of $1000 1966-1985 (Real Values) 
This chart was created by my Mountain Spreadsheet (xls)
Note 1 The following articles are historical studies that have examined SWRs and P/E10. E10/P and SWR by Gummy
Safe Withdrawal Rates versus Valuations by Gummy
Safe Withdrawal Rates and P/E Ratios by John Greaney
Resolving the Paradox -- Is the Safe Withdrawal Rate Sometimes Too Safe? (pdf) by Michael Kitces
For those unfamiliar with P/E10, here is a description. Instead of using 12 months of earnings, P/E10 uses an average of the previous 10 years of earnings.
Return to the Text
