Our Strong Theoretical Foundations

Research on Safe Withdrawal Rates

Moderator: hocus2004

Post Reply
JWR1945
***** Legend
Posts: 1697
Joined: Tue Nov 26, 2002 3:59 am
Location: Crestview, Florida

Our Strong Theoretical Foundations

Post by JWR1945 »

Although we generate numbers, that is not what we emphasize. We stress cause-and-effect relationships along with common sense. Our calculations add another dimension. They help us gain insight. They are not the foundation.

The observation that selling one's stock when prices are low is what kills retirement portfolios predates our own research. We accept this as fact without adding further proof. We see evidence of this when we look at historical sequences. We see evidence of this when we look at Gummy's equation for portfolio balances, which is also applied to calculate Safe Withdrawal Rates.

It is trivially obvious but exceedingly controversial that Safe Withdrawal Rates are influenced by purchase prices. It seems trivially obvious but exceedingly controversial and often denied that valuations affect portfolio safety during retirement.

Valuations are important because you control their influence. You can decide whether or not to buy an investment at a specified price.

We can separate Gummy's equation into two components, as he does. One part is the total return of the investments when there are no withdrawals. It does not depend upon the sequence of returns. It depends only upon the initial balance and the final balance without withdrawals. It is the product of gain multipliers. This first part is multiplied by another part that depends upon the sequence of returns and the magnitude of withdrawals as well as the overall return from investments.

Valuations clearly affect the first part of Gummy's equation. To the extent that valuations predict the overall return of the stock market, they tell us everything about the first component and quite a bit about Safe Withdrawal Rates in general. Yale professor Robert Shiller (along with Dr. Campbell) has established that P/E10 has such a predictive capability. It is a partial capability. It does not guarantee future outcomes. But it does remove a large amount of uncertainty.

[P/E10 is the current price of the S&P500 divided by the average of the previous ten years of earnings. Benjamin Graham had recommended looking at 5 to 10 years of trailing earnings when analyzing stocks.]

We have looked at a great variety of measures of valuations and determined how well they would have predicted the Historical Surviving Withdrawal Rates of the past. P/E10 is our best predictor so far. It is not the only valuation measure with a predictive capability.

We have used P/E10 not only to predict portfolio survival, but we have also used it to separate the effects of the sequence of returns (i.e., the second part of Gummy's equation) from the overall return of the stock market.

We have noted a difference in behavior of investments from earlier time periods to that of more recent times. We have found that commercial paper was once sufficiently profitable to guarantee a comfortable retirement. We have noted that P/E10 has had a much greater predictive capability in the modern era than in the days of the gold standard, very low federal taxes, a fledgling Federal Reserve system with minimal influence, a limited federal government and very little in the way of economic statistics.

[I will add that only those using Professor Robert Shiller's stock market data set have looked at these earlier years. Data from other sources typically start around 1926 or much later.]

We have found that making reliable predictions based on P/E10 requires very few years of data. Data from a single decade are useful. Adding more years helps. It refines our estimates and often reduces the degree of extrapolation needed when looking at other periods.

Shifting Allocations

The notion that one should change his investment allocations depending upon the relative attractiveness of the stock market dates back at least as far as Benjamin Graham's original writings. We have applied that notion in our studies of switching allocations according to P/E10. We have seen that this approach is tolerant of errors in allocations and in the thresholds for changing allocations. Straightforward comparison shows that it would have produced worst case Historical Surviving Withdrawal Rates of 5% as compared to the traditionally reported result of 4% when allocations are held constant.

The second part of Gummy's equation makes it clear that investment returns during the earliest years have the greatest importance in determining a portfolio's survival. Professor Shiller's research shows that P/E10 has a predictive capability in the medium-term: ten years later. This makes a compelling case that shifting allocations according to P/E10 should improve surviving withdrawal rates.

The foundation underneath these numbers is much stronger than a simple comparison of numbers. It goes back first to Gummy's equation and then to what we have learned from Professor Shiller's research.

We can top this off with our own empirical observations as to how much the sequence of returns influences survival as opposed to the total returns by themselves. The indicated improvement makes sense.

Stepping Aside

Today's market is still priced outside of the historical range of our calculators. We have looked at stepping aside until sanity returns to the stock market. We are waiting for stocks to become less expensive. We are not demanding that they become cheap.

This has led us to examine a variety of TIPS-based strategies. We have consistently found that they allow us to sit on the sidelines for a very long time, more than a decade, until things get better.

These are based on straightforward mathematical calculations.

In fact, as we look at realistic extrapolations for portfolio survival based on today's valuations, TIPS consistently produce higher Safe Withdrawal Rates. Even if stocks never return to pre-bubble valuations, TIPS-based strategies are still viable. They will pale by comparison only if a new, super-bubble follows the bubble.

If it were not for the bubble, TIPS-based strategies would have been vastly superior to high stock strategies in the late 1990s.

Dividend-Based Strategies

More recently, we have begun to look more deeply into dividend-based strategies. They appear to be especially attractive in today's environment. Broad-based assertions that high dividend stocks underperform the market are now known to be false. We should expect the advantages of dividend-based strategies to decrease as they become more popular because of arbitrage. But we should not assume that they will disappear entirely. Dividend-based strategies are attractive to one group of investors, retirees, but not necessarily to all investors.

Dividend-based strategies typically produce an income stream that lasts throughout the foreseeable future and that grows in buying power.

Dividends are highly predictable in the short and medium-terms. Dividend amounts are tied closely to several years of earnings, not to the results of a single year. We can apply the Gordon Equation to estimate the total return in the absence of withdrawals. We also know how the Gordon Equation can fail. Although dividend growth is somewhat predictable, it cannot be predicted reliably beyond just a few years. Predictions about dividends can be based on traditional measures of value such as earnings quality and asset values.

Very recently, we have applied our calculators by using TIPS as a surrogate to determine the conditions at which high dividend stocks become more attractive than the market as a whole (i.e., the S&P500 index).

Our approach is based on the solid foundation of Gummy's equation, which highlights the importance of the first few years, and other mathematical equations. We have added stock fundamentals. We have build upon this foundation. We have supplemented our knowledge by making additional calculations.

Have fun.

John R.
dbphd
* Rookie
Posts: 6
Joined: Tue Dec 24, 2002 8:33 am
Location: Montecito, CA

Post by dbphd »

I'm not sure why I'm left with the feeling this is less than a rigorous defense of market timing, perhaps the notion that if only stocks hadn't performed so well during the "bubble," i.e. if only there were not this data that doesn't fit our model, our case would be clearer, adds to my uneasy feeling. Identifying outlying data as a bubble seems a convenience that shouldn't be part of a rigorous explanation. A standard technique of evaluating a model is to look at its predictions at the extremes. Given our inability to predict the future, why shouldn't another "bubble" or "crash" be among the expect outcomes?

Sticking with an asset allocation and correcting the portfolio accordingly seems to have the common sense appeal of selling high and buying low, but I do realize that what appears to be comon sense can be wrong.

db
JWR1945
***** Legend
Posts: 1697
Joined: Tue Nov 26, 2002 3:59 am
Location: Crestview, Florida

Post by JWR1945 »

I wrote a one-paragraph overview related to market timing in Many Different Objectives dated Sun, Aug 22, 2004.
http://nofeeboards.com/boards/viewtopic.php?t=2904

For additional information, read Dynamic Portfolio Book dated Fri, Jul 02, 2004.
http://nofeeboards.com/boards/viewtopic.php?t=2725

Have fun.

John R.
hocus2004
Moderator
Posts: 752
Joined: Thu Jun 10, 2004 7:33 am

Post by hocus2004 »

JWR1945: "If it were not for the bubble, TIPS-based strategies would have been vastly superior to high stock strategies in the late 1990s."

dbphd: "Identifying outlying data as a bubble seems a convenience that shouldn't be part of a rigorous explanation."

The Data-Based SWR Tool is intended for use by buy-and-hold investors. For most buy-and-hold investors, TIPS-based stratagies WERE superior to high-stock stratagies in the late 1990s. The SWR for TIPS was at one point more than three times the SWR for an 80 percent S&P portfolio.

The effect of the temporary rise in prices experienced by stocks in the late 90s was to push the already low SWR for stocks even lower. That is not a benefit to the long-term buy-and-hold investor. It IS a benefit for short-term market timers. Short-term market timers will in some circumstances do better not making use of the SWR tool. But the tool was not intended for their use in any event.

Long-term buy-and-hold investors always benefit from knowing what the historical data says re the long-term income stream likely to be delivered by their investment asset classes. It is always possible that one of the two caveats will apply, that stocks may perform worse in the future than they ever have in the past (a pessimistic assumption) or that a worst-case scenario will not turn up in one's retirement (an optimistic assumption). But knowing the SWR always informs one's investment choices. And the data-based tool's "predictions" cannot fail so long as the caveats do not apply. So long as the caveats do not apply, the "predictions" are nothing more than OBJECTIVE REPORTS of things that have already happened.

The book is closed on retirements that began in January 2000. High-stock portfolios for such retirements possess extremely low SWRs. It is possible that worst-case scenarios will not pop up for those retirements. But even the best-case scenarios (of those that turned up in the past) are not appealing for high-stock retirements that began at the valuation levels that applied at that time. The only hope for retirees who are taking withdrawals in excess of those permitted even in best-case return sequences is that we will see a returns sequence from 2000 through 2030 more favorable than any we have seen in the past.

Is it possible? I believe that it is. My understanding is that the premise of the Jim Glassman book "DOW 3600" is that the longstanding equity premium is about to disappear, providing stock investors with exceptional returns in the near future (followed by exceptionally poor returns in the time-period following that). The data-based tool says that a 4 percent withdrawal might work for a high-stock retirement beginning in 2000 (but that it is not at all safe to presume that it will) and it is possible that even higher withdrawals will work.

The essential point remains, however. It is not "100 percent safe" to presume that things that have never happened in the past are going to begin happening on the day that your retirement begins. The study published at RetireEarlyHomePage.com assumes that stocks will perform in the future in ways that they have never performed in the past, that for the first time in history changes in valuation will have zero effect on long-term returns. This is an extremely dangerous premise on which to build one's hopes for a happy retirement. Those who claim that the take-out numbers generated by the conventional methodology are "100 percent safe" are at best uninformed as to what the historical data actually says and at worse are engaging in financial fraud that could cause severe life setbacks for early retirees for many years to come.
hocus2004
Moderator
Posts: 752
Joined: Thu Jun 10, 2004 7:33 am

Post by hocus2004 »

In the post immediately above, this sentence:

"But even the best-case scenarios (of those that turned up in the past) are not appealing for high-stock retirements that began at the valuation levels that applied at that time."

should instead read:

"But even the best-case scenarios (of those that turned up in the past) are not terribly appealing for high-stock retirements that began at the valuation levels that applied at that time.
Post Reply