Any Studies on Long-term Market Timing?

Discuss all general (i.e. non-personal) investing questions and issues, investing news, and theory.
User avatar
wade
Posts: 539
Joined: Fri Sep 17, 2010 12:38 am
Location: Main Line
Contact:

Any Studies on Long-term Market Timing?

Post by wade » Thu Jan 20, 2011 3:42 am

May I ask about long-term market timing strategies?

I know that there is an extensive literature about the predictability of long-term stock returns dating back to Campbell and Shiller's work in the mid-1990s.

I also know that there is an extensive literature about short-term market timing strategies. Don't worry, I have enough background with the efficient market hypothesis to dismiss short-term market timing strategies for my own personal investment situation.

But my question is about LONG-TERM market timing strategies. In other words, using market timing over periods of at least 10 years to obtain better returns than a buy and hold strategy. The literature seems slim. I will list what I know about, and I hope someone might fill in more details. I am particular interested, at this point, in studies which reject long-term market timing.

Here's the short list of what I have been able to find:

Ben Stein and Phil DeMuth, "Yes, You Can Time the Market", 2003
strong support for long-term market timing

Norbert Shlenker has a posting at Financial Webring forum on this topic
my reading based on his evidence is that it is favorable, but he
concludes the opposite

Meir Statman and Ken Fisher, "Market Timing at Home and Abroad" Journal of Investing, 2006

Meir Statman and Ken Fisher, "Market Timing in Regression and Reality" Journal of Financial Research, 2006

both of these studies conclude that long-term market timing is not a
feasible strategy

Anything else? Thank you!
Last edited by wade on Fri Jan 21, 2011 3:30 am, edited 1 time in total.

cjking
Posts: 1742
Joined: Mon Jun 30, 2008 4:30 am

Post by cjking » Thu Jan 20, 2011 5:54 am

In "Valuing Wall Street" there is a suggestion to get out of shares altogether when valuations are at a premium (say 50% too high, but the figure is not critical) and get back in once fairly valued.

Although this showed an improvement in backtesting at the time, I'm not sure the improvement will look as good now that another ten years of data is in. The slow deflation of the 2000 bubble (still going on) would mean being out of stocks for a long time. I think a better strategy is to prefer other assets when stocks are unattractive relative to current alternatives, rather than when they are unattractive compared to their own historical or (predicted) future prices.

Regardless of whether it serves your purpose, "Valuing Wall Street" is worth reading anyway. Probably the most rigorous and single-minded investment book aimed at the general public there is, in the way it beats one particular very important topic to death.

(Edited to refer to correct book!)
Last edited by cjking on Thu Jan 20, 2011 6:57 am, edited 2 times in total.

User avatar
wade
Posts: 539
Joined: Fri Sep 17, 2010 12:38 am
Location: Main Line
Contact:

Post by wade » Thu Jan 20, 2011 6:36 am

cjking wrote:In "Wall Street Revalued" there is a suggestion to get out of shares altogether when valuations are at a premium (say 50% too high, but the figure is not critical) and get back in once fairly valued.


Thank you for the recommendation. Just to clarify, Andrew Smithers has two similarly titled books. "Wall Street Revalued" came out in 2009. "Valuing Wall Street" came out in 2002. Are you recommending the newer one? It looks like they are both about the same general topics, and they both look quite relevant for me. I am going to order them both. Thanks so much.

cjking
Posts: 1742
Joined: Mon Jun 30, 2008 4:30 am

Post by cjking » Thu Jan 20, 2011 6:55 am

Sorry, brain failure here, I meant to recommend the older one, "Valuing Wall Street."

I also have both. (They aren't the same.)

richard
Posts: 7961
Joined: Tue Feb 20, 2007 3:38 pm
Contact:

Post by richard » Thu Jan 20, 2011 7:00 am

The problem with studies of long-term strategies is that there isn't enough data for any conclusion to be statistically significant. There at most 10 reliable independent data points for 10 year returns.

If conditions are meaningfully different today than in the earlier years, then you have even fewer data points. In addition to the obvious differences in conditions, risks, government actions, etc., over the past 100 years, investors today are aware of the past and take it into account in investing.

You appear to believe market efficiency rules out short term timing but not long term timing. If so, what's the basis for this belief?

User avatar
Taylor Larimore
Advisory Board
Posts: 26125
Joined: Tue Feb 27, 2007 8:09 pm
Location: Miami FL

Market timing ?

Post by Taylor Larimore » Thu Jan 20, 2011 7:52 am

Hi Wade:

My question is about LONG-TERM market timing strategies. In other words, using market timing over periods of at least 10 years to obtain better returns than a buy and hold strategy.


Long-term, intermediate-term, short-term, I doubt if it makes any difference. This is what Mr. Bogle wrote:

"After nearly fifty years in this business, I do not know of anybody who had done it (market timing) successfully and consistently. I don't even know anybody who knows anybody who had done it successfully and consistently."


As a former market-timer myself, I urge you to forget about it.
"Simplicity is the master key to financial success." -- Jack Bogle

User avatar
wade
Posts: 539
Joined: Fri Sep 17, 2010 12:38 am
Location: Main Line
Contact:

Post by wade » Thu Jan 20, 2011 7:57 am

richard wrote:The problem with studies of long-term strategies is that there isn't enough data for any conclusion to be statistically significant. There at most 10 reliable independent data points for 10 year returns.

I didn't really mean to make this into a thread discussing the merits of long-term market timing, and I would feel more comfortable discussing it after doing more research first. But I can offer some ideas about your points.

First, this issue of overlapping observations. If you have 130 years of data, then that means you have 120 overlapping 10 year periods, or 13 independent 10-year observations. I know 120 overlapping 10-year periods is not the same as 120 independent observations. I'm not comfortable with theoretical statistics, and my intuition may be wrong, but my intuition is that 120 rolling 10-year periods still provides more information than does 13 independent observations. Is your idea based on some theoretical findings?

richard wrote:If conditions are meaningfully different today than in the earlier years, then you have even fewer data points. In addition to the obvious differences in conditions, risks, government actions, etc., over the past 100 years, investors today are aware of the past and take it into account in investing.

You have two main points here, and I accept them both as important caveats. But just a few observations:

1. I reject market-timing as being either 100% stocks or 100% Treasury bills. One example of what I am thinking of has been attributed to Benjamin Graham, though I haven't yet verified whether this is an urban legend. He seems to have advocated a 50% stock allocation that could be adjusted between 25% and 75% depending on market valuations. That type of strategy would provide some protection in case conditions did change.

2. Though I haven't seen a list of the tenets of Bogleheadism, isn't one of them that history repeats itself, that things don't change, that there is no new era of investing, etc. etc.?

3. I think investors have been aware of valuations since the time of Benjamin Graham, so investor behavior should have been affected for a long time. Again, this is an area where I need to do more research, but if markets are being driven by large fund managers who must focus on short-term goals like beating the benchmark, and that the only way to arbitrage this "market inefficiency" is to patiently wait for at least 10 years, then I don't necessarily think it will be priced out of the market by investor knowledge. It's not like the January effect or some other strategy that disappears as soon as it is discovered.

4. Target date funds: Now it is accepted that age (investing horizon) is an important factor for determining asset allocation, as well as risk aversion and other factors. What is so controversial about adding market valuation levels to this list?
richard wrote:You appear to believe market efficiency rules out short term timing but not long term timing. If so, what's the basis for this belief?


I just read the Amazon Review for "Valuing Wall Street" and I believe the beginning explains it better than I ever could:

"Most books about the stock market tell you how to make money. This one ... will show you how to avoid losing it," begins this smart, blunt, cautionary tale based on Nobel laureate James Tobin's 1969 "q ratio," which posits, among other things, that no matter how bullish a market gets, it's bound to snap back into place at some point--and those who don't brace for the reversal will feel its sting."

That's the idea. The problem about this snapping back is that you don't know if it will happen in 1 day, 1 week, 1 month, 1 year, or 5 years, but sooner or later it is bound to happen. That is why I think long-term market timing has more potential as a strategy.

User avatar
wade
Posts: 539
Joined: Fri Sep 17, 2010 12:38 am
Location: Main Line
Contact:

Re: Market timing ?

Post by wade » Thu Jan 20, 2011 8:02 am

Taylor Larimore wrote:Hi Wade:

My question is about LONG-TERM market timing strategies. In other words, using market timing over periods of at least 10 years to obtain better returns than a buy and hold strategy.


Long-term, intermediate-term, short-term, I doubt if it makes any difference. This is what Mr. Bogle wrote:

"After nearly fifty years in this business, I do not know of anybody who had done it (market timing) successfully and consistently. I don't even know anybody who knows anybody who had done it successfully and consistently."


As a former market-timer myself, I urge you to forget about it.


Taylor, thank you kindly for the comments, and I respect your opinion. Just to be sure, I'm no where near the stage of actually changing my asset allocation based on valuation levels.

About Mr. Bogle's quotation, I have the feeling that he is referring to short-term market timing, right?

You say that you doubt if it makes a difference, and I do respect your opinion, but I would like to study it some more. Spending a month or two researching this topic will be fun, even if it doesn't end up amounting to anything.

richard
Posts: 7961
Joined: Tue Feb 20, 2007 3:38 pm
Contact:

Post by richard » Thu Jan 20, 2011 8:50 am

wade wrote:I'm not comfortable with theoretical statistics, and my intuition may be wrong, but my intuition is that 120 rolling 10-year periods still provides more information than does 13 independent observations. Is your idea based on some theoretical findings?

It's standard statistics. Not to be abrupt, but your intuition is contrary to standard statistics.

For the reasons given in my post, I don't believe we have that many independent 10 year periods

wade wrote:1. I reject market-timing as being either 100% stocks or 100% Treasury bills. One example of what I am thinking of has been attributed to Benjamin Graham, though I haven't yet verified whether this is an urban legend. He seems to have advocated a 50% stock allocation that could be adjusted between 25% and 75% depending on market valuations. That type of strategy would provide some protection in case conditions did change.

Graham advocated staying between 25/75 and 75/25. I've thought of Graham as more a stock picker than a market timer. "If I have noticed anything over these 60 years on Wall Street, it is that people do not succeed in forecasting what’s going to happen to the stock market."

wade wrote:2. Though I haven't seen a list of the tenets of Bogleheadism, isn't one of them that history repeats itself, that things don't change, that there is no new era of investing, etc. etc.?

It's hard to see what is fundamental, beyond costs matter. I've been watching this stuff for a while and haven't seen a lot of "history repeats itself," certainly not in a way that suggests timing is actionable. It's more of a hostility to "this time it's different."

wade wrote:3. I think investors have been aware of valuations since the time of Benjamin Graham, so investor behavior should have been affected for a long time. Again, this is an area where I need to do more research, but if markets are being driven by large fund managers who must focus on short-term goals like beating the benchmark, and that the only way to arbitrage this "market inefficiency" is to patiently wait for at least 10 years, then I don't necessarily think it will be priced out of the market by investor knowledge. It's not like the January effect or some other strategy that disappears as soon as it is discovered.

As Taylor points out, there really aren't any long-term successful market timers. You should listen to Taylor.

wade wrote:4. Target date funds: Now it is accepted that age (investing horizon) is an important factor for determining asset allocation, as well as risk aversion and other factors. What is so controversial about adding market valuation levels to this list?

Your question is a non-sequitor. It's in the form "X is a good idea, so why should it be controversial to add Y?"

Age is a rule of thumb. Some reasons it is useful is the younger have more time to make up for mistakes and that people who are older tend to have more wealth, which means a lower need to take risks. Market timing is controversial because it doesn't appear to work.

richard wrote:You appear to believe market efficiency rules out short term timing but not long term timing. If so, what's the basis for this belief?

You didn't answer this question, you just provided some rationales for market timing.

dbonnett
Posts: 726
Joined: Tue Jun 10, 2008 3:59 pm

Post by dbonnett » Thu Jan 20, 2011 9:32 am

Even Dr. Bill Bernstein tells us to "keep your eyes on the price of tomatoes"

User avatar
wade
Posts: 539
Joined: Fri Sep 17, 2010 12:38 am
Location: Main Line
Contact:

Post by wade » Thu Jan 20, 2011 9:47 am

richard wrote:It's standard statistics. Not to be abrupt, but your intuition is contrary to standard statistics.


I haven't taken a statistics class since my first year of college. But can you possibly refer me to the chapter number of a statistics textbook that discusses this? I would love to read about it.

richard wrote:Graham advocated staying between 25/75 and 75/25. I've thought of Graham as more a stock picker than a market timer. "If I have noticed anything over these 60 years on Wall Street, it is that people do not succeed in forecasting what’s going to happen to the stock market."


Okay, but anyway, something like 25-50-75 depending on valuation levels is more like what I have in mind than 100-0. That was my point.


richard wrote:It's more of a hostility to "this time it's different."

Let's come back to this after I finished my research.

wade wrote:Your question is a non-sequitor. It's in the form "X is a good idea, so why should it be controversial to add Y?"


I don't wish to make a non-sequitor. Let me say differently. In the past, new research undercovered the idea that target date funds can improve the outcomes for long-term investors compared to straight buy-and-hold strategies. That doesn't mean valuations should also be added to the list. But why can't valuations be considered as another avenue for research that could possibly lead to new advice in the future? Why am I discouraged from even exploring the idea?

richard wrote:Market timing is controversial because it doesn't appear to work.


Okay. This is what I am asking in this thread. What studies show this? In particular, are there some academic studies that I haven't found yet? That's all I want to know. At this point, the two papers by Fisher and Statman do support your view. I already have some concerns about their methodology though, and I am not convinced by their findings. I am looking for other studies that support your view.

richard wrote:You appear to believe market efficiency rules out short term timing but not long term timing. If so, what's the basis for this belief?
You didn't answer this question, you just provided some rationales for market timing.


I didn't answer your question? I don't want to be evasive. Maybe I don't understand your question. If I can try again: should I conclude that long-term market timing is possible, I think it will be because market valuation levels tend to revert to their mean over long periods of time. But because the precise timing of this mean reversion is not known in advance, and is indeed random, expecting the result to happen in the short-term will not be possible. But long-term investors who can be patient can wait for this mean reversion, while they may lag behind buy-and-holders for years at a time, will eventually come out ahead by the end of the game.

User avatar
wade
Posts: 539
Joined: Fri Sep 17, 2010 12:38 am
Location: Main Line
Contact:

Post by wade » Thu Jan 20, 2011 10:10 am

wade wrote:
richard wrote:It's standard statistics. Not to be abrupt, but your intuition is contrary to standard statistics.


I haven't taken a statistics class since my first year of college. But can you possibly refer me to the chapter number of a statistics textbook that discusses this? I would love to read about it.


Richard, before I gave the example: 130 years of data means 13 independent 10-year observations, or 120 rolling 10-year observations. I said my intuition is that 120 rolling 10-year observations provides more information than 13 independent observations. You tell me I am wrong, and it is standard statistics. Just in case you were next going to tell me to do my own homework and find the answer for myself, here is an excerpt from a paper by Ulrich Mueller, who is now a professor at Princeton University:

In many statistical studies of time series, a special type of variables is observed: variables that belong to a time interval in the time series. A typical example is price change statistics, such as the analysis of the scaling law, see (M¨uller et al., 1990). The mean value of a variable, e. g. the absolute value of price changes over an interval of fixed length, has to be determined over a certain sample. The more observations in the sample, the smaller the random error of the mean and the more significant the result.
Oneway to increase the number of data points is to also consider overlapping time intervals, so the distance between the starting points of two subsequently analyzed intervals becomes smaller than the interval size. The obtained series of observations certainly exhibits serial
dependence, we cannot regard the observations as independent. However, we intuitively feel that adding overlapping intervals to the sampling scheme might increase the precision of the result. Is this intuitive feeling justified by theory?
The exact answer depends on the statistical properties of the analyzed variable.
...


source:

http://www.olsen.ch/fileadmin/Publicati ... verlap.pdf

It seems like my intuition does not deserve such an abrupt response :D

User avatar
Taylor Larimore
Advisory Board
Posts: 26125
Joined: Tue Feb 27, 2007 8:09 pm
Location: Miami FL

What does Mr. Bogle mean ?

Post by Taylor Larimore » Thu Jan 20, 2011 11:04 am

Hi Wade:
Mr. Bogle wrote:

"After nearly fifty years in this business, I do not know of anybody who had done it (market timing) successfully and consistently. I don't even know anybody who knows anybody who had done it successfully and consistently."


You wrote:

Taylor--About Mr. Bogle's quotation, I have the feeling that he is referring to short-term market timing, right?

If you read his quote carefully, he didn't say that.

Respectful suggestion: If you are going to do your own research, it is important to avoid data and interpreting quotations to fit preconceived opinions.
"Simplicity is the master key to financial success." -- Jack Bogle

User avatar
bob90245
Posts: 6511
Joined: Mon Feb 19, 2007 8:51 pm

Re: Any Studies on Long-term Market Timing?

Post by bob90245 » Thu Jan 20, 2011 11:10 am

wade wrote:But my question is about LONG-TERM market timing strategies. In other words, using market timing over periods of at least 10 years to obtain better returns than a buy and hold strategy. The literature seems slim. I will list what I know about, and I hope someone might fill in more details. I am particular interested, at this point, in studies which reject long-term market timing.

I wrote an article that examines P/E10 and 10-year returns of stock and bonds. You might say that it mildly supports long-term market timing. Or rather, to consider reducing stock allocation when P/E10 is very high. Here's one chart that plots the data:

Image
Source: http://www.bobsfinancialwebsite.com/Los ... sPE10.html
Ignore the market noise. Keep to your rebalancing schedule whether that is semi-annual, annual or trigger bands.

User avatar
Adrian Nenu
Posts: 5228
Joined: Thu Apr 12, 2007 6:27 pm

Post by Adrian Nenu » Thu Jan 20, 2011 11:19 am

One example of what I am thinking of has been attributed to Benjamin Graham, though I haven't yet verified whether this is an urban legend. He seems to have advocated a 50% stock allocation that could be adjusted between 25% and 75% depending on market valuations. That type of strategy would provide some protection in case conditions did change.



Ben Graham could do it, Buffett can do it but +99% of investors can't implement this strategy. In my opinion, conservative "buy & hold" portfolios, risk of loss tested for suitability offer average investors the best odds of success towards achieving their financial goals. Not perfect but it is the best option in the real world.

Adrian
anenu@tampabay.rr.com

User avatar
BlueEars
Posts: 3531
Joined: Sat Mar 10, 2007 12:15 am
Location: West Coast

Re: Any Studies on Long-term Market Timing?

Post by BlueEars » Thu Jan 20, 2011 1:39 pm

wade wrote:May I ask about long-term market timing strategies?
...

Siegel in Stocks for the Long Run presented the details of market timing using moving averages over several decades. The presentation is pretty even handed which is unusual.

There is a lot of talk about secondary sources here but very few have actually tried to wrestle with the data. Probably because that takes a lot of time and effort to do it right (and I don't mean just reading charts off the web) and it'd be pretty meaningless if you felt MT was a losers game. Most of the MT stuff presented seems to be about moving averages. Here's a link with 23 pages of back-and-forth http://www.bogleheads.org/forum/viewtopic.php?t=27460&highlight=200dma

In this forum you will find lots of confirmation bias regarding the impossibility of timing markets in any way. Personally, I'd like to keep an open mind. But keeping MT discussions confined to just a few threads in this forum is a good idea IMO.

Yipee-Ki-O
Posts: 167
Joined: Sun May 24, 2009 12:42 pm

Post by Yipee-Ki-O » Thu Jan 20, 2011 1:39 pm

Norbert has anotherstudy, "Assett Allocation Strategies: Age + Valuation + Momentum," posted at Morningstar you might find of interest.

Wagnerjb
Posts: 7015
Joined: Mon Feb 19, 2007 8:44 pm
Location: Houston, Texas

Post by Wagnerjb » Thu Jan 20, 2011 2:18 pm

richard wrote:The problem with studies of long-term strategies is that there isn't enough data for any conclusion to be statistically significant. There at most 10 reliable independent data points for 10 year returns.



That is the problem I have with this topic as well. First of all, you need to be quoting from books published in say 1951 that show how to time the markets to your advantage. Of the hundreds of "get rich quick" books published in 1951 you have to be confident that people knew this one particular strategy was the right one. And you will have 60 years of results to see. The strategy would be massively popular by now and would be embraced by many intelligent and respected finance experts.

I don't believe looking at books published in the recent past can tell you anything. Let's see how actual results fared AFTER the book was published. How did real people do?

Best wishes.
Andy

User avatar
Random Musings
Posts: 5209
Joined: Thu Feb 22, 2007 4:24 pm
Location: Pennsylvania

Post by Random Musings » Thu Jan 20, 2011 2:35 pm

Current PE10 around 23.50. Sure, there were run-ups in the late 20's and a very nice one in the late 90's, but unless you were a timer and sold high, longer period returns (3-7 year range) were disappointing on a real basis.

For buy and holders, the best intermediate outcome might be spinning your wheels with high volatility.

IMHO, for both equities and bonds, nothing screams "value" to me right now.

RM

matt
Posts: 2305
Joined: Sun Mar 04, 2007 3:47 pm

Post by matt » Thu Jan 20, 2011 3:04 pm

Yipee-Ki-O wrote:Norbert has anotherstudy, "Assett Allocation Strategies: Age + Valuation + Momentum," posted at Morningstar you might find of interest.


I don't think that's the same Norbert.

User avatar
Scott S
Posts: 1388
Joined: Mon Nov 24, 2008 3:28 am
Location: SW CR IA US NA PE

Post by Scott S » Thu Jan 20, 2011 3:28 pm

I suppose if we're stretching the term "market-timing" to include 10+ years, I guess I'm "market timing" as well. I'm making a bet that stocks will be much higher in 30 years, when I plan to retire. :wink:

- Scott
My Plan: * Age-10 in bonds until I reach age 60, 50/50 thereafter. * Equity split: 50/50 US/Int'l, Bond split: 50/50 TBM/TIPS. * Everything over 2 months' expenses gets invested.

dbonnett
Posts: 726
Joined: Tue Jun 10, 2008 3:59 pm

Post by dbonnett » Thu Jan 20, 2011 4:59 pm

Several studies of long term market timing have shown marked differences between decades using the same approach. It's only real advantage is when the market goes into a deep hole (-35%) Otherwise it sorely trails buy and hold. Especially when you factor in taxes in a taxable account.

peter71
Posts: 3768
Joined: Tue Jul 24, 2007 8:28 pm

Post by peter71 » Thu Jan 20, 2011 6:03 pm

While the too few INDEPENDENT periods critique is a reasonable one, it doesn't stop econ profs like Shiller and Delong from finding statistical significance using overlapping periods . . . and as someone who teaches stats I think people overestimate the importance of statistical significance in both directions . . . a very common problem (though not on this board) is thinking that a tiny effect size observed in a very large sample is "important" because it crosses the threshold of statistical significance . . . anyway, here's one of many capably-done papers:

http://www.kansascityfed.org/PUBLICAT/R ... p02-01.pdf

All best,
Pete

dbonnett
Posts: 726
Joined: Tue Jun 10, 2008 3:59 pm

Post by dbonnett » Thu Jan 20, 2011 7:28 pm

Could someone paraphrase and summarize the findings of the above papers. The main point seems that when PEs and/or interest rates are high, stocks tank.

bhmlurker
Posts: 305
Joined: Mon Apr 14, 2008 11:16 pm

Post by bhmlurker » Thu Jan 20, 2011 7:52 pm

dbonnett wrote:Could someone paraphrase and summarize the findings of the above papers. The main point seems that when PEs and/or interest rates are high, stocks tank.


Isn't 'interest rate' as set by the Fed merely reacting to the state of the economy? If it's a lagging indicator, I don't see how it can predict future market performance, i.e. a fire engine speeding down the road is because a fire occurred, not because one will occur.

Using PE in any form to predict future performance for a broad index is problematic, because while one can get up-to-the-millisecond Price data, past EPS are, well, past, and getting accurate forward EPS is difficult to say the least. Additionally, EPSs are riddled with extraordinary events, depreciation, and all sort of finance trickery that makes it less than easy to discern how a business is really doing.

In my own experience, if one spends a lot of time to understand a handful of companies, such that one can consistently predict forward Cash Flow, one might have a good chance at plotting real-time Price vs Cash Flow and actually buy near the trough. This, I believe, is only possible for companies with minimal extraordinary events each quarter, and preferably with forward revenue semi-predictable by long-term contracts. In essence, this kind of market timing appears similar to timing TIPS real yield, and is really not plausible for most stocks.
Last edited by bhmlurker on Thu Jan 20, 2011 8:00 pm, edited 2 times in total.

User avatar
alec
Posts: 2845
Joined: Fri Mar 02, 2007 2:15 pm

Post by alec » Thu Jan 20, 2011 7:54 pm

"It is difficult to get a man to understand something, when his salary depends upon his not understanding it!" - Upton Sinclair

User avatar
dmcmahon
Posts: 1856
Joined: Fri Mar 21, 2008 10:29 pm

Post by dmcmahon » Thu Jan 20, 2011 8:06 pm

wade wrote:The problem about this snapping back is that you don't know if it will happen in 1 day, 1 week, 1 month, 1 year, or 5 years, but sooner or later it is bound to happen. That is why I think long-term market timing has more potential as a strategy.


Well, the problem with that is that it might not "snap back" but could "soft-land". Suppose the market is, today, about 20% overvalued. Now suppose that the overvaluation is corrected over the next decade by underwhelming market performance, perhaps (say) 1-2% less growth per year in equity values. At the end of the period, in 2020, the market is now fairly valued again. During that time it delivered less-than-stellar returns, but the returns still beat the pants off of cash or bonds. You die broke waiting for a crash that never comes. So, your timing strategy will somehow have to take into account the degree of overvaluation. Arguably you're not going to go too far wrong shifting your AA to 90% bonds if the market's trading at some astronomical level like (say) 100 P/E. But what about 25 P/E? Here's another problem: what if the other part of the AA (bonds) is also overvalued? Sitting in cash may then underperform any AA that splits between stocks and bonds.

User avatar
wade
Posts: 539
Joined: Fri Sep 17, 2010 12:38 am
Location: Main Line
Contact:

Post by wade » Thu Jan 20, 2011 11:20 pm

VERY PRELIMINARY AND SUBJECT TO REVISION

"Valuation-Informed Indexing"

by Wade D. Pfau

These are some preliminary findings from the research I am working on now. I will present one example. Obviously my result would be data mining if I choose this particular example because it is the only one that worked. I will take steps in my final paper to test a wide variety of assumptions about asset allocation, valuation-based decision rules, whether the period is 10, 20, 30, or 40 years, lump-sum vs. dollar-cost averaging, and so on, and to show that the results are quite robust to changes in any of these assumptions. The basic story can be seen with the following example though.

Data for this study is from Robert Shiller's webpage. The two asset choices are stocks (S&P 500 index) and Treasury bills (1-year yields).
http://www.econ.yale.edu/~shiller/data.htm

My example will be for a person with a 30-year long career. This person will invest 10% of salary each year over 30 years. The salary is constant in real terms, so it keeps pace with inflation, but there is no real wage growth for this investor. These particular assumptions, though, make very little difference.

I don't like the term "long-term market timing." Market-timing is much too pejorative, and is also too easy to confuse with short-term market timing. Also, market-timing seems to be associated with extreme strategies: either 100% stocks or 100% Treasury bills. I don't like that either. I think the best term for this is "Valuation-Informed Indexing," (VII) which is a term coined by Rob Bennett, who is banned from posting at the Bogleheads Forum. In discussing asset allocation strategies for valuation-informed indexing (VII), Rob Bennett in his podcast:
RobCast #137 "Nine VII Portfolio Allocation Strategies"indicates some preference for his high-medium-low strategy, which would be 60% stocks in the baseline, but would switch to 30% stocks when the PE10 ratio moves above 21, and would switch to 90% stocks when the PE10 ratio moves below 12. My example will be a variation of this strategy that captures the spirit of what he suggested. I don't think it is proper to use the ex-post rule of deciding on valuation cut-offs after the fact (but his specific decision rule does work just as well), but I will consider the evolving interquartile range of PE10 ratios between 1871 and each date in the historical period. When PE10 enters the top 25% percentile of for the historical data up to that point, the investor switches to 30% stocks, and when the PE10 falls into the bottom 25% percentile, the investor switches to 90% stocks.

Shiller's data for PE10 as well as the evolving median value and evolving interquartile range [which shows how the decision rules are formed] is seen in the following figure:

Image

Just to be clear about this again, when PE10 is above the top red line, the stock allocation is 30%, and when PE10 is below the bottom red line, the stock allocation is 90%. For the period in between, the stock allocation is 60%.

This VII strategy will be compared to the classic buy-and-hold strategy: 60% stocks and 40% bills. I guess the classic strategy is really 60/40 for stocks and bonds. I haven't tested that yet, but later I should confirm the results when having bonds replace bills for each strategy.

I assume that each year rebalancing will be used to restore the asset allocation to the targeted level for that year.

Now to present the results for 110 rolling 30-year periods with underlying data from 1871 to 2009, for careers beginning between 1871 and 1980. What you see in the top part of the graph for each year is the amount of wealth accumulated after 30 years for someone following buy-and-hold against someone following VII. This is scalar for the savings rate. I use 10%, but a higher savings rate would push the numbers up in the same consistent way. The particular savings rate makes no difference to the relative difference between the two strategies. When the red line is above the blue line, VII provided more wealth, and when the red line is below the blue line, buy-and-hold did better. The bottom part of the figure, which is based on the PE10 decision rule described above, shows the asset allocation for each strategy in each year.

Image

Because of the sound thrashing delivered by VII for investors living through the Great Depression, this figure makes it hard to see that buy-and-hold did occasionally do better.

VII provides more wealth for 102 of the 110 rolling 30-year periods, while buy-and-hold did better in 8 of the periods. If that isn't enough to satisfy you, let's look in more detail at the 8 cases where buy-and-hold provided more wealth. The following figure is the same as the top of the previous figure, except that the points for VII are only plotted for the 8 cases when VII underperformed buy-and hold:

Image

That figure is still going to be hard to see, so here is a Table of those 8 cases.
The columns are year, buy-and-hold wealth accumulation, VII wealth accumulation

1970.00 5.61 5.19
1971.00 5.25 5.03

1974.00 4.89 4.49
1975.00 6.02 5.86
1976.00 5.27 4.68
1977.00 5.48 4.71
1978.00 5.89 5.39

1980.00 5.37 5.01

Now some observations about this:
First, the underperformance of VII is minimal and is not going to put the investor in the poor house. Remember that these calculations are in real terms, in which wealth is expressed as a multiple of final salary. In these final years, buy-and-hold was performing as well as it ever had at any point since 1880, and the slightly lower VII numbers also provided wealth that was on the high end of the spectrum compared to what buy-and-hold had provided in the past. This seems like a small price for VII to pay for insurance against market crashes when valuations are out-of-control.

Second, you may be concerned by the fact VII only started to fail just recently means that it is no longer going to work any more in the future. I don't think that is the appropriate conclusion. That conclusion would mean that stocks can now enjoy a permanently higher-plateau for their PE10 ratio. If that's true though, then the VII strategy will evolve and that interquartile range will adjust upwards. There will still be periods of high valuations and low valuations. VII did cause investors to miss out somewhat on the full bull market run of the 1990s, as their stock allocation was lowered between 1992 and 2008. But, as shown the price they paid for this was minimal. I don't really think we are in a new era.

Okay, that's the basic idea. I know that considering taxes is important. These results would apply for a Roth IRA. Accounting for taxes properly is difficult, but I hope that by the end of the year I am able to do that. I welcome your comments and criticisms about this. Thank you.

Edit: I've added this post to my blog: http://wpfau.blogspot.com/2011/01/valua ... inary.html
Last edited by wade on Fri Jan 21, 2011 12:33 am, edited 1 time in total.

User avatar
bob90245
Posts: 6511
Joined: Mon Feb 19, 2007 8:51 pm

Post by bob90245 » Thu Jan 20, 2011 11:49 pm

wade wrote:I will take steps in my final paper to test a wide variety of assumptions about asset allocation, valuation-based decision rules, whether the period is 10, 20, 30, or 40 years, lump-sum vs. dollar-cost averaging, and so on, and to show that the results are quite robust to changes in any of these assumptions. ...

My example will be for a person with a 30-year long career. This person will invest 10% of salary each year over 30 years.

I am quite surprised 30 years produced such good results. I examined 10-year periods and the number of "Lost Decades" were relatively small. I'll repost my chart from upthread:

Image
Source: http://www.bobsfinancialwebsite.com/Los ... sPE10.html
Ignore the market noise. Keep to your rebalancing schedule whether that is semi-annual, annual or trigger bands.

User avatar
wade
Posts: 539
Joined: Fri Sep 17, 2010 12:38 am
Location: Main Line
Contact:

Post by wade » Fri Jan 21, 2011 12:03 am

Bob90245:

After 10 years, VII is just starting to work. I will show the results now for 10-year rolling periods. VII outperforms buy-and-hold in 91 out of 130 (70%) of 10-year periods:

Image

Image

User avatar
market timer
Posts: 5801
Joined: Tue Aug 21, 2007 1:42 am

Post by market timer » Fri Jan 21, 2011 12:38 am

Wade, you may want to compare VII against the leveraged lifecycle models suggested by Ayres and Nalebuff in Lifecycle Investing. Their suggestion to maintain constant risk in dollar terms has the investor selling as stocks rise and buying as stocks fall, provided the collateral constraint is not binding.

cjking
Posts: 1742
Joined: Mon Jun 30, 2008 4:30 am

Post by cjking » Fri Jan 21, 2011 4:36 am

I would be interested to see the results of a slight variation on this strategy, one where you only switch back to 60% (from either direction) when the median of PE10 is crossed. I recognise this strategy does carry risks for the future that the original doesn't, so it would be legitimate to rule it out even if improves on the original historically.

Edit: To be sure I'm being clear, I mean you would switch away from 60% at the same quartile thresholds as before, but switch back at a different threshold, the median.

peter71
Posts: 3768
Joined: Tue Jul 24, 2007 8:28 pm

Post by peter71 » Fri Jan 21, 2011 6:37 am

Hi Wade,

A few comments:

1) At first I thought the inter-quartile thing was atheoretical but I do like that it anticipates the "permanently high plateau" critique.

2) I do think you have to think about the theory behind P/E 10, such as it is, as opposed to valuations more generally . . . the effect sizes that Delong finds in updating Shiller are huge -- a one unit change in P/E 10 leads to an approximately 50 basis point decrease in annualized returns over 20 years, but IIRC Shiller is almost explicit about trying out about eight different valuations predictor models and converging on P/E 10 as the one that worked best at the time he first started doing this work in the 90's (I trust his co-authored stuff with Campbell even less, by the way). So while I think the theory behind valuations mattering is incredibly sound, I worry a lot about situations like the present one in which the IMO theoretically sound "P/E 1" is far lower than P/E 10. And even if you don't think P/E 1 is a theoretically sensible predictor, why P/E 10 in particular out of all the possible temporal and non-temporal adjusted P/E variants out there?

3) You presumably have a ton of training in multivariate modeling, so while zeroing in on a single predictor may have pedagogical value it seems worth thinking about at least one more . . . I'm personally persuaded that both interest rates and momentum are worth taking into consideration too.

All best,
Pete

User avatar
wade
Posts: 539
Joined: Fri Sep 17, 2010 12:38 am
Location: Main Line
Contact:

Post by wade » Fri Jan 21, 2011 8:06 am

cjking wrote:I would be interested to see the results of a slight variation on this strategy, one where you only switch back to 60% (from either direction) when the median of PE10 is crossed. I recognise this strategy does carry risks for the future that the original doesn't, so it would be legitimate to rule it out even if improves on the original historically.

Edit: To be sure I'm being clear, I mean you would switch away from 60% at the same quartile thresholds as before, but switch back at a different threshold, the median.


I ordered "Valuing Wall Street" today, and I also realized that I already have "Wall Street Revalued" but just hadn't looked at it yet. I thought it was a book about the financial crisis. I brought it home to start reading this weekend.

I think I understand your proposed strategy. It would result is spending more time at the extreme allocations, but still would occasionally spend some time at the middle allocation. I will try it out and let you know later about what happens. Next week is going to be quite busy for me at work, so it might take about a week before I get to it.

Another variation on this theme might be that once you switch an allocation, you are not allowed to switch again for x amount of time. This would reduce the occasions where allocations start moving around a lot within a short amount of time.

peter71 wrote:1) At first I thought the inter-quartile thing was atheoretical but I do like that it anticipates the "permanently high plateau" critique.

2) I do think you have to think about the theory behind P/E 10, such as it is, as opposed to valuations more generally . . . the effect sizes that Delong finds in updating Shiller are huge -- a one unit change in P/E 10 leads to an approximately 50 basis point decrease in annualized returns over 20 years, but IIRC Shiller is almost explicit about trying out about eight different valuations predictor models and converging on P/E 10 as the one that worked best at the time he first started doing this work in the 90's (I trust his co-authored stuff with Campbell even less, by the way). So while I think the theory behind valuations mattering is incredibly sound, I worry a lot about situations like the present one in which the IMO theoretically sound "P/E 1" is far lower than P/E 10. And even if you don't think P/E 1 is a theoretically sensible predictor, why P/E 10 in particular out of all the possible temporal and non-temporal adjusted P/E variants out there?

3) You presumably have a ton of training in multivariate modeling, so while zeroing in on a single predictor may have pedagogical value it seems worth thinking about at least one more . . . I'm personally persuaded that both interest rates and momentum are worth taking into consideration too.


Thank you for the comments.

About PE10, I do think that earnings are too noisy for PE1, and so PE10 is better than PE1. Beyond that, perhaps PE6 or PE13 would work even better, but I'm not sure that finding the PEx that maximizes the results is wise. PE10 is a commonly used measure, and that is why I choose it. Actually, I was not aware of Delong's work on this topic. I will look it up and read it carefully. Thanks.

About a multivariate model, when I first started thinking about how to approach this, I did have in mind a multivariate model that would be similar to what I used in the following paper to select optimal asset allocations in the final part of the paper:

"Predicting Sustainable Retirement Withdrawal Rates Using Valuation and Yield Measures."
http://ideas.repec.org/p/pra/mprapa/27487.html

But as I thought about it more, I decided that at least for a first look into the issue, I can go ahead and keep things simple. Then, if this simpler version can generate some interest and get published, I would go ahead and put more effort into developing a more complete and technical paper about this issue. Also, as you nicely put, one variable does have pedagogical value and is easy to understand.

Also, earlier, you made a comment about the difference between statistical significance and real world significance. It is important. Shiller finds that PE10 as an R2 in the neighborhood of 0.3 for predicting 10-year stock returns, and while that it significant, it would be hard to make money with. But I think he acknowledges those limitations quite well. The interesting thing is how applying PE10 to investors saving for retirement (as discussed in this thread) or withdrawing during retirement (as discussed in the paper linked above, where the R2 is 0.76, but that includes other variables too) results in much better fits than for those 10-year stock predictions.

afan
Posts: 2943
Joined: Sun Jul 25, 2010 4:01 pm

Post by afan » Fri Jan 21, 2011 8:35 am

Wade,

Thanks for bringing this up, and letting us know about your blog.


I am curious how you are going to handle the concern of data mining. Searching through for a variety of time periods and assumptions does not seem to address the issue. If you have one time series of data, cannot assume the points are independent (in fact, the premise to test is that they are not independent, then how can you test for significance

Maybe try international data

Thanks

dbonnett
Posts: 726
Joined: Tue Jun 10, 2008 3:59 pm

Post by dbonnett » Fri Jan 21, 2011 9:33 am

Would seem to defy fine tuning and only provide value at extreme levels of exuberance and despair...

User avatar
DRiP Guy
Posts: 2237
Joined: Tue Feb 20, 2007 4:54 pm

Post by DRiP Guy » Fri Jan 21, 2011 9:38 am

1) Wade, I offered criticism of your article via your partner's site (your own site did not appear to be accepting comments from my browser).

2) I can easily understand why you would love to use the Boglehead's many resources (amazingly giving contributors, genius-level participants, respected researchers and authors, a ready large audience in place, huge gravitas in the myriad field of internet investing forums, etc), especially as a contrasting canvas against which you can bounce your apparently developing theories, which you wish to work into academic papers under your own authorship. I don't begrudge you that, and wish you much success.

However, since your own work is overtly at odds with the ethos of the board -- here, the theme is John Bogle's philosophy, which eschews market timing, I myself will no longer obliquely support it by giving you a whetstone on which to sharpen your knife. You must certainly know that this very board came into existence in order to ESCAPE the lunatic behaviors of one individual -- the very individual with which you have publicly and openly aligned yourself, and who you are openly quoting and sourcing in your column and are forming your intended paper around.

While there is much merit in open discussion of competing, differing, and varied approaches, as to you, sir, I personally will have no more of it here on this forum, given the poison well from which you are now openly drawing your own water.

I ask no one to join me in my position, but merely to follow the dictates of your own conscience and intellect and beware reinforcing a dangerous mania, even if by proxy.

cjking
Posts: 1742
Joined: Mon Jun 30, 2008 4:30 am

Post by cjking » Fri Jan 21, 2011 9:46 am

wade wrote:I ordered "Valuing Wall Street" today, and I also realized that I already have "Wall Street Revalued" but just hadn't looked at it yet. I thought it was a book about the financial crisis. I brought it home to start reading this weekend.


"Valuing Wall Street" is the book I really meant to recommend.

"Wall Street Revalued" is a response to the financial crisis, perhaps the most relevant bit is its explanation of markets as imperfectly efficient, which might answer some questions asked here about why long-term timing should be possible.

"Valuing Wall Street" is totally focused on investigating/explaining one big idea from every possible angle, "Wall Street Revalued" is more of a discursive response to recent events.

afan
Posts: 2943
Joined: Sun Jul 25, 2010 4:01 pm

Post by afan » Fri Jan 21, 2011 10:31 am

Drip,

As a relatively new person on this forum, I have no idea what you are talking about. There is someone, not Wade, whose "lunatic" behavior lead to the existence of this board? I can understand avoiding the classic abusive internet behavior of toxic contributors. However, that is a far cry from having no more of open discussion of competing approaches. From what little I have seen on this forum and Wade's site, I don't see anything harmful.

Again, I am new here, but I hope people can post ideas that do not conform to others ideas of what Bogle would say. After all, this is finance, not religion. Bogle is a smart guy who has done a tremendous service to American investors, including the majority who do not do business with Vanguard. Does that mean no one is allowed to disagree with him on any topic?

afan
Posts: 2943
Joined: Sun Jul 25, 2010 4:01 pm

Post by afan » Fri Jan 21, 2011 10:37 am

On the original topic, Vanguard has a short paper reviewing some of the evidence on tactical asset allocation. Includes references to the original literature. Not sure it is over a long enough term for your interests.

Asset Allocation: A Primer on Tactical Asset Allocation Strategy Evaluation

https://personal.vanguard.com/us/litera ... mentguides

User avatar
DRiP Guy
Posts: 2237
Joined: Tue Feb 20, 2007 4:54 pm

Post by DRiP Guy » Fri Jan 21, 2011 10:56 am

afan wrote:Drip,

As a relatively new person on this forum, I have no idea what you are talking about. There is someone, not Wade, whose "lunatic" behavior lead to the existence of this board? I can understand avoiding the classic abusive internet behavior of toxic contributors. However, that is a far cry from having no more of open discussion of competing approaches. From what little I have seen on this forum and Wade's site, I don't see anything harmful.

Again, I am new here, but I hope people can post ideas that do not conform to others ideas of what Bogle would say. After all, this is finance, not religion. Bogle is a smart guy who has done a tremendous service to American investors, including the majority who do not do business with Vanguard. Does that mean no one is allowed to disagree with him on any topic?


I've made my personal point, and it needed to be said. As you say, if you have no idea of why this board came to be, it is reasonable you would not understand, and there is no need to go into it further.

If the information doesn't regard you, then pay no mind -- post in peace and ease about whatever you like, to your heart's content!

User avatar
bob90245
Posts: 6511
Joined: Mon Feb 19, 2007 8:51 pm

Post by bob90245 » Fri Jan 21, 2011 11:33 am

wade wrote:Bob90245:

After 10 years, VII is just starting to work. I will show the results now for 10-year rolling periods. VII outperforms buy-and-hold in 91 out of 130 (70%) of 10-year periods:

Image

After I had a night to mull this over, I think understand the reason why VII has outperformed. And your two charts seem to confirm by my eyeball estimations.

I notice that in certain periods when VII outperforms, the stock allocation is set to 90%. It would be interesting if this is the major reason for the outperformance. Thus buy-and-hold is held back by being at only 60% during these periods.

If I can extrapolate from this interpretation, VII switches to 90% only at low P/E10 levels. Which tend to forecast higher future returns. However, keep in mind the reason why P/E10 might have fallen to low levels in the first place: the economic environment was more uncertain than usual.

As an example, recall only the most recent period when P/E10 fell to around 13 in early 2009. The financial markets had been in turmoil and governments around the world were struggling with attempts to shore up the system. It was indeed a scary time for investors as reflected by low stock prices. At the time, hardly anyone could have predicted that stocks would recover strongly, let alone rocket over 80% from the lows.
Ignore the market noise. Keep to your rebalancing schedule whether that is semi-annual, annual or trigger bands.

User avatar
Orion
Posts: 506
Joined: Mon Feb 19, 2007 11:52 pm

Post by Orion » Fri Jan 21, 2011 1:14 pm

afan wrote:There is someone, not Wade, whose "lunatic" behavior lead to the existence of this board?


Yep - although it has obviously moved well beyond those early days. It was the third case I know of where a new board was created to be just like the old board only without that person. He said he made up the name of his investing system so that people will google it and end up at his site. As far as I have been able to tell, his system is basically his name applied to Shiller's work and he does not follow the system himself. I say "basically" because no one has gotten a precise and stable definition. Over the years various people have backtested it using their own various guesses as to the parameters and sometimes it tests well and sometimes it doesn't.

Last I read, a newspaper reporter interviewed him and he said he may have to return to work because his website hadn't taken off. My interest in him waned after that. However, I still find Shiller's work interesting. Let's not throw the baby out with the bathwater.

User avatar
bob90245
Posts: 6511
Joined: Mon Feb 19, 2007 8:51 pm

Post by bob90245 » Fri Jan 21, 2011 1:38 pm

wade wrote:I don't like the term "long-term market timing." Market-timing is much too pejorative, and is also too easy to confuse with short-term market timing. Also, market-timing seems to be associated with extreme strategies: either 100% stocks or 100% Treasury bills. I don't like that either. I think the best term for this is "Valuation-Informed Indexing," (VII) which is a term coined by Rob Bennett, who is banned from posting at the Bogleheads Forum. ...

I don't think it is proper to use the ex-post rule of deciding on valuation cut-offs after the fact (but his specific decision rule does work just as well), but I will consider the evolving interquartile range of PE10 ratios between 1871 and each date in the historical period. When PE10 enters the top 25% percentile of for the historical data up to that point, the investor switches to 30% stocks, and when the PE10 falls into the bottom 25% percentile, the investor switches to 90% stocks. ...

The bottom part of the figure, which is based on the PE10 decision rule described above, shows the asset allocation for each strategy in each year.

Image

My argument is not to your methdology but with semantics. You may like the term "Valuation Informed Indexing" better than "long-term market timing". But frankly, I don't see the connection with "long term". Yes, there are some long stretches when the stock allocation doesn't change. But you have other periods like the decade of the 1930's when the stock allocation changed 5 times. That would be short term by most standards.

So to be intellectually honest, I think you should not coin new phrases. Just stick to the common if not pedestrian term of "tactical asset allocation".
Ignore the market noise. Keep to your rebalancing schedule whether that is semi-annual, annual or trigger bands.

User avatar
Norbert Schlenker
Posts: 540
Joined: Mon Feb 19, 2007 11:06 pm
Location: The Dry Side of the Wet Coast

Post by Norbert Schlenker » Fri Jan 21, 2011 2:48 pm

I don't really have time for this but a PM alerted me that my name had been mentioned above. It's an interesting topic. The thread at FWF is here. It's hard to believe it was three years ago.

I do conclude that the evidence in favour of long term timing is tenuous at best. All studies of time series that try to draw conclusions are subject to cherry picking and end date bias, so you have to go into them with skepticism from the word go. With really short time series - and we only have good data for a very short period of time - you had better be careful. Really careful.

Furthermore, when one is dealing with human behavior and economics, you had better convince yourself ahead of time that today really isn't different than a hundred years ago. Quite frankly, I think it's pretty different. I think economics pre 1970 is different than economics today. I think financial markets pre about 1990 are different than financial markets today. So I think one walks on quicksand when one extrapolates from a trading technique's historical outperformance when most of the history lies far in the past and is thus potentially completely irrelevant. I honestly believe that we have only about 20 years of excellent data, maybe 50 years of good data.

There is a real shortage of good data in this field. If you study this at all, you want statistical significance. The only way to get that is more data. So you go to the only place it can come from, which is history. How good is that data? How accurate is that data? How relevant is that data to the way in which the economy and markets operate today? All of this gets glossed over by appealing to Shiller's spreadsheet. That spreadsheet is an abstraction of what was going on back to about 1880. The first question to ask is what relevance imperfect and incomplete market data from 130 years ago has today, if any. The second is what else we are ignoring. Taxes? Frictional costs?

Instead, we ask Excel to draw a chart. I'm not pointing fingers; I do it too. Every picture does tell a story. The trouble is that, sometimes, the story makes you forget all the conditions and caveats to consider. Because stories are fun and easy to remember and tell. And some of them are even true, provable beyond a reasonable doubt.

This isn't one of them. I'd like a magic bullet too. Right now, there isn't enough data to support this one. I don't believe there will be in my lifetime.

Regarding the question of overlapping versus non-overlapping data, there's a real problem there. This board has a huge population. Somebody must know something about signal theory. AFAIK - but I am not an expert - there is no extra information in 120 overlapping sub-intervals compared to 12 non-overlapping. It's at best neutral, I believe, and it may actually lose information (because the act of calculating a compounded ten year return is, like almost every statistical operation, intentionally lossy). If you want 120 independent data points, you have an option: Use the annual returns. Good luck finding the signal in that.

All animals are pattern matchers. We are animals. We are desperate to see the pattern, not only in finance, but in everything in life. The pattern is a shortcut, a mental map of how the world works, and it's really easy to expend considerable effort to generalize from data to the pattern, i.e. extract signal from noise. It is easy to be fooled. It is easy to construct the map from insufficient data. We are hard wired to do this. And it is hard - REALLY hard - to step back, re-examine assumptions, change the map, even to wonder if the map is wrong.
Nothing can protect people who want to buy the Brooklyn Bridge.

peter71
Posts: 3768
Joined: Tue Jul 24, 2007 8:28 pm

Post by peter71 » Fri Jan 21, 2011 4:03 pm

Hi Norbert,

Few quick notes . . .most of what you say is uncontroversial but will be well-traveled ground for Wade, who's an economics prof.

On signalling, we've talked about it here before, but I think it's ultimately a philosophy of science issue rather than a technical issue . . . but even if we concede the point that shiller and shiller's peer reviewers are all utterly and profoundly incompetent for using overlapping periods, you can, IIRC, look at everything from 6 non-overlapping 20 year periods to 40 non-overlapping 3-year periods and get regression coefficients with the same sign . . . and that puts the fact that, without the overlapping periods, one can't be 95% confident that a two-tailed test of statistical significance allows one to reject the null that valuations don't matter in context . . .

But I certainly agree re "no magic bullet," etc., etc.

All best,
Pete

peter71
Posts: 3768
Joined: Tue Jul 24, 2007 8:28 pm

Post by peter71 » Fri Jan 21, 2011 4:03 pm

dup post deleted
Last edited by peter71 on Fri Jan 21, 2011 4:12 pm, edited 1 time in total.

matt
Posts: 2305
Joined: Sun Mar 04, 2007 3:47 pm

Post by matt » Fri Jan 21, 2011 4:04 pm

I believe that there are occasional periods when the broad stock market is overvalued enough that one might choose to reduce exposure or step aside completely. However, I don't think this long-term timing idea will work well for most investors.

If you want to be a valuation-oriented investor, you'll probably do better over time by staying fully invested in value stocks as opposed to trying to time the overall market.

SP-diceman
Posts: 3968
Joined: Sun Oct 05, 2008 9:17 am

Post by SP-diceman » Fri Jan 21, 2011 4:33 pm

The problem with long-term market timing is it
takes too long to find out if your right or wrong.

User avatar
wade
Posts: 539
Joined: Fri Sep 17, 2010 12:38 am
Location: Main Line
Contact:

Post by wade » Sat Jan 22, 2011 3:59 am

Thank you to everyone for the comments.

afan: Data mining is always going to be a legitimate concern. I would really like to do the same study with international data, for sure. The trouble is, at this point I have been unable to find a source for P/E data, or even a source that divides between dividends and capital gains/losses, over a long period of time. I know that MSCI has P/E data going back to 1970 for some countries, but I haven't figured out how to gain access to that yet, and I haven't tried real hard to find it since 1970 is not so far back anyway. If you know of anything, please let me know. And thank you for the reading recommendation and other comments.

DRiP Guy: I'm not sure if you are criticizing me or not. I read "A Random Walk Down Wall Street" when I first starting drawing a salary, and I follow the advice there quite closely for my own personal investments. I'm a fully indexed Vanguard customer, except for the TIPS fund. And I will tell you how I came across the Boglehead Forum. When I was doing background research for the paper,

"An International Perspective on Safe Withdrawal Rates: The Demise of the 4 Percent Rule?"
http://www.fpanet.org/journal/CurrentIs ... awalRates/

(which you complimented before, thank you)

I was trying to figure out whether anyone had conducted such a study before. From Google, I came across a thread here where someone had just made an offhand comment that it would be nice if someone would use the DMS data (that's the short hand name for the dataset I used) to study this question. Well, that is an idea I already had by that point, but naturally I started reading more to see if I can find research ideas and/or ideas about what the current hot topics of interest are.

I don't think that is enough to warrant a co-authorship with Bogleheads, but I do acknowledge Bogleheads. I'm very grateful for the thoughtful comments of the many intelligent people who post here and who have real world investing experience. In the last paper I finished, I thanked the Bogleheads Forum. Here is the acknowlegements section from that paper:

Acknowledgements: The author is grateful to the Bogleheads Forum for bringing his attention to
this research topic and for financial support from the Japan Society for the Promotion of Science
Grants-in-Aid for Young Scientists (B) #20730179.
http://ideas.repec.org/p/pra/mprapa/27487.html

And I will definitely be acknowledging a number of people here by name for this current paper.

What exactly do you want from me in this regard?

Next, you say that my philosophy is at odds with John Bogle's investment philosophy. I did check _Enough_ again, and I agree that Taylor is right, and that Mr. Bogle "eschews market timing".

But reading other parts of that book, I wonder if there can be room for some advances. I'm at home without my copy of the book now, so I can't quote directly, but what I want to quote is typical textbook style stuff:

Mr. Bogle explains that stock returns are divided into 3 components: underlying earnings growth, the dividend yield, and changes in the P/E multiple. He calls earnings and dividends the source of INVESTMENT returns, and he calls changing P/E multiples the source of SPECULATIVE returns. He notes that the P/E ratio was about 15 in 1900 and was about 15 in 2007.

That part is clear enough. Next, may be my misinterpetation, but I believe he is saying that speculative returns are transitory, random, and it is foolish to try to chase trends in the P/E ratio. No market timing.

But the big question I have, and which would help my understanding a great deal about this issue is:

WHY IS IT SUCH AN ODIOUS VIOLATION OF THE TENETS OF BOGLEHEADISM TO EXPLORE WHETHER SOMEONE WHO HAS ENOUGH PATIENCE AND ENOUGH TIME ON THEIR HANDS MIGHT BE ABLE TO BENEFIT FROM THE TRANSITORY NATURE OF SPECULATIVE RETURNS (I.E. THE IDEA THAT THE P/E RATIO EVENTUALLY ENDS UP WHERE IT STARTED)?

Acceptable answers at this point are not:
-because the investor might be sitting on the sidelines forever and miss stock returns
-because the investor doesn't have enough time to do it

Because the example I provided before shows this is not the case (at least with backtested historical data) . But there might be a good answer, and I look forward to hearing it.

On a lighter note, I promise that my next research topic will not be controversial in this regard. Lots of finance research has a hard time incorporating taxes (including mine), making it difficult to answer questions like: should I invest in a Roth IRA or a regular IRA, and how frequently should I rebalance my portfolio? I hope to develop a way to incorporate taxes into Monte Carlo simulations to be able to provide better answers for these types of questions.

Orion: Are you aware of Shiller offering asset allocation advice based on PE10? And other studies like Stein and DeMuth, and the two papers by Fisher and Statman only consider all-or-nothing stocks or bills strategies. If you read Rob Bennett's stuff carefully, I think he did provide an important contribution in terms of describing a way for PE10 to guide asset allocation for long-term conservative investors. I also think he was right on the issue of safe withdrawal rates. Even if VII ends up being wrong-headed, his heart does seem to be in the right place no matter whatever it is YOU might think about other aspects of his personality.

Bob90245: You are absolutely right. My idea about "long-term" is that you have a particular asset allocation strategy that you commit yourself to following over the long-term, but indeed you may have to make changes to your asset allocation rather frequently at some points in time. In that regard, tactical asset allocation may be a better term to use and I will think about this some more.

Norbert: Thank you for your valuable comments as well, and I will think about them a lot. It almost seems, though, that you are suggesting we know nothing. In this case, what is the default recommended approach for someone to develop a lifetime asset allocation plan?

peter71: Yes, adjusting standard errors for overlapping observations is a complicated topic. I don't have much to add, except that if this is a topic that interests you, then you might like this paper:

Britten-Jones, Mark, Anthony Neuburger, and Ingmar Nolte. 2010. "Improved Inference and Estimation in Regressions with Overlapping Observations." The Review of Financial
Studies. forthcoming.

A working paper version is available if you search for that title.

The problem is that even this recent and sophisticated method can't be used for my research. I discuss this issue in a paper that I've now linked to several times. Here it is again:
http://ideas.repec.org/p/pra/mprapa/27487.html

Thanks.

Edit: I forgot to mention regarding acknowledgements, in my last paper I also cited the work of Boglehead Forum member and active contributor on this thread, Bob90245:

Bob’s Financial Website. 2008. "SWRs versus Returns." Available from
http://www.bobsfinancialwebsite.com/ReturnsVsSWRs.html (Accessed on September 27,
2010).
Last edited by wade on Sat Jan 22, 2011 5:53 am, edited 1 time in total.

Post Reply