Worst ten years

Discuss all general (i.e. non-personal) investing questions and issues, investing news, and theory.
Post Reply
Topic Author
cjking
Posts: 2039
Joined: Mon Jun 30, 2008 4:30 am

Worst ten years

Post by cjking »

I was updating my spreadsheet with the Shiller data, in which I calculate the total return going forward for periods from 1 to 40 years, from each month. The ten years from March 1999 is the worst ten years on record, with a return of -6.1% per year.

My spreadsheet shows the index at 253% of fair value according to PE10, and 255% according to "q", in March 1999.

Previous bad starting dates were in August 1972 and December 1964, when there was a return of -4.0% a year over the following ten years.

Surprisingly, the return after the 1929 crash was not as bad, the worst period was from August 1929, -1.8% a year.

Prior to 1964, the previous similarly bad starting year was June 1911, with a return of -4.7% a year.

The best ten years in the run-up to 2000 was from September 1990, 15% a year. The all-time best ten years was from September 1919, 18% a year.
livesoft
Posts: 85971
Joined: Thu Mar 01, 2007 7:00 pm

Post by livesoft »

Wow, "the worst ten years on record" doesn't seem so bad at all in hindsight. Since we survived those years quite nicely and in style, I guess we can predict that we will do OK for the next 10 years as well?
User avatar
Adrian Nenu
Posts: 5228
Joined: Thu Apr 12, 2007 6:27 pm

Post by Adrian Nenu »

You have to be careful and avoid those "bad starting dates" and only invest during the "good starting dates". :wink:

Steven Evanson shows on his website that entry point is important and buying stocks during bear markets can lead to higher return averages. Conversly, buying at the top of bull markets/speculative bubbles leads to bad return averages.

http://www.evansonasset.com/index.cfm?Page=13

Like I have shown before, a 50/50 composed of equal parts TIPS, Total Bond Market, Total Stock Market and Total International indexes returned 2.65% annually over the last 10 years.

So when you determine your asset allocation, you have to account for the fact that stocks can have long periods of very low returns.
2) Real returns differ substantially depending upon time of market entry, and for the typical investor with a twenty or twenty-five year time-frame, this can make very large differences in money available during retirement. Frequently presented "mountain" charts smooth out and conceal what would have happened to unfortunate investors who happened to enter the market at or near long-term secular bull market peaks.

An investor in the S & P 500 from 1929 through 1949 received an inflation adjusted return of 4.54%. Yet, an investor beginning in 1932, and holding until 1951, received an inflation adjusted annual return of 10.84%, over 6% greater per year. Our 1929 investor received a compound total return of 84.36% for twenty years; our 1932 investor received a compound total return of 818.13% for twenty years, and ended up with almost ten times as much as our unfortunate 1929 investor. Extend the holding period out to twenty-five years, and our 1929 investor does better, receiving a compound total return of 319.33%, but our lucky investor still receives far more, 2202%.

Let's compare returns for the last big market top and bottom. An investor in 1968, a market top, had to wait until 1983, fourteen years to just breakeven after inflation. If they waited until 1987, twenty years, their annual return was 4.19% after inflation, and their compound total return was 489.24%. If they waited until 1992, twenty-five years, their annual return was 5.83%, compounded to1132%. Today's big bull market began in the early 1980's, so only twenty years or so data is available at this point if we consider the 2000-2002 sell-off as a correction. But, an investor in the market from 1981 through 2000 received a whopping after inflation return of 12.91% annually, and a compound total return of 1741%; it was the best twenty years in U.S. market history. Yet, an investor in Japan at its peak in 1989 still has not recovered to breakeven 17 years later in 2007. Japanese residential real estate and equities are still down around 60%. Time of market entry matters yet there's not evidence markets can be timed.
Adrian
anenu@tampabay.rr.com
chaz
Posts: 13604
Joined: Tue Feb 27, 2007 1:44 pm

Post by chaz »

Maybe we are in the worst 10 year period now - only time will tell.
Chaz | | “Money is better than poverty, if only for financial reasons." Woody Allen | | http://www.bogleheads.org/wiki/index.php/Main_Page
Triple digit golfer
Posts: 10430
Joined: Mon May 18, 2009 5:57 pm

Post by Triple digit golfer »

It's probably very likely that no one could successfully time the market repeatedly. But that doesn't mean that after a crash, going more into stocks, or after a big run up, taking some out of stocks, is necessarily market timing or a bad thing. It's just buying less stocks because you feel they're expensive. I'd buy more lunches out if Jimmy John's sold sandwiches for a buck. I'd buy far less if they started selling them for $10 each. Same thing with stocks. If the S&P 500 suddenly skyrocketed and a year from now was at 2,500, I think many, many people on this board would lock in their gains and move a lot of money into bonds.

And people not on this board would just start going into stocks at that point :)

The problem with this strategy is that in 1997, someone could have said the same thing. Stocks have had many good years, I'll move to bonds. Then it kept going up. And up. And up. So "expensive" is hard to tell in the stock market until after the fact.

Me personally, I plan on sticking to my AA regardless of what the market does, but will I actually stick to it? I don't know. If we have five years of 25% annualized gains, I may be inclined to change that desired AA and up the bond portion.
Topic Author
cjking
Posts: 2039
Joined: Mon Jun 30, 2008 4:30 am

Post by cjking »

Adrian Nenu wrote:You have to be careful and avoid those "bad starting dates" and only invest during the "good starting dates". :wink:
I know you were joking, but the measure that "q" is shown to predict very well in "Valuing Wall Street" is called "hindsight value" because the goodness/badness of starting dates is exactly what it tries to capture.

"Hindsight value" is defined as the average annual return experienced by 40 investors with holding period of 1 to 40 years respectively.

The averaging across the 40 different end-dates causes the effect of good and bad end-dates to cancel each other out, so the average then becomes a measure of the merit of the starting date only.

When they overlay the chart of "q" on that of hindsight value, they track each other very closely.

My current view is that it's not how good or bad the starting date is in historical terms that matters, but how equities at that valuation compare to alternative investments.

For comparison purposes I have to translate the valuation level into an expected return going forward.

To use an example I've quoted a few times, at the end of 1998 the projected yield on equities was less than the yield on TIPS.

(Here's my version of the chart that shows "PE10" and "q" predictors versus hindsight value, where hindsight value is in green end ends 40 years ago. It was created last year so is not totally up-to-date.)

<a href="http://www.flickr.com/photos/31048087@N05/3062871365/" title="yield_chart4 by cjk1896, on Flickr"><img src="http://farm4.static.flickr.com/3008/306 ... 8189_o.jpg" width="779" height="562" alt="yield_chart4"></a>
Rodc
Posts: 13601
Joined: Tue Jun 26, 2007 9:46 am

Post by Rodc »

The averaging across the 40 different end-dates
Which effectively means you have 3 independent data points...

The basic idea is not so bad, but using 20 years would at least give you 6 data points, still hardly enough to say much, but better.
We live a world with knowledge of the future markets has less than one significant figure. And people will still and always demand answers to three significant digits.
grayfox
Posts: 5569
Joined: Sat Sep 15, 2007 4:30 am

Post by grayfox »

cjking wrote: For comparison purposes I have to translate the valuation level into an expected return going forward.

To use an example I've quoted a few times, at the end of 1998 the projected yield on equities was less than the yield on TIPS.

(Here's my version of the chart that shows "PE10" and "q" predictors versus hindsight value, where hindsight value is in green end ends 40 years ago. It was created last year so is not totally up-to-date.)
How do you translate valuation level into expected return for PE10 and q?
Topic Author
cjking
Posts: 2039
Joined: Mon Jun 30, 2008 4:30 am

Post by cjking »

Rodc wrote:Which effectively means you have 3 independent data points...
I can't help feeling that if anyone looks at that chart and concludes it contains insufficient information to affect their non-valuation based strategy, and that there's no reason, everything else being equal, why their equity allocation should be any different in 1929, 1968 and 2000 than it was in say 1920, 1933 or 1982, then no amount of data points will change their mind.

For me, the definition of PE10 is sufficient reason to use its reciprocal as a reference point, the rolling corellation through-out all available history is merely a bonus.
Topic Author
cjking
Posts: 2039
Joined: Mon Jun 30, 2008 4:30 am

Post by cjking »

grayfox wrote:
cjking wrote: For comparison purposes I have to translate the valuation level into an expected return going forward.

To use an example I've quoted a few times, at the end of 1998 the projected yield on equities was less than the yield on TIPS.

(Here's my version of the chart that shows "PE10" and "q" predictors versus hindsight value, where hindsight value is in green end ends 40 years ago. It was created last year so is not totally up-to-date.)
How do you translate valuation level into expected return for PE10 and q?
For PE10 I have two methods. The quick one is just to take the reciprocal. The one my historical spreadsheet uses is to divide the current value by the average of all recorded values. Actually I prefer the quick method, as it is more intellectually justifiable. If I want to data mine, I can show that 0.884/PE10 improves on 1/PE10, if I want to maximise the correlation with 40-year hindsight value.

For q I simply adjust a long-term expected return of 6% up or down proportionately according the the level of over or under-valuation.

I chose the 6% figure by as far as possibly measuring the whole of recorded history subject to the constraint of near identical valuation levels at the beginning and end. In practise this means choose the current value of the index as the end date, then find the earliest recorded value when valuation levels were the same as today for the starting date.

All methods produce similar results, but I like 1/PE10 as it's easy to understand why it should tell you something useful. One could justify using it even if one had no historical return data.

Here's a chart that shows 0.884/PE10 as a predictor, versus history, the blue line. History means the average hindsight value corresponding to each PE10 starting value. You will see that the predictor works well for most values of PE10, but at both extremes breaks down, reality performs worse than predicted. This is also visible on the chart above, where the green line spikes lower than the predictors.

<a href="http://www.flickr.com/photos/31048087@N05/3588549144/" title="PE10_predictor by cjk1896, on Flickr"><img src="http://farm3.static.flickr.com/2442/358 ... 7de9_o.jpg" width="779" height="440" alt="PE10_predictor"></a>
User avatar
spam
Posts: 907
Joined: Tue Jun 10, 2008 9:47 am

Post by spam »

William J. Bernstein copyrighted this in 1999:
The dividend yield was 2.6% in September 1929, and for the 30 years after that earnings growth was only 1.8%. Thus, had the crash not occurred then stocks would have returned 4.4% per year, resulting in a "break-even" point with what actually occurred of January 1952, or 22 years, almost exactly the same period predicted by the duration model. Viewed from this perspective, today's market is a good deal more frightening than that of 1929, since a 75% stock decline produces a duration of 19 years at the 2.6% 1929 yield, versus 33 years at the current 1.4% yield.

Certainly, such a wrenching market decline today would wreak havoc on the financial and social structure of the republic, as it did 70 years ago. But at the same time, today's high prices and resultant low yields are no great blessing either.
Interesting what Bernstein had to say about the stock market in 1999. Thanks to Dan Kohn, I got to read the entire article. It is here:

http://www.efficientfrontier.com/ef/999/duration.htm
Topic Author
cjking
Posts: 2039
Joined: Mon Jun 30, 2008 4:30 am

Post by cjking »

In case anyone needs cheering up, it occurred to me to measure what happened in the decade after the four bad decades documented above.

In ten years from June 1921, the average annual real return was 14.2%, from August 1939 the return was 2.9%, from December 1974 6.6%, from August 1982 13.4%. Overall average 9.3%.
Rodc
Posts: 13601
Joined: Tue Jun 26, 2007 9:46 am

Post by Rodc »

cjking wrote:
Rodc wrote:Which effectively means you have 3 independent data points...
I can't help feeling that if anyone looks at that chart and concludes it contains insufficient information to affect their non-valuation based strategy, and that there's no reason, everything else being equal, why their equity allocation should be any different in 1929, 1968 and 2000 than it was in say 1920, 1933 or 1982, then no amount of data points will change their mind.

For me, the definition of PE10 is sufficient reason to use its reciprocal as a reference point, the rolling corellation through-out all available history is merely a bonus.
That is fine, but arguing by definition is entirely different than arguing from data.

On the one hand valuations having something to do with long term returns makes sense. But how much and how long and how reliably require data and we don't have much. If you think you should take action X, how likely is to help or hurt and by how much. Rebalancing is a mild reaction to valuations, jumping between 0% and 100% stocks is a wild reaction; which is best or where in between is best? Absolutely no one knows.

So, we are left with mere gut reaction as to how to use valuation data.

My point is not that one should simply ignore valuations, but one should be aware of how little any claims about what should be done with valuations hang on.
We live a world with knowledge of the future markets has less than one significant figure. And people will still and always demand answers to three significant digits.
Post Reply