Dynamic AA - does it make sense?

Discuss all general (i.e. non-personal) investing questions and issues, investing news, and theory.
leonard
Posts: 5993
Joined: Wed Feb 21, 2007 10:56 am

Re: Dynamic AA - does it make sense?

Post by leonard »

AppelSienSapFrietjes wrote:Hi,

I have been pondering about the following. Say you have a standard AA between stocks and bonds, for example 70/30. Does it, in your opinion, ever make sense to forgo your 70/30, and go to 90/10 (or 80/20 or 100/0 or ...) during a certain period because of bad past results of the equities? If you had 70/30 before 2008, and went to 100/0 (the extreme) after/during the market collapsed, you would be buying low. Years later, you will benefit.

To begin with, I don't think this is market timing, since you are over weighting assets which have done bad in the past.

You need some kind of threshold to decide that you forgo your AA in favor of more equities. A threshold might be that the market has been down for XX% (for example 30%) since it's highs. You can never predict (and you don't want to try - market timing) the lowest point, so you need some kind of threshold. Once the equities have been recovered - again, a threshold is needed here -, you can again convert to 70/30.

Does this make sense? What is your opinion?

Dynamic AA = Market Timing = Bad Move.
Leonard | | Market Timing: Do you seriously think you can predict the future? What else do the voices tell you? | | If employees weren't taking jobs with bad 401k's, bad 401k's wouldn't exist.
Topic Author
AppelSienSapFrietjes
Posts: 93
Joined: Thu Jun 20, 2013 5:50 am

Re: Dynamic AA - does it make sense?

Post by AppelSienSapFrietjes »

One quick interesting outcome, for now.

(1) 100/0 from 1972 to 2012

Code: Select all

period: 1972 - 2012
default AA: 100.0 / 0.0
tilt AA: 100.0 / 0.0
threshold down: 30.0
threshold up: 42.85714285714286

=> portfolio after yearly balancing: 4688.455280997256

tiling starts at begin of 1975
tiling stops at end of 1976
tiling starts at begin of 2003
tiling stops at end of 2004
tiling starts at begin of 2009
tiling stops at end of 2010

=> portfolio after tilted balancing: 4688.455280997256
(2) 50/50 from 1972 to 2012. If the market drops 30% we switch to 100/0. Once this drop has been recovered, we switch back to 50/50:

Code: Select all

period: 1972 - 2012
default AA: 50.0 / 50.0
tilt AA: 100.0 / 0.0
threshold down: 30.0
threshold up: 42.85714285714286

=> portfolio after yearly balancing: 3613.033969438202

tiling starts at begin of 1975
tiling stops at end of 1976
tiling starts at begin of 2003
tiling stops at end of 2004
tiling starts at begin of 2009
tiling stops at end of 2010

=> portfolio after tilted balancing: 5746.280777338504
I would say the first portfolio has much more risk than the second one, since in the first, you are always 100/0, while in the second, you are only a few years 100/0 and mostly 50/50.

The second appears to return better in a tilted balancing setup: x57 versus x46.
In case of a default rebalancing setup, the first returned better, as expected: x46 versus x36.
User avatar
Clearly_Irrational
Posts: 3087
Joined: Thu Oct 13, 2011 3:43 pm

Re: Dynamic AA - does it make sense?

Post by Clearly_Irrational »

leonard wrote:Dynamic AA = Market Timing = Bad Move.
I don't necessarily agree. It would depend heavily on what was being proposed and whether or not the system was "mechanical". The first hurdle is, can you produce a system that will backtest with better risk adjusted returns on both in and out of sample data. For example, if you develop a great system using the US market, does it work equally well on the European market?
leonard
Posts: 5993
Joined: Wed Feb 21, 2007 10:56 am

Re: Dynamic AA - does it make sense?

Post by leonard »

Clearly_Irrational wrote:
leonard wrote:Dynamic AA = Market Timing = Bad Move.
I don't necessarily agree. It would depend heavily on what was being proposed and whether or not the system was "mechanical". The first hurdle is, can you produce a system that will backtest with better risk adjusted returns on both in and out of sample data. For example, if you develop a great system using the US market, does it work equally well on the European market?
You seem to be assuming machinery that was not in the OP's post.
Leonard | | Market Timing: Do you seriously think you can predict the future? What else do the voices tell you? | | If employees weren't taking jobs with bad 401k's, bad 401k's wouldn't exist.
Topic Author
AppelSienSapFrietjes
Posts: 93
Joined: Thu Jun 20, 2013 5:50 am

Re: Dynamic AA - does it make sense?

Post by AppelSienSapFrietjes »

leonard wrote:
Clearly_Irrational wrote:
leonard wrote:Dynamic AA = Market Timing = Bad Move.
I don't necessarily agree. It would depend heavily on what was being proposed and whether or not the system was "mechanical". The first hurdle is, can you produce a system that will backtest with better risk adjusted returns on both in and out of sample data. For example, if you develop a great system using the US market, does it work equally well on the European market?
You seem to be assuming machinery that was not in the OP's post.
Indeed, machinery. No smart ass decision making, but only decisions based on price drops and gains.
Topic Author
AppelSienSapFrietjes
Posts: 93
Joined: Thu Jun 20, 2013 5:50 am

Re: Dynamic AA - does it make sense?

Post by AppelSienSapFrietjes »

Another interesting comparsion:

(1) always 70/30

Code: Select all

period: 1972 - 2012
default AA: 70.0 / 30.0
tilt AA: 70.0 / 30.0
threshold down: 30.0
threshold up: 42.85714285714286

=> portfolio after yearly balancing: 4169.3385594629235

tiling starts at begin of 1975
tiling stops at end of 1976
tiling starts at begin of 2003
tiling stops at end of 2004
tiling starts at begin of 2009
tiling stops at end of 2010

=> portfolio after tilted balancing: 4169.3385594629235
(2) 0/100 and 70/30 during downturns of -30%

Code: Select all

period: 1972 - 2012
default AA: 0.0 / 100.0
tilt AA: 70.0 / 30.0
threshold down: 30.0
threshold up: 42.85714285714286

=> portfolio after yearly balancing: 2055.884804740879

tiling starts at begin of 1975
tiling stops at end of 1976
tiling starts at begin of 2003
tiling stops at end of 2004
tiling starts at begin of 2009
tiling stops at end of 2010

=> portfolio after tilted balancing: 4142.966723605086
Returns are both x41 but risk is much lower in second portfolio.

(3) 0/100 and 70/30 during downturns of -37% instead of -30%

Code: Select all

period: 1972 - 2012
default AA: 0.0 / 100.0
tilt AA: 70.0 / 30.0
threshold down: 37.0
threshold up: 58.73015873015872

=> portfolio after yearly balancing: 2055.884804740879

tiling starts at begin of 1975
tiling stops at end of 1976
tiling starts at begin of 2003
tiling stops at end of 2006
tiling starts at begin of 2009
tiling stops at end of 2012

=> portfolio after tilted balancing: 4727.117025294424
This yields even better, since we are staying longer in a rising equity market. Risk compared with (2) is of course higher. It is of course still much lower than in (1).

:shock: :shock:

(4) Is the downturn threshold 38% instead of 37%, we miss a lot of years in equities. This is a disaster.

Code: Select all

period: 1972 - 2012
default AA: 0.0 / 100.0
tilt AA: 70.0 / 30.0
threshold down: 38.0
threshold up: 61.290322580645174

=> portfolio after yearly balancing: 2055.884804740879

tiling starts at begin of 1975
tiling stops at end of 1976

=> portfolio after tilted balancing: 2691.5425823097175
User avatar
Ranger
Posts: 446
Joined: Tue Feb 20, 2007 9:27 pm
Location: Wolverine Nation

Re: Dynamic AA - does it make sense?

Post by Ranger »

Valuation matters. Here is the withdrawal rate based on tactical allocation. They change asset allocation from 50-50 to stock allocation to 25,50 and 75 based on valuations. Not exactly answers your question, but along the same lines.

http://www.fpanet.org/journal/withdrawa ... valuation/

Here is wade's analysis in Japan's market along the same lines.

http://wpfau.blogspot.com/2011/03/could ... ation.html

Here is crestmont research graph

Image

Image
YDNAL
Posts: 13774
Joined: Tue Apr 10, 2007 4:04 pm
Location: Biscayne Bay

Re: Dynamic AA - does it make sense?

Post by YDNAL »

Clearly_Irrational wrote:
YDNAL wrote:
Clearly_Irrational wrote:
YDNAL wrote:How do you know when the market stops "collapsing" ?
Obviously there is no way to know for sure, however based on previous events you'd have reasonably good odds betting that most crashes will max out at a 50% loss.
Think again ! There were multiple upswings, followed by downswings, down to near 90% drop in 3 years.
Yes, I was well aware of the great crash and the fact that it was over 50%, that doesn't invalidate my statement.
"Reasonably good odds" are best-left to a betting person or those willing to engage in a definition contest. Thus, a potential remedy [just saying] would be to qualify such statements.

By the way, you made your statement using my quote. If would be rude not to respond and correct - where applicable.

Question: how many "crashes" of significant nature would you consider having occurred in history ?
Landy | Be yourself, everyone else is already taken -- Oscar Wilde
Topic Author
AppelSienSapFrietjes
Posts: 93
Joined: Thu Jun 20, 2013 5:50 am

Re: Dynamic AA - does it make sense?

Post by AppelSienSapFrietjes »

Another interesting outcome. This time from 1982 to 2012.

(1) 100/0 all the time

Code: Select all

period: 1982 - 2012
default AA: 100.0 / 0.0
tilt AA: 100.0 / 0.0
threshold down: 35.0
threshold up: 53.84615384615383

=> portfolio after yearly balancing: 2409.8062178866767

tiling starts at begin of 2003
tiling stops at end of 2005
tiling starts at begin of 2009
tiling stops at end of 2012

=> portfolio after tilted balancing: 2409.8062178866767
(2) 30/100 and only 100/0 during downturns of 35%.

Code: Select all

period: 1982 - 2012
default AA: 30.0 / 70.0
tilt AA: 100.0 / 0.0
threshold down: 35.0
threshold up: 53.84615384615383

=> portfolio after yearly balancing: 1533.7263848472774

tiling starts at begin of 2003
tiling stops at end of 2005
tiling starts at begin of 2009
tiling stops at end of 2012

=> portfolio after tilted balancing: 2439.819208149068
If somebody can provide me downloadable/parsable data, I can improve this greatly.

There is now one big caveat: balancing happens at the end of the year. This has impact:

- If there is a crash of 35% and a restore of 80% (for example) in the next year, the whole 80% is included, while we only requested to allocate according the tilt AA during a restore of 53.84%. We change too late.
- Likewise, if there is a crash but it is restored within the year, we will never change to the tilt AA. We do not change while we should have.
- If there is a crash of 80% in one year, we change after 80% (at the end of the year), and not after 35% (somewhere during of year), as requested. We change too late.
- ...

I need daily stock prices instead of yearly returns to be table to fix that.
Last edited by AppelSienSapFrietjes on Tue Jul 16, 2013 6:01 pm, edited 2 times in total.
User avatar
nisiprius
Advisory Board
Posts: 52211
Joined: Thu Jul 26, 2007 9:33 am
Location: The terrestrial, globular, planetary hunk of matter, flattened at the poles, is my abode.--O. Henry

Re: Dynamic AA - does it make sense?

Post by nisiprius »

Ranger wrote:Here is crestmont research graph

Image
That is an unconvincing graph. Do they have confidence limits for the slope? Is it statistically significantly different from zero?
Annual income twenty pounds, annual expenditure nineteen nineteen and six, result happiness; Annual income twenty pounds, annual expenditure twenty pounds ought and six, result misery.
Rodc
Posts: 13601
Joined: Tue Jun 26, 2007 9:46 am

Re: Dynamic AA - does it make sense?

Post by Rodc »

Unless I missed in a quick read you still are not comparing apples to apples. In your timed portfolios you have greater average exposure to stocks. You really need to fix that before you have anything interesting.

Over that time period, just about anything you do to increase stock exposure will increase returns.

And you report no stats at all on risk metrics.
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.
Topic Author
AppelSienSapFrietjes
Posts: 93
Joined: Thu Jun 20, 2013 5:50 am

Re: Dynamic AA - does it make sense?

Post by AppelSienSapFrietjes »

Rodc wrote:Unless I missed in a quick read you still are not comparing apples to apples. In your timed portfolios you have greater average exposure to stocks. You really need to fix that before you have anything interesting.

Over that time period, just about anything you do to increase stock exposure will increase returns.
Hmm, I am missing something. I did three comparisons:

(1) 1972-2012: 100/0 always <=> 50/50 most of the time and only 100/0 during certain periods.
(2) 1972-2012: 70/30 always <=> 0/100 most of the time and only 70/30 during certain periods.
(3) 1982-2012: 100/0 always <=> 30/70 most of the time and only 100/0 during certain periods.

So I never have greater average exposure to stocks, if I am not mistaken. Actually, the exposure to stocks is much much much less in these 3 examples.

(I might be wrong, it is getting late here.)

To be sure, XX/YY means STOCKS/BONDS here.
Rodc wrote:And you report no stats at all on risk metrics.
Give me a few days (if I can find time), this requires more time to implement. As far as I can see, risk is always lower in the three comparisons above, since these are less exposed to stocks.
Rodc
Posts: 13601
Joined: Tue Jun 26, 2007 9:46 am

Re: Dynamic AA - does it make sense?

Post by Rodc »


Hmm, I am missing something. I did two comparisons:

(1) 1972-2012: 70/30 always <=> 0/100 most of the time and only 70/30 during certain periods.
(2) 1982-2012: 100/0 always <=> 30/70 most of the time and only 100/0 during certain periods.

So I never have greater average exposure to stocks, if I am not mistaken.
in (1) the stock heavy portfolio wins if I am reading things correctly (ie timing loses, but it does not have much invested in stocks)

In (2) the stock heavy portfolio again wins if I am reading things correctly (ie timing loses, but does have much invested in stocks)

In the cases I saw where timing won, timing was more heavily invested in stocks. Not the most easy to read format which when you have time to write up I am sure will be better.

FWIW: long day for me too, so I might not be entirely clear headed myself.
Last edited by Rodc on Tue Jul 16, 2013 6:14 pm, edited 1 time in total.
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.
User avatar
Clearly_Irrational
Posts: 3087
Joined: Thu Oct 13, 2011 3:43 pm

Re: Dynamic AA - does it make sense?

Post by Clearly_Irrational »

YDNAL wrote:
Clearly_Irrational wrote:Yes, I was well aware of the great crash and the fact that it was over 50%, that doesn't invalidate my statement.
"Reasonably good odds" are best-left to a betting person or those willing to engage in a definition contest. Thus, a potential remedy [just saying] would be to qualify such statements.

Question: how many "crashes" of significant nature would you consider having occurred in history ?
I'm sure someone has a better list somewhere, but how about something like this: http://dprogram.net/2008/10/08/the-10-w ... s-history/ and add the 2008 crash.
Rodc
Posts: 13601
Joined: Tue Jun 26, 2007 9:46 am

Re: Dynamic AA - does it make sense?

Post by Rodc »

nisiprius wrote:
Ranger wrote:Here is crestmont research graph

Image
That is an unconvincing graph. Do they have confidence limits for the slope? Is it statistically significantly different from zero?
No listing of R^2 either.

The problem in finance is that a lot of not terribly stats sophisticated people run data through canned stats packages without really understanding the results.

To be fair, I don't always take the time to do things carefully either. But I'm just an anonymous guy posting on the internet. Somethings I do are more rigorous and some more seat of the pants. :)
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.
Topic Author
AppelSienSapFrietjes
Posts: 93
Joined: Thu Jun 20, 2013 5:50 am

Re: Dynamic AA - does it make sense?

Post by AppelSienSapFrietjes »

Just to be sure, how to read my output:

Code: Select all

period: 1982 - 2012
default AA: 30.0 / 70.0
tilt AA: 100.0 / 0.0
threshold down: 35.0
threshold up: 53.84615384615383

=> portfolio after yearly balancing: 1533.7263848472774

tiling starts at begin of 2003
tiling stops at end of 2005
tiling starts at begin of 2009
tiling stops at end of 2012

=> portfolio after tilted balancing: 2439.819208149068
The default AA is 30% stocks, 70% bonds. The tilt AA is 100% stocks and 0% bonds.

The tilt AA is chosen when the market drops 35%. The default AA is picked again when the market restores 53.84%.

"=> portfolio after yearly balancing: 1533.7263848472774" means that your end portfolio is 1533 if you do standard rebalancing, as everyone knows.

"=> portfolio after tilted balancing: 2439.819208149068" means that your end portfolio is 2439 if you follow the tilted/timed rebalancing according to the variables set above.

The years during which the tilt AA is chosen are also printed.
Last edited by AppelSienSapFrietjes on Tue Jul 16, 2013 6:24 pm, edited 2 times in total.
Topic Author
AppelSienSapFrietjes
Posts: 93
Joined: Thu Jun 20, 2013 5:50 am

Re: Dynamic AA - does it make sense?

Post by AppelSienSapFrietjes »

Rodc wrote:

Hmm, I am missing something. I did two comparisons:

(1) 1972-2012: 70/30 always <=> 0/100 most of the time and only 70/30 during certain periods.
(2) 1982-2012: 100/0 always <=> 30/70 most of the time and only 100/0 during certain periods.

So I never have greater average exposure to stocks, if I am not mistaken.
in (1) the stock heavy portfolio wins if I am reading things correctly (ie timing loses, but it does not have much invested in stocks)

In (2) the stock heavy portfolio again wins if I am reading things correctly (ie timing loses, but does have much invested in stocks)

In the cases I saw where timing won, timing was more heavily invested in stocks. Not the most easy to read format which when you have time to write up I am sure will be better.

FWIW: long day for me too, so I might not be entirely clear headed myself.
For your interest, I had updated my post during your comment and I have added a how-to-read. I hope this clarifies.
User avatar
Ranger
Posts: 446
Joined: Tue Feb 20, 2007 9:27 pm
Location: Wolverine Nation

Re: Dynamic AA - does it make sense?

Post by Ranger »

nisiprius wrote:That is an unconvincing graph. Do they have confidence limits for the slope? Is it statistically significantly different from zero?
No, They don't list any other stat with that. This is another graph accompanying with that.

Image

Since these are based on Shiller data, it shouldn't be too difficult to recreate these charts. I will do it, when i have some spare time.
Topic Author
AppelSienSapFrietjes
Posts: 93
Joined: Thu Jun 20, 2013 5:50 am

Re: Dynamic AA - does it make sense?

Post by AppelSienSapFrietjes »

After a night sleep, I was able to make some surprising - to me - conclusions.

- 1972 to 2012: 100% TSM -> 100 becomes 4688
- 1972 to 2012: 100% TBM -> 100 becomes 2055

Equities are fatter better than bonds, nothing new. Now, an extreme timed approach:

- 1972 to 2012: 100% TBM and only 100% TSM when equities dropped 35% until they won 53,8% (= 1/0.65 - the inverse of -35%) relative to that point -> 100 becomes 5910

A few years in the stock market is far better than always in the stock market. Even though 1982 to 1999 was a pure goldmine for stocks, the timed approach still wins. We had 0% equity exposure during 1982 to 1999!

Let's try to explain this. According to Simba's spreadsheet, the CAGR of TBM was 7,65%. So, indeed, a portfolio of 100 after 41 years is:

100 * (1.0765^41) = 2054

This matches with the value above. So far, so good.

During the period from 1972 to 2012, there were 3 crashes of -35% (even more): 1972-1973, 2000-2002 and 2008. The market recovered all three times. This means we had three times the opportunity for an extra 53,8% if we went all in stocks after a drop of 35%:

2054 * (1 / 0.65) * (1 / 0.65) * (1 / 0.65) = 7479

This is not completely correct, since this still includes the gains from the bond market during these recovery periods. This explains the difference between 7479 and 4688 above. This is not easy to calculate, so let's ignore this for now. It doesn't change much to get the point of what I am trying to say.

The CAGR is now: 74.79 ^ (1 / 41) = 1.11 = 11%

According to Simba's spreadsheet, CAGR of 100% TSM was 9.84%. So, indeed, the timed approach wins, given the assumption made earlier. Because of that assumption, please do not look at the exact numbers, but think about the effect of three times 53,8%.


This example of course didn't say much, since no real sane person would go from 100% TBM to 100% TSM. However, it gives you a feeling why this approach might work.

In previous posts, I made a more interesting example: 1972-2012: 70/30 always <=> 0/100 and only 70/30 during recovery periods. In this approach, the risk of the timed portfolio was significantly lower, since equity exposure was much lower. Yet, this performs the same of even better. See http://www.bogleheads.org/forum/viewtop ... 1#p1751641 for results. We had 0% equity exposure during 1982 to 1999, the mother of recent secular bull markets, and still win in 2012! We had 70/30, a standard portfolio, during a small set of recovery periods.

A conclusion from these very basic tests is: a lot of money can be made during a recovery. Actually, it doesn't harm to be over weighed in bonds for most of the time (and therefore losing value because of lack of equities), if you have the guts to migrate to equities during a downturn. This doesn't necessarily mean that you have to go for 100%: 0/100 most of the time and 70/30 during recoveries doesn't yield worse than 70/30 all the time, but it's much less risky! Having guaranteed money to invest during a downturn is golden. Giving up equity exposure in favor of bonds during bull markets to be able to invest with as much as possible guaranteed money during a recovery, isn't a bad idea, surprisingly. Returns are the same or even better, but risk is much lower, due to decrease of equity exposure.

I am happy to run more tests if somebody wants me to. If somebody has links to downloadable/parsable data, I am happy to improve my application. I will try to implement rolling returns somewhere in the next days, so we can do more interesting tests.
IlliniDave
Posts: 2388
Joined: Fri May 17, 2013 7:09 am

Re: Dynamic AA - does it make sense?

Post by IlliniDave »

AppelSienSapFrietjes wrote:Hi,

I have been pondering about the following. Say you have a standard AA between stocks and bonds, for example 70/30. Does it, in your opinion, ever make sense to forgo your 70/30, and go to 90/10 (or 80/20 or 100/0 or ...) during a certain period because of bad past results of the equities? If you had 70/30 before 2008, and went to 100/0 (the extreme) after/during the market collapsed, you would be buying low. Years later, you will benefit.

To begin with, I don't think this is market timing, since you are over weighting assets which have done bad in the past.

You need some kind of threshold to decide that you forgo your AA in favor of more equities. A threshold might be that the market has been down for XX% (for example 30%) since it's highs. You can never predict (and you don't want to try - market timing) the lowest point, so you need some kind of threshold. Once the equities have been recovered - again, a threshold is needed here -, you can again convert to 70/30.

Does this make sense? What is your opinion?
Certainly the idea makes sense at least on a superficial level. Accumulation techniques like value averaging indirectly use a similar tack, and John Bogle in his writing acknowledges it's something a "bold, aggressive investor" could consider (although only infrequently, only in small doses--no more than 10%--and it needs to be based on a rational assessment that certain asset classes will do do better/worse than others in the foreseeable future).

Simple rebalancing will accomplish similar things, although for different reasons.

The problem is when one makes such a move there's an underlying assumption that the "down" asset class will rebound in a reasonably short time (such time being of different length for every investor). In fact, it may actually fall farther and stay down longer than the investor can tolerate, so anything more than modest "tilts" increases one's risk substantially. It's not something you want to sit around and do on a weekly basis. But I must admit that if stocks plunged 40% next week, I'd be sorely tempted... :twisted:
Don't do something. Just stand there!
User avatar
HomerJ
Posts: 21281
Joined: Fri Jun 06, 2008 12:50 pm

Re: Dynamic AA - does it make sense?

Post by HomerJ »

AppelSienSapFrietjes wrote:
YDNAL wrote:
AppelSienSapFrietjes wrote:I have been pondering about the following. Say you have a standard AA between stocks and bonds, for example 70/30. Does it, in your opinion, ever make sense to forgo your 70/30, and go to 90/10 (or 80/20 or 100/0 or ...) during a certain period because of bad past results of the equities? If you had 70/30 before 2008, and went to 100/0 (the extreme) after/during the market collapsed, you would be buying low. Years later, you will benefit.
How do you know when the market stops "collapsing" ? How do you know when to buy for "years later to benefit" ?
  • We can look back and say... there, we should go 100% stocks around August 2003 and March 9, 2009.
  • Try to do that going forward !! - consistent market timing doesn't work (you may get lucky here/there).
Image
You don't know when the markets stop collapsing. That's not the point. You have a certain threshold (for example -30%), and then you alter in favor of equities. You might hit the gold spot (the very low), or not, you don't know at that point in time. But that's not really important. You are still better by this, than you when you keep your AA.

Whether you are 100/0, 90/10 or 60/40, you still believe in the markets. If not, you would have been (or would have wanted to go) to 0/100 or so. Given that assumption, confidence, it isn't a bad thing to alter your AA in favor of equities at certain times?

Once the market has been stabilized, you can again change to your original AA.
Here's what I know from personal experience... In 2008 I had a 60/40 AA. When the 2009 crash came, and the DOW dropped from 14000 to 10000, I I rebalanced... That means I sold bonds and bought stocks to bring myself back to 60/40. It was VERY VERY hard for me to do even this. Especially as the DOW continued to drop into the 8000s, 7000s, 6000s. I saw my portfolio drop from $750,000 to $500,000. That was huge and scary... It had taken me and my wife decades to save $750,000 and wham $250,000 of it was gone... I didn't rebalance again, because I didn't want to risk anymore. I can't imagine going to 100/0 or 90/10 during such a time... Some people did, but there's no way I could have handled it (my wife was already freaking out over the losses we had already experienced).

At no time did I think about selling out of the stock market... I knew it would come back someday... but I couldn't throw my bond money into stocks just in case it didn't.

I needed to know that, if even if the DOW crashed to 3000 and stayed there for 10 years, and we lost our jobs, we'd still have a solid chunk of money.

Note that it was very EASY for me to put NEW money into 100% stocks though... All during that crash, I changed our 401k contributions to 100% stock, and we even both got bonuses during that time which I put 100% in stocks. Strangely, that felt good.. That was investing NEW money at low prices... I just couldn't bring myself to risk more of my old safe money.
User avatar
Chan_va
Posts: 869
Joined: Wed Dec 05, 2012 6:15 pm

Re: Dynamic AA - does it make sense?

Post by Chan_va »

AppelSienSapFrietjes wrote:After a night sleep, I was able to make some surprising - to me - conclusions.

- 1972 to 2012: 100% TSM -> 100 becomes 4688
- 1972 to 2012: 100% TBM -> 100 becomes 2055

Equities are fatter better than bonds, nothing new. Now, an extreme timed approach:

- 1972 to 2012: 100% TBM and only 100% TSM when equities dropped 35% until they won 53,8% (= 1/0.65 - the inverse of -35%) relative to that point -> 100 becomes 5910

A few years in the stock market is far better than always in the stock market. Even though 1982 to 1999 was a pure goldmine for stocks, the timed approach still wins. We had 0% equity exposure during 1982 to 1999!

Let's try to explain this. According to Simba's spreadsheet, the CAGR of TBM was 7,65%. So, indeed, a portfolio of 100 after 41 years is:

100 * (1.0765^41) = 2054

This matches with the value above. So far, so good.

During the period from 1972 to 2012, there were 3 crashes of -35% (even more): 1972-1973, 2000-2002 and 2008. The market recovered all three times. This means we had three times the opportunity for an extra 53,8% if we went all in stocks after a drop of 35%:

2054 * (1 / 0.65) * (1 / 0.65) * (1 / 0.65) = 7479

This is not completely correct, since this still includes the gains from the bond market during these recovery periods. This explains the difference between 7479 and 4688 above. This is not easy to calculate, so let's ignore this for now. It doesn't change much to get the point of what I am trying to say.

The CAGR is now: 74.79 ^ (1 / 41) = 1.11 = 11%

According to Simba's spreadsheet, CAGR of 100% TSM was 9.84%. So, indeed, the timed approach wins, given the assumption made earlier. Because of that assumption, please do not look at the exact numbers, but think about the effect of three times 53,8%.


This example of course didn't say much, since no real sane person would go from 100% TBM to 100% TSM. However, it gives you a feeling why this approach might work.

In previous posts, I made a more interesting example: 1972-2012: 70/30 always <=> 0/100 and only 70/30 during recovery periods. In this approach, the risk of the timed portfolio was significantly lower, since equity exposure was much lower. Yet, this performs the same of even better. See http://www.bogleheads.org/forum/viewtop ... 1#p1751641 for results. We had 0% equity exposure during 1982 to 1999, the mother of recent secular bull markets, and still win in 2012! We had 70/30, a standard portfolio, during a small set of recovery periods.

A conclusion from these very basic tests is: a lot of money can be made during a recovery. Actually, it doesn't harm to be over weighed in bonds for most of the time (and therefore losing value because of lack of equities), if you have the guts to migrate to equities during a downturn. This doesn't necessarily mean that you have to go for 100%: 0/100 most of the time and 70/30 during recoveries doesn't yield worse than 70/30 all the time, but it's much less risky! Having guaranteed money to invest during a downturn is golden. Giving up equity exposure in favor of bonds during bull markets to be able to invest with as much as possible guaranteed money during a recovery, isn't a bad idea, surprisingly. Returns are the same or even better, but risk is much lower, due to decrease of equity exposure.

I am happy to run more tests if somebody wants me to. If somebody has links to downloadable/parsable data, I am happy to improve my application. I will try to implement rolling returns somewhere in the next days, so we can do more interesting tests.
Interesting Appel. Couple of avenues for your to explore

1. How sensitive is your model to the 35% drop number? What if you made it 30%? or 20% or 40%?
2. Run the analysis for different time periods. 70-80, 80-90, 90-00 etc.
3. Run the analysis upto to 2007 - see if there is recency bias
DoWahDaddy
Posts: 319
Joined: Sat Jul 02, 2011 9:32 am

Re: Dynamic AA - does it make sense?

Post by DoWahDaddy »

AppelSienSapFrietjes wrote:Hi,

I have been pondering about the following. Say you have a standard AA between stocks and bonds, for example 70/30. Does it, in your opinion, ever make sense to forgo your 70/30, and go to 90/10 (or 80/20 or 100/0 or ...) during a certain period because of bad past results of the equities? If you had 70/30 before 2008, and went to 100/0 (the extreme) after/during the market collapsed, you would be buying low. Years later, you will benefit.

To begin with, I don't think this is market timing, since you are over weighting assets which have done bad in the past.

You need some kind of threshold to decide that you forgo your AA in favor of more equities. A threshold might be that the market has been down for XX% (for example 30%) since it's highs. You can never predict (and you don't want to try - market timing) the lowest point, so you need some kind of threshold. Once the equities have been recovered - again, a threshold is needed here -, you can again convert to 70/30.

Does this make sense? What is your opinion?
It makes sense. It's the definition of rebalancing. If you're 70/30, and stocks drop 20%, but bonds don't move, you are now 65/35. So selling bonds and buying bonds related to bad equity performance to get back to 70/30 is prudent.

That is, if you're responding to bad equity performance. If you're predicting bad equity performance, another story.
"I've worked in the private sector. They expect results." - Dr. Raymond Stantz
User avatar
ObliviousInvestor
Posts: 4212
Joined: Tue Mar 17, 2009 9:32 am
Contact:

Re: Dynamic AA - does it make sense?

Post by ObliviousInvestor »

Appel, you may find this thread to be interesting reading:
http://www.bogleheads.org/forum/viewtopic.php?t=75585
Mike Piper | Roth is a name, not an acronym. If you type ROTH, you're just yelling about retirement accounts.
Chris M
Posts: 88
Joined: Tue May 28, 2013 6:25 pm

Re: Dynamic AA - does it make sense?

Post by Chris M »

Chan_va wrote:
Interesting Appel. Couple of avenues for your to explore

1. How sensitive is your model to the 35% drop number? What if you made it 30%? or 20% or 40%?
2. Run the analysis for different time periods. 70-80, 80-90, 90-00 etc.
3. Run the analysis upto to 2007 - see if there is recency bias
Appel, it makes intuitive sense to me, looking forward to seeing more results. I agree with above suggestions, and have another. The 35% cutoff would have left you all bonds for over 25 years (1975 to 2000), and at least up till 1982 you'd have been earning negative real returns. Inflation is really a bigger risk than volatility. What if instead of 100% bonds you went to some other allocations, say 50/50?

What I'd really like to see is this test extended back in time, but I can't suggest a data source for you...
Topic Author
AppelSienSapFrietjes
Posts: 93
Joined: Thu Jun 20, 2013 5:50 am

Re: Dynamic AA - does it make sense?

Post by AppelSienSapFrietjes »

Hi,

Quick update: I plan to update my application such that the questions above can be answered quite easily. Unfortunately, I won't have time for this in the next week, but I will let you know.

@ObliviousInvestor: thanks for link, very interesting!
User avatar
HomerJ
Posts: 21281
Joined: Fri Jun 06, 2008 12:50 pm

Re: Dynamic AA - does it make sense?

Post by HomerJ »

Almost every single investing theory backtested well at one point.

And then failed when the future didn't match the past.

What variable have YOU forgotten?

You only get one chance at a 30-year plan...

You really think you'd hold all bonds through 1982-2000 and not rethink your plan?
Dandy
Posts: 6701
Joined: Sun Apr 25, 2010 7:42 pm

Re: Dynamic AA - does it make sense?

Post by Dandy »

Market timing can sneak up on you. It often seems like a no brainer e.g. market plunged so let's buy a lot more equities than our risk tolerance calls for. If you had 70% equities and rebalanced yearly you would have done nicely and not taken more risk than, in calmer times, you decided you could handle. To avoid market timing traps the emphasis is on having an IPS with allocations that make sense and then sticking to it. Yes, there are some times you could do better and yes sometimes worse - but you will avoid getting into risk you probably can't handle.
Post Reply