Rebalancing strategies

[cathyS]

I read Harry Browne's "Fail Safe Investing", and it struck me that while the asset combination was rationally chosen and clearly very effective, the allocation and rebalancing method were based only on simplicity.  There's nothing wrong with that, but it got me wondering:  what IS the best rebalancing strategy, or does it really matter?

I did a few simple Matlab simulations using S&P 500 prices since 1950.  I quickly realized that long term simulations would be very misleading.  Market behavior has changed greatly over time, and the results will usually be dominated by one era.  Not to mention that the stock market from 1950-2012 will never again be duplicated; what I want to know is how different strategies work for different types of market behavior.

So I tested in 3 eras separately:  slow steady growth (1950s-1960s), rapid but steady growth (1980s-1990s), and wild fluctuations with little overall growth (2000-present).  For purposes of comparing strategies, I used a very simplified model:  portfolio of cash (assumed to grow at a steady rate of 2%/year) and SPY.

The strategies were:  1) simple rebalancing to maintain a 50/50 portfolio with 10% bands, 2) rebalancing with overcompensating (i.e. go from 60% stocks to 40% stocks, rather than 50%, and 3) a "momentum"-based strategy where the portfolio is either 100% stocks or 100% cash depending on recent average performance.

The momentum strategy is the exact opposite of simple rebalancing, so it's likely that one will work better than the other consistently.  If the stock market followed a weighted "random walk" pattern, it shouldn't work.  However, stock price changes most definitely are not random when you look at monthly data.  I verified this in each era separately (stats would not be valid for the entire data set at once - data are non-gaussian and non-stationary).  The best predictor turns out to be 1 month performance.  Interestingly, correlation goes negative when you go back 24-36 months.

I can post data if anyone is interested, but here's the bottom line:  Strategy 2 was the worst performer in all eras, and Strategy 3 was the best performer in all eras.  In the 1950s and 1980s-1990s, Strategy 3 was dramatically better; in the 2000s, Strategy 3 was better than Strategy 1 even if you use a 60/40 split rather that 100/0.  For #3, rebalancing events occurred just under once a year through the 1990s, then about once every 2-3 months in the 2000s.

I want to rerun this scenario using gold instead of cash (I can't find treasury price data in usable form), but in the meantime it looks to me that momentum-based rebalancing might indeed be the way to go if you are willing to put in the time, and if trading costs are low relative to portfolio size.

[Xan]

Why 50/50 stocks and cash?  What about the rest of the portfolio?

[bswift]

How does your strategy perform in the future?

[MediumTex]

cathyS,

What are your goals in using the PP?

Taking something that works as promised (i.e., stable inflation adjusted-returns in all market environments) and turning it into a momentum-based trading strategy is utterly inconsistent with the theory behind the PP (i.e., being prepared for any market conditions at all times).

Even if a bulletproof vest made a lovely headwrap or sarong, I think you would still want to be cautious about modifying its application in this way.

[ngcpa]

I can't find treasury price data in usable form

I use price & dividend data from VUSTX in my simulation program.  It has a long history (started 5/19/86) and is close enough (average maturity 21.4 years and average duration 15.0 years).    I obtain gold price data from http://www.onlygold.com/TutorialPages/PricesY2KFS.asp which goes back to 1973.  The only problem is I had to key in a lot of data (I use qtr closing prices and dividend reinvesting prices as the close of the day before the dividend is paid).
Norm

[alvinroast]

Even if a bulletproof vest made a lovely headwrap or sarong, I think you would still want to be cautious about modifying its application in this way.

I love this. You may have outdone yourself here. I really hope that the book includes a generous dose of MT analogies.

[cathyS]

I'll remember the bulletproof vest next time.

I am trying to understand the rationale behind a key piece of the PP, which hasn't really been closely examined and seems to be accepted as "proof by eminent authority".  It is interesting that two completely OPPOSITE strategies both seem to work.  My number crunching suggests that once you get past day to day noise, markets move slowly - which makes perfect sense when you think about it.  This inertia has been well documented, too, and looks to be just as much a property of the economy as inflation, deflation etc.  Simple rebalancing during an up cycle, then, may actually be like shooting yourself in the foot.  With that in mind, I might well turn bswlft's question around and ask, how do you know that simple rebalancing will work in the future?  The answer is:  likely all three strategies will work okay - but I'm impressed that the strategy that sounds riskiest actually had slightly lower volatility than the standard approach, AND higher returns, in 3 very different market eras.

At the time that the PP was being conceived, there weren't many tools accessible to the average small-fry investor so simple rebalancing made the most sense.  In 1986, a computer with far less power than the laptop you're reading this on took up an entire room and could only exist at major universities.  And there was no such thing as Excel.  Now that we do have these kinds of resources, I'd say it is time to take a fresh look.

And no, I'm not sold on the idea yet.  Yes, this was an extremely simple model, because I wanted to understand how rebalancing works, not road test the whole portfolio (yet).  Thanks for the tip about VUSTX.  Would like to know if anyone else has done these kinds of simulations.  I'd be happy to email matlab code on request.

[MediumTex]

I am trying to understand the rationale behind a key piece of the PP, which hasn't really been closely examined and seems to be accepted as "proof by eminent authority".  It is interesting that two completely OPPOSITE strategies both seem to work.  My number crunching suggests that once you get past day to day noise, markets move slowly - which makes perfect sense when you think about it.  This inertia has been well documented, too, and looks to be just as much a property of the economy as inflation, deflation etc.  Simple rebalancing during an up cycle, then, may actually be like shooting yourself in the foot.  With that in mind, I might well turn bswlft's question around and ask, how do you know that simple rebalancing will work in the future?  The answer is:  likely all three strategies will work okay - but I'm impressed that the strategy that sounds riskiest actually had slightly lower volatility than the standard approach, AND higher returns, in 3 very different market eras.

At the time that the PP was being conceived, there weren't many tools accessible to the average small-fry investor so simple rebalancing made the most sense.  In 1986, a computer with far less power than the laptop you're reading this on took up an entire room and could only exist at major universities.  And there was no such thing as Excel.   Now that we do have these kinds of resources, I'd say it is time to take a fresh look.

And no, I'm not sold on the idea yet.  Yes, this was an extremely simple model, because I wanted to understand how rebalancing works, not road test the whole portfolio (yet).  Thanks for the tip about VUSTX.  Would like to know if anyone else has done these kinds of simulations.  I'd be happy to email matlab code on request.

I think that most people who use the PP are skeptical of backtested tweaks that have the benefit of hindsight. 

The PP is fundamentally not a strategy that uses past results to predict future returns; rather, the approach is based upon an economic theory that the economy has a limited number of configurations and if you can position yourself to have stable returns in all possible configurations you will be protected no matter what the future brings.

Against this theoretical backdrop the PP uses backtesting to confirm that past results do not disprove the theory, but past results are not used as an indicator of what might happen in the future.

If a PP-based momentum strategy is what makes the most sense to you, then I would say go for it and enjoy the excitement.  Other people are looking for something a bit more dull and plodding, and that's what the PP really is.

From my own experience, virtually any momentum-based strategy will eventually just burn you out, and you will be on to the next idea.  To me, the PP is sort of the place you go when you get tired of all of that.

[cowboyhat]

Daniel Kahneman's (Nobel Prize in Economics) new book "Thinking, Fast and Slow" is worth your time. Lots of information for investors, but written in a lively and accessible style.

http://www.amazon.com/Thinking-Fast-Slo ... 0374275637

There is an interesting section in the book where Kahneman discusses the officer screening system he developed as a young man for the Israeli Army. He uses this story to develop some ideas about what types of algorithmic systems are most robust for problem solving in complex environments.

When I first read Harry Browne's justification for his asset selection in "Fail Safe Investing" it struck me as simplistic and unlikely. I felt more comfortable with the traditional and more complex justification in William Bernstein's essay on the PP:

http://www.efficientfrontier.com/ef/0adhoc/harry.htm

Kahneman has made me think again about Harry Browne's reasoning. Yet another humbling lesson.

[clacy]

CathyS,

I employ momentum in my holdings.  I use a standard PP for about half of my assets and rebalance a little more frequently than the 40% bands.

For the other half, I use 2 momentum strategies. 

My theory is that momentum does exist, but I find it difficult to put all my eggs solely in the momentum basket. 

Since I use momentum for half of my money and mean reversion (buy low, sell high rebalancing) for the other half, I feel that one of these two methods will always be out-performing the other, but both will have a real growth over time.

[cathyS]

I employ momentum in my holdings.  I use a standard PP for about half of my assets and rebalance a little more frequently than the 40% bands.

Interesting, what have you found so far?

For the record, everyone....I'm not yet sold on the idea.  I'm currently sticking with the standard approach.  The results of the simple test I did surprised me (beat standard rebalance by 14% under the worst possible conditions, without increasing volatility), but when you think about it....whether the market will go up or not next month has to be a weighted coin flip.  If it were truly random, we'd all just stuff our money under the mattress and call it a day, because over time the markets wouldn't go anywhere.  If I promised you a dollar every time a 60/40 weighted coin came up heads, if you agreed to pay me a dollar every time it came up tails, I'm pretty sure you'd take the deal.  The question really is, whether the actual added returns are worth the additional work and transaction costs - which I don't know yet.

There was a great mathematical puzzle that comes to mind.  Remember the "Let's Make a Deal" game show?  It goes like this:  You're shown 3 closed doors, and told that behind one of them is a Cadillac, and behind the other two are a can of tunafish.  You pick a door, then are shown another door with a tuna can behind it.  You are then asked if you want to keep your original choice or switch.  If you're a mathematician, you would always switch because it doubles your chances of winning compared to sticking with your original choice.  However, most people did not switch because of the psychology involved - and the actual show relied on this fact in order to stay profitable.

Last comment - sticking with a tried and true system with solid theoretical underpinnings I can understand, but realize the first part of that is no different from back-testing.  I love the idea of the PP because of the theory behind the asset selection, which is a beautiful example of simple, inductive logic winning the day.  I like the momentum concept for the exact same reason.

[AdamA]

...it looks to me that momentum-based rebalancing might indeed be the way to go if you are willing to put in the time, and if trading costs are low relative to portfolio size.

Using such a strategy with the PP assets, I think you'd be 100% in 30 year treasuries right now. 

Or, you'd be 50% 30 year treasuries and 50% in gold (if you used the top 2 assets for the past year).

Would you be able to pull the trigger on that?

[cowboyhat]

Momentum investing probably works, but it takes real effort and may make you more anxious about your money than is healthy. It is also a lot more expense than passive investing when you take taxes and trading costs into account.

Might be worth trying a small experiment before plunging into something with lots of your money.

You could set up two accounts with $10,000 in each. Invest one in a classical EFT-based PP (TLT, GLD, SPY, SHY). Invest the other in the Decision Moose. Last time I looked (two years ago?) Decision Moose cost something like $50 per year for the trading signal plus you got a bunch of reasonably interesting market commentary. With Decision Moose you buy and sell once a week, so it's not too much to keep up with if you have a day job.

For a fair comparison you would need to keep track of how much time you spend on your Decision Moose account as well as all of your expenses. At the end of the year compare the net after tax profits between the accounts and figure out your net gain on Decision Moose compared to the PP. If you divide your total invest-able assets by 10,000 and multiply that number times your net DM profits divided by the number of hours you spent on DM you will have pretty good estimate of how much you are capable of making per hour momentum investing using system with a reasonable track record.

My potential excess profits came out to less than minimum wage. Plus I was emotionally invested in what happened in the markets.

[MediumTex]

My potential excess profits came out to less than minimum wage. Plus I was emotionally invested in what happened in the markets.

Reality.

[Reub]

Decision Moose hasn't made much money over the past year so be cautious about employing this strategy.

[clacy]

cathy,

To your question, I had a very good year last year, but keep in mind for the momentum portion, I use half (so 25% of my total portfolio) on the top 2 performing asset classes.  I look at this monthly, but switches aren't as frequent as you might expect.

The other half (25% of my total portfolio) I use a very wide range of asset classes, including many foreign developed and emerging country funds, as well as a wider range of hard asset classes.  

**Edit to say that this year I'm trailing the S&P fairly substantially, but it seems to ebb and flow with momentum.  So for 6 months to a year, it may enter into a DD or stay relatively flat, often lagging the S&P, followed by a handful of very powerful months.

That last assumption was based on backtesting, as I've only been using momentum for under two years.

[MachineGhost]

I read Harry Browne's "Fail Safe Investing", and it struck me that while the asset combination was rationally chosen and clearly very effective, the allocation and rebalancing method were based only on simplicity.  There's nothing wrong with that, but it got me wondering:  what IS the best rebalancing strategy, or does it really matter?

The interplay between money management, position sizing, rebalancing, momentum and trend following are all distinct and very, very tricky.  There is still not available easy to use software to allow objective comparisons of all such variables.  What is needed is something like the "Trading Recipes" software of old.

I have found that trend following (both technical and fundamental) for each of the non-cash three asset classes gives the best risk/reward ratio.  As a general rule, momentum increases returns and risk.  Trend following reduces returns and risk.  Combing both together is the ideal solution.  But that is not a risk I'm willing to take since I find the default HBPP to be way too risky and you may lose "Black Swan" protection.  But, if you're comfortable with 25% or so intrayear maximum drawdowns, there's certainly room for improvement on the reward side.

MG

[murphy_p_t]

"...since I find the default HBPP to be way too risky and you may lose "Black Swan" protection...."

what have you found to be less risky?