## Thursday, March 27, 2008

### A Market Timing Experiment

All available evidence and logic tell us that the vast majority of investors can’t beat the market by market timing. This applies to professional money managers as well. I tried a little experiment to see how accurate a market timer’s predictions would have to be to succeed at beating the market.

Market timing refers to the practice of jumping in and out of the stock market in an attempt to avoid market declines.

The Experiment

Suppose that a market timer decides at the beginning of each month whether to have all of his money in the S&P 500 stocks or all of it in cash. His goal is to avoid being in stocks during the months where stocks perform worse than cash.

Each month the market timer has a certain probability of guessing right. If he just tosses a coin, this probability is 50%. The question is how high this probability has to be for the market timer to beat the strategy of just buying and holding through thick and thin.

I gathered data on the S&P 500 from December 1990 to March 2008 and ran some simulations. For each fixed probability of the market timer guessing right each month, I ran 10 million Monte Carlo simulations and averaged out the results.

To account for commissions, costs due to stock spreads, and interest on cash, I assumed that the net return of interest minus costs averaged 3% per year while the market timer has his money out of stocks.

The Results

A buy and hold investor over this period of time would have received an average compound return of 11.0% per year. From the results chart, we see that the market timer has to be right 60% of the time just to break even with the investor who buys and holds.

If the market timer is right only 50% of the time, he will underperform the buy and hold strategy by 4.2% per year, on average. This is a huge penalty for just guessing.

Over the full period, an initial investment of \$100,000 grew to \$606,000 by buying and holding. A market timer who tosses a coin each month will have a median outcome of only \$312,000! This shows that our market timer missed many months where stocks rose significantly.

It seems to be human nature to be tempted to believe that we can do better than buying and holding by anticipating market declines. The next time you are tempted to try market timing, remember that you have to be right 60% of the time (and the people you trade against have to be wrong 60% of the time) just for you to break even. Do you really believe that you are that much better than all the other market timers?

1. Michael: I'm wondering if you've considered taxes in your experiments. I'm guessing that the market timer would have to pay more in taxes considering that interest income is fully taxed and capital gains are owed when the total gains becomes positive. I'm guessing taxes will put the market timer into a bigger hole.

2. CC: You're abosulutely right. My analysis is based on the assumption that the investment is made in a tax-sheltered account. For many people, this is the only way they will ever invest. I'll see about following up next week with what happens when taxes have to be paid.

3. Micahel, great post and very interesting results. It is one thing to read that market timing doesn't work another to see it in numbers based on probability and simulation models.

My only issue is that market timers don't really toss a coin when deciding, so it is not completely random. One strategy that I know some people follow is to cash out after a significant run (say 5% from a recent low) in a certain period (could be week/month/quarter) and get back in after a significant fall (say 5% from a recent low).

I don't believe in market timing strategies either, but what are your thoughts on this and how much does it affect the simulation you ran?

Thanks

4. Jay: I agree that market timers base their decisions on some sort of information rather than just making random choices. However, my conclusion that you have to be right 60% of the time (and trade with someone who is wrong 60% of the time) just to break even with buy-and-hold is still true.

I'll take another look at this issue next week. I could try something along the lines of what you describe. This can be a dangerous practice, though, because I may end up finding a strategy that would have worked out well over the last 17 years, but will be ineffective in the future. Sometimes with back-tested strategies, all you are doing is fitting a strategy to the particular gyrations of the stock market over the time period you are studying.

5. From Peter Lynch:

People spend all this time trying to figure out "What time of the year should I make an investment? When should I invest?" And it's such a waste of time. It's so futile. I did a great study, it's an amazing exercise. In the 30 years, 1965 to 1995, if you had invested a thousand dollars, you had incredible good luck, you invested a the low of the year, you picked the low day of the year, you put your thousand dollars in, your return would have been 11.7 compounded. Now some poor unlucky soul, the Jackie Gleason of the world, put in the high of the year. He or she picked the high of the year, put their thousand dollars in at the peak every single time, miserable record, 30 years in a row, picked the high of the year. Their return was 10.6 That's the only difference between the high of the year and the low of the year. Some other person put in the first day of the year, their return was 11.0. I mean the odds of that are very little, but people spend an unbelievable amount of mental energy trying to pick what the market's going to do, what time of the year to buy it. It's just not worth it.

Full interview here: http://www.pbs.org/wgbh/pages/frontline/shows/betting/pros/lynch.html

6. Preet: I agree with Peter Lynch about not obsessing about what time of year to make investments. His experiment is different from mine, though. Peter supposed that you were going to buy and hold for 30 years, and he concluded that it doesn't matter very much what time of year you buy before your 30-year hold. I'm looking at market timing where the "investor" jumps in and out of the market. For my experiment, timing makes a huge (negative) difference. Our conclusions are about the same, though: market timing isn't worth it.

7. Correction:
"and get back in after a significant fall (say 5% from a recent HIGH)".

Michael: are you saying that no matter what the strategy is, one would still have to be right 60% of the time to be able to match the buy-hold returns?

Thanks again for another enlightening post. Your writing is so clear and your blog is one of the best I have ever read about investing.

8. Jay: Under normal market conditions, yes, a market timer investing in a tax-sheltered account making a decision once a month about whether or not to be in the market would have to guess right 60% of the time to keep pace with a buy-and-hold investor.

It is possible for there to be extreme market conditions that affect this, though. Suppose that there is one month with a -99% return. In this case, your overall performance would depend on little other than whether you were invested during this month or not. In general, you can be right less than 60% of the time if you happen to be right in the most extreme months. But, I see no reason to believe that anyone can reliably predict which months will be extreme.

9. Michael: Isn't the 60% threshold dependent on the market timer randomly exiting/entering the market every month?

It just seems to me that if one employs a different strategy (like buy only after a drop and sell only after a rise), then they may not have to be right 60% of the time. Possibly less.

But then again, not all truths are intuitive in Math.

I hope you get the time to re-run your experiment and will be check often to see the results.

Thanks

10. Jay:

I am working on something related to the strategy you describe. It should be ready next week.

11. Hi Michael ..

I been to one of your presentation ... it i interesting ... my question is which tools can we use to set a simulation or enable a trade ...
cheers
simon

12. Simon: I'm not sure I fully understand your question. When I perform investing simulations, I develop my own tools. I have heard of web sites that let you set up pretend accounts where you can make practice trades and have your results tracked. However, they are usually geared to option trading or day trading.