Wednesday, April 2, 2008

Buy Low, Sell High Market Timing Strategies

A reader, Jay, left comments on my market timing experiment asking specifically about strategies that buy after a market drop, and sell after the market rises. The theory behind these strategies is to buy low and sell high.

I ran some experiments to test this approach. As in previous experiments, the investor decides each month whether to be invested in the S&P 500 or not. The goal is to avoid months where stocks drop in value. I ran the experiments on S&P return data from December 1990 to March 2008.

The first thing I tried was to assume that the investor would be invested in the stock market at the beginning and would proceed as follows:

1. If the money is in the stock market, and if stocks are priced at least 5% higher than they were one, two, or three months ago, then sell. Otherwise, stay in the market.
2. If the money is in cash, and if stocks are priced at least 5% lower than they were one, two, or three months ago, then buy stocks. Otherwise, stay out of the market.


The market timer’s compound annual return was only 3.2% compared to 11% for a buy and hold investor. How could the result be so bad? What happened was that the market timer jumped out of the market after prices rose and then waited for a decline that took a long time to come. In the mean time, the market timer was out of the market while prices rose.

Instead of just using 5% thresholds, I tried varying them over all combinations from 1% to 15%. That’s a total of 15*15 = 225 strategies. In every single case the market timer did worse than a buy and hold investor. The only time it was close was when the thresholds were set so that the market timer stayed in the market almost all the time.

As we saw in the version of my original experiment that took into account taxes, the market timer’s results compared to buy and hold are even worse if the money is invested in an account that isn’t tax deferred.

We could keep trying variations on this approach, but we’d run risk of creating a strategy that is tuned to old data but won’t work in the future like what happened in this April Fool’s joke.

All the market timing strategies I’ve examined fail in basically the same way. They stay out of the market too much of the time. The market rises most of the time and the odds are that the market will be up on average during whichever periods of time you choose to sit on the sidelines.


  1. Michael: All I can say is WOW ... Thinking about it, it actually makes a lot of sense. This kind of strategy would make someone to sit on the sidelines for too long, and we all know there are a lot more up days than down days. Hence one would miss a lot of up days.

    Again, thank you for your detailed work and for a very enlightening post. I look forward to reading your next post.

  2. I have one more variation that I was wondering if you could test out for me. It's where you start with nothing, and regularly save some amount per month (let's say $500) and then keep it in cash until there is a down or flat day (or week or month, if that's the sampling you have the market at) to put it in the market. Once it's in the market, it stays there forever.

    My hypothesis is that this strategy might be slightly better than buying immediately on the first of the month if one is able to use the noise of fast sampling (i.e.: waiting for a down day or week shouldn't take very long, but a down month might take too long and one would miss out on a lot of gains). With longer time periods (waiting for a down quarter) I guess this would probably lag the buy immediately strategy. I suspect that market timing may still be a bad strategy, but at least this way you're only guessing once, and might be able to take advantage of some of the noise.

    Looking at the past 6 months by hand, this doesn't look like a bad strategy... but then again, each of the last 6 months had a down day in one of the first two days. The strategy might not be good with sampling at even a week, and if I picked starting at the last day of the month, instead the 1st day, there's at least 2 months where this would have been worse than buying immediately at the last day's close (though waiting for a down day would be very slightly better than buying immediately on the last day overall).

  3. Potato: The data I have are monthly. I ran a two simulations where $1000 is invested each month. In one simulation, the money is always invested immediately. In the other, the money stays in cash (generating 3%interest) until the return over the previous month is negative or zero. The results were very close. The "immediate" investor ended up with $442,000, and the market timer ended up with $437,300.

  4. Thanks! I think that should put market timing to rest for me... at least until I get that time machine figured out.

  5. I think all investment strategy tests are dependant on the start/end dates used. In this test, for example, what would happen if we used 1990-2009 (after the market crashed)? Buy-and-hold investors may not have better results than those who got out before the crash and got back in after the crash...

  6. @Alex: Everyone dreams of getting out before a crash and back in at the bottom, but almost nobody is able to do this.