Here's some pretty standard advice: Stocks are long-term investments, so you should invest with at least a five-year horizon, ideally longer. What happens to investors who sell before then? A recent study by financial services research firm Dalbar, Inc. offers one answer, and it's not pretty.

By studying the cash that flowed into and out of mutual funds from 1984 to 2002, Dalbar concluded that, on average, investors held on to a stock fund just 29.5 months -- less than three years. Generally, money flowed into stock funds during good times, and flowed out during the bad. In other words, investors adhered to a "buy high, sell low" strategy.

This is not a recipe for success. While the Standard & Poor's 500 gained an average of 12.22% a year during the period, investors earned just 2.57%. And with inflation averaging 3.14%, in terms of actual buying power, investors lost ground. Bond fund investors did a little better, staying in a fund for 34.3 months and earning 4.24% annually. However, this still lagged the long-term government bond's average annual return of 11.70%.

As pointed out in a Wall Street Journal article yesterday, some experts take issue with the Dalbar study because it analyzes money flows and not the returns of actual funds and investors. But all agree on one overarching conclusion: Investors chase returns, and do so to their great misfortune.

Do your net worth a favor: Don't jump in and out of funds, and don't pick a fund, or stock, based on recent returns. Studies have shown that one year's top-performers are often the next year's dogs. For just one example, take the Janus Global Technology fund, which returned 212% in 1999 but plunged 84% from March 2000 to October 2002, according to Kiplinger's magazine.

If you're looking for a fund that you can live with for a decade or three, consider an index fund that mirrors the performance of the stock market. Any investor who chose to do that in 1984 would be sitting rather pretty right now.

For help with choosing an index fund, visit our 60-Second Guide.