While researching a stock, you learn that eight Wall Street analysts publish opinions on it. Five of them recommend a purchase and three say hold. Is that good?

Maybe. A bit more information will make those recommendations more telling.

Finance researchers have devoted heaps of time over the past decade to judging the worth of analyst advice. That was motivated in part by scandal. The Internet stock bubble of the 1990s raised suspicions that large investment banks had published chipper research on companies to better their chances of selling financial services on the side. In April 2003, about three years after the bubble popped, 10 of America s largest investment firms agreed to pay more than $1.3 billion to settle the matter with the Securities and Exchange Commission.

Here s what studies have turned up so far: Buy recommendations are far more common then sells, especially in the U.S. Sell recommendations seem to hold more predictive power. One study showed that stocks with the most buys outperformed those with the most sells, but later studies found that some hidden factors explain the results better than the recommendations. Analysts tend to favor glamorous stocks -- those with rising share prices and fast-growing earnings -- just like most investors. Stocks that demonstrate that kind of momentum are more likely than not to outperform over the following year or so -- with or without analysts recommendations. That suggests that the level of analysts recommendations isn t that useful as a predictor. (Also, it means we should pay careful attention to the study periods when reviewing analyst performance. Glamour stocks tend to outperform in bull markets, so analysts can be expected to look good then, too.)

One finding stands out as more useful, though. Upgrades tend to be followed by market-beating stock performance. In other words, what matters about the stock in my opening example isn t that it has five buy recommendations, but whether any of them were recently changed from holds or sells. Most likely, that has to do with freshness. Some analysts revisit their recommendations only every six months or so. Standing recommendations might reflect stale opinions. Recent upgrades reflect new thinking.

Of course, investors shouldn t blindly follow analysts without doing their own research. The five stocks below have attracted recommendation upgrades in recent weeks and have modest valuations, good dividends and strong balance sheets.