N March 10, 2000, five years ago today, the Nasdaq composite index hit its high, 5,132.52, at the peak of the dot-com bubble. It is fitting to commemorate this anniversary with a column about what financial economists have learned from this episode.

Perhaps the most fundamental question one can ask about the bubble is how it could have happened in the first place. How could stock prices be pushed up to such irrational and unsustainable levels?

Few economists would deny that fools and gamblers participate in the stock market. But the participation of such irrational traders does not necessarily imply that stock prices themselves should be irrational.

In principle, irrational exuberance should be self-correcting. If overly optimistic investors bid up the price of a stock, rational investors should step in and sell shares, moving the price back down to a realistic level.

This adjustment process does not even require that sellers own shares of the overpriced stock. Someone who thinks that the price of a stock will fall but does not own any shares can borrow shares to sell, a practice known as selling short.

If an investor sells short and the stock price does indeed fall, he can buy shares to pay back the loan and pocket the difference as profit. As with other sorts of borrowing, the borrower typically has to pay interest on the loan.

But, in the case of the Internet boom, short selling was apparently not strong enough to damp the stock price increases during the Internet bubble. The question is, Why not?

Owen A. Lamont, a professor of finance at the Yale School of Management, reviews some recent work in the economics of short selling in the latest issue of the NBER Reporter (available at www.nber.org).

As he points out, there were striking examples of apparent overpricing of stocks in 2000. For example, in March of that year, 3Com sold a fraction of its holding of Palm; it announced that by the end of the year it would disburse the rest of its holdings by giving 3Com shareholders 1.5 shares of Palm for each share of 3Com they owned.

One would expect that 3Com shares would be worth at least 1.5 times the value of Palm shares. But on the first day of trading after the announcement, Palm shares were worth $95.06 a share while 3Com shares fell to $81.81. The market was valuing the non-Palm part of 3Com's business at minus $63.

This pricing anomaly was widely reported in the financial press. The most likely explanation was that day traders and other overly optimistic investors bid up the price of Palm stock to excessive levels. These traders were presumably unaware that they could acquire Palm indirectly by buying 3Com stock.

The apparent mispricing created a low-risk arbitrage opportunity. A savvy investor could buy some 3Com shares outright, borrow some Palm shares, sell them, and repay the borrowed Palm stock in a few months when 3Com issued the Palm shares.

Indeed, many investors did exactly that. At one point, the number of Palm shares borrowed to sell short was 147 percent of the shares outstanding. (The number could exceed 100 percent since shares could be borrowed only to be lent out again.)

Even this additional supply of shares was not enough to quell the Palm enthusiasts. According to Professor Lamont and his co-author on one paper, Richard H. Thaler, a big part of the problem is that the market for borrowing shares is not a centralized market with quoted prices, but rather a highly disaggregated market.

In many cases, it was quite difficult to find shares of Palm that could be sold - and when they could be found, the interest rate charged to borrow them was quite high.

Short selling was not the only way to bet against Palm. One could also buy put options, which allowed the stock to be sold for a fixed price. But the options were also mispriced during this period, making such investments unattractive.

So the anomaly persisted for many months. Eventually, of course, it disappeared: a few weeks before 3Com issued its remaining Palm shares, the prices reflected the appropriate ratio. But the short selling constraints seemingly allowed the mispricing to persist for an awfully long time.

Many intelligent investors believe that short selling is a strong signal of subsequent price declines. Professor Lamont's work and that of several other academics find that this is so: stocks that are subject to large short selling tend to have lower subsequent returns.

One might ask why short selling is not immediately reflected in the price of a stock. The reason appears to be related to the basic problem with short selling: there is no centralized market for borrowing stock, so it can take time to find shares to sell short. Typically, there is no problem in finding shares of large companies that are traded frequently. But shares of small companies, whose stock is lightly traded, may be hard to find.

This finding suggests that the total amount of short selling might be a good predictor of stock market movements in general. Somewhat surprisingly, this turns out not to be true. In fact, during the bubble years, the aggregate amount of short selling declined as stock prices were bid up. As Professor Lamont and his co-author on another paper, Jeremy C. Stein, remark, "Short selling does not play a particularly helpful role in stabilizing the overall stock market."

But again one must ask why not. One suggested answer is that short selling arbitrage has more risk than appears at first glance. For example, the owner of the shares can force the borrower to return them under certain conditions. Hence, the short seller might find his position unwound at an inconvenient time. This risk factor means that short sellers may want to take smaller positions than they would otherwise prefer.

It appears that at least on some occasions, short selling constraints can disrupt the normal operation of supply and demand: when supply is constrained, stock prices end up being determined by those who are overly optimistic.

Hal R. Varian is a professor of business, economics and information management at the University of California, Berkeley.