But there is reason to believe that the bargains will not continue. For the last few years, Internet retailers have engaged in ''penetration pricing,'' setting low prices to make a big splash in new markets. But recently, Internet retailers appear to be nudging up prices, so as to reduce the burn rate and improve the bottom line.
While online retail selling matures, online wholesale markets are heating up as businesses rush to set up exchanges to cut supply costs and, incidentally, to create a more competitive environment for their suppliers. Proponents of online exchanges and auctions claim cost savings of 10 to 30 percent.
What does economics have to say about these rosy scenarios? Will online markets really result in lower prices in the long run? The answer is somewhat ambiguous: there are good reasons why prices should go down, but there are also some good reasons why they might go up.
The Internet lowers the cost of search. It is a lot easier to click on a hyperlink than it is to drive to the store, or for that matter, to make a phone call or send a fax. This means buyers can shop around for a low price. High-priced sellers have either to meet the competition or go out of business.
But there are more subtle effects. It is true that buyers find it easier to discover what the sellers are charging, but each seller also finds it easier to discover what other sellers are charging. It there are a lot of competing sellers, this will probably tend to push down prices even more rapidly. But if there are only a few sellers, low-cost information about what they are charging could, perversely, lead to higher prices.
Why? Because when price information is widely available, it becomes easier for suppliers to coordinate their pricing, with each seller agreeing, implicitly or explicitly, to charge a high price so as to avoid ''ruinous competition.'' A classic example is the infamous Joint Executive Committee set up by the major American railroads in the 1880's. The committee, actually a cartel, predated the Sherman Antitrust Act of 1890 and was, at the time, perfectly legal. Newspapers of the day reported on meetings and statements from the committee, just as newspapers today report on meetings of OPEC.
The primary role of the committee was to collect information and publish weekly accounts of shipments by the individual railroads. The committee also collected data on prices charged, but this was notoriously unreliable since committee members tended to cheat on the cartel prices by making secret agreements with shippers to shave prices in exchange for traffic.
It is important to recognize that price cutting by the railroads had to be secret -- public cuts were counterproductive because they just precipitated price wars. But imagine what would happen if the railroad cartel had used a public Internet exchange. Each railroad could then see the prices charged by the others and attempts to shave prices could be quickly countered, making the cartel less vulnerable to price wars.
Sound far-fetched? Look at the way airline prices are posted online today: each airline announces its prices, along with the associated terms and conditions, allowing competitors to scrutinize one another's offerings. In the late 1980's, the airlines used online reservation systems to signal price changes to one another. One airline would post a price increase at 2 a.m. on Monday. If the competition followed, the change would stick. If the competition did not respond, the airline would roll its price back by 6 a.m. In 1992, the Justice Department sued several large airlines, requiring them to cease such practices.
Online merchants are clearly monitoring one another's prices today. Go to www .dealtime.com and type in the name of a well-known book. Odds are the the best prices are within a few cents of one another. This is not just chance.
The usual story of competition has an implicit timing assumption that buyers move faster than sellers. If a seller cuts its price, it is rewarded by a flood of buyers before the other sellers can respond. But if the sellers move faster than the buyers, they can match one another's price cuts immediately, damping the rewards of price cutting. Similarly, if a seller raises a price and all other sellers quickly match it, the seller will lose few, if any, customers. When sellers move faster than buyers, prices tend to drift upward.
If there are a lot of competitors, this effect is swamped by the usual dynamics of competition. The problem is that the cost structure of online retailing suggests that there are likely to be only a few big companies in the long run, making implicit collusion more likely. When Toys ''R'' Us and Amazon.com announced a joint venture a few weeks ago, Toby Lenk, chief executive of eToys, said: ''This is great news for us. Last year we had half a dozen competitors. Now our two remaining competitors are merging into one.''
If there are only a few online retail vendors, the real competition they face will not be from one another, but will be from offline substitutes -- traditional retailers. As long as this market remains fragmented, full-fledged retail collusion is unlikely. Prices will not be as low as they were in the heady days of penetration pricing, but they will have to be attractive enough to compete with alternative retail channels.
What about the online business-to-business exchanges? If an exchange operates in a wholesale market where there are few buyers and many sellers, the extra flow of price information is likely to work for the benefit of the buyers, pushing down prices. But in exchanges in which there are few sellers and many buyers, the availability of timely price information may well have the opposite effect, pushing up prices.
The Federal Trade Commission is now trying to set standards of behavior for online exchanges. The objective is to make sure that such exchanges foster competition rather than dampen it.