Between 1904 and 1908, more than 240 companies entered the nascent automobile business. In 1910 there was a shakeout, and many of those companies went under. The railroads exhibited the same behavior: in the 1880's there were more miles of railroad track laid than in any other decade in American history. By the 1890's there were more miles in bankruptcy than in any other decade.

What causes these technology-driven investment booms and busts? Why do we always seem to overinvest in new technologies?

In the case of railroads and automobiles, the major economic force at work was economies of scale. The primary costs associated with a railroad are the fixed costs -- the costs of servicing the debt incurred in laying the track and buying the rolling stock. In the 1880's, about two-thirds of the total costs of operating a railroad were fixed.

When fixed costs are high, large companies have an inherent advantage, since they have a lower total cost per shipment. The railroads recognized this and invested heavily in building capacity.

But once the capacity was installed, there was inevitable cutthroat competition for freight. There was no way around the fact that there was just too much rail stock relative to demand. Companies went bankrupt, wiping out their obligation to make debt payments, leading to even more aggressive pricing. The industry sank into a slump from which it took decades to recover.

The automobile story involved a different kind of competitive dynamics. The entrepreneurs who entered the industry in the 1900's were mostly small operators -- carriage makers, bicycle makers and the like -- who were enthusiastic about the new technology of the horseless carriage. Some operators specialized in parts supply and some in assembly, but their fortunes were intimately intertwined since so many supplies were custom made.

One of these small assemblers was Henry Ford, who introduced the Model T in 1908. Ford experimented relentlessly with mass-production techniques. He built big and reaped the cost advantages from economies of scale and vertical integration. The other, much smaller manufacturers just couldn't match his price and quality, and many went out of business. By 1923, Ford was producing nearly half the automobiles sold in America.

But even Ford got his comeuppance. He had built his company on the idea of a single, low-cost product, the Model T. In the 1920's, Alfred P. Sloan, president of General Motors, recognized that the public had tired of the same product's being offered year after year. G.M. pioneered the idea of the yearly model change, an innovation Ford had a difficult time matching and that significantly reduced his dominance of the industry.

What is to be learned from these episodes? The railroad boom and bust arose because there were large fixed costs and a commodity product -- freight transportation. Since there were many providers, price wars were all too likely. The cost structure of this industry bears a remarkable resemblance to long-haul fiber optics. The big fixed cost comes from laying the fiber, with operating expenses being relatively small. This is a recipe for price wars, and, indeed, the cost of long-distance telephony -- especially in negotiated contracts for large businesses -- has been plummeting.

Automobiles, on the other hand, were a highly differentiated product.

Ford managed to make economies of scale work in his favor by building big, enforcing high quality standards and forcing constant innovation in the manufacturing process. A modern parallel would be Michael Dell, founder of Dell Computer, who recognized early on the potential of integrating online ordering and supply-chain management. Now, as the computer industry has become commoditized, consumer tastes are shifting to networked appliances: special-purpose, networked-enabled devices that are highly differentiated.

This is much like the switch the auto industry made in the 1920's. The network appliance industry will soon look much more like the automobile industry or the consumer appliance industry, with multiple companies, extensive product-line competition, new models every year and feature proliferation.

In addition to the classic supply-side economies of scale that drove the industrial evolution in railroads and automobiles, new-economy companies often exhibit demand-side economies of scale. These occur when a user's demand for a product rises along with the number of users. A large part of the value of Microsoft Word, for example, comes from the fact that it is the industry standard.

If someone is thinking of buying a word processor, this makes it a particularly attractive choice.

EBay is another good example. Buyers want to be where there are a lot of sellers, and sellers want to be where there are a lot of buyers. The biggest online auction automatically has an advantage over smaller ones, not necessarily because of cost or quality, but because any given person wants to use the same service that others use.

Last year, hundreds of start-ups wrote business plans saying they were going to be the ''eBay of B-2-B commerce.'' Most of these are rapidly disappearing because they just couldn't create a thick enough market to ensure survival.

Economists have pondered the technology boom-and-bust cycle for years. Joseph Schumpeter called it ''creative destruction'' and thought of it as one of the primary drivers of capitalism.

Overinvestment may seem wasteful, but then it's always easy to identify winners once the race is over. Who would have thought that a business model built on aggregating garage sales and flea markets would turn out to be successful? Yet eBay is one of the few profitable dot-coms. There's no substitute for experimentation -- with business models as well as with technology. One of the great strengths of American-style capitalism is its ability to finance crazy ideas -- because every now and then, those ideas have a very, very big payoff.