NBER Reporter: Spring 2001
Conference on E-Commerce
The NBER's first Conference on E-Commerce, organized by Research Associates Severin Borenstein, University of California, Berkeley, and Garth Saloner, Stanford University, was held in Bodega Bay, California on January 26 and 27. The following papers were discussed:
Carlton and Chevalier examine manufacturers' decisions about whether and how to offer their products for sale over the Internet, focusing on three types of products: fragrances, DVD players, and side-by-side refrigerators. Their evidence suggests that manufacturers who limit distribution in the physical world also use various mechanisms to limit distribution online. In particular, these manufacturers attempt to prevent the sale of their products by online retailers who sell at deep discounts. Furthermore, the authors show that manufacturers who distribute their goods directly through manufacturer web sites tend to charge very high prices for the products.
Gertner and Stillman study how firms in the apparel industry have adapted to the Internet. They choose apparel because it is one of the leading categories of online sales and because various types of organizational form exist within the industry. For example, firms like The Gap are vertically integrated, but many other brands are owned by vendors and sold mainly through department stores and other third party retailers. Using data from 30 firms, the authors find that vertically integrated specialty retailers tended to start online sales sooner, regardless of whether the firm had pre-existing catalog operations. They also find that products of vertically integrated specialty retailers and catalog companies are more available online than the products of nonintegrated vendors.
Kaplan and Garicano study the changes in transaction costs related to the introduction of the Internet in transactions between firms (that is, business-to-business [B2B] e-commerce). They argue that B2B e-commerce likely reduces coordination costs and increases efficiency. The authors classify these efficiencies into three broad categories: process improvements, marketplace benefits, and indirect improvements. At the same time, B2B e-commerce affects incentive costs and possibly informational asymmetries. Using detailed data from one Internet-based firm, and less detailed data for one other firm, they find that process improvements and marketplace benefits are potentially large. There is little evidence that informational asymmetries are more important in the electronic marketplace than the existing physical ones, though.
One of the earliest and best-known Internet reputation systems is run by eBay, which gathers comments from buyers and sellers after each transaction. Resnick and Zeckhauser examine a large dataset from 1999 that reveals several interesting features of this system. For example, feedback was provided more than half the time, and it was almost always positive. Reputation profiles predicted future performance, but the net feedback scores that eBay displayed encouraged overly positive assessments of reputations and were far from the best predictor available. Further, although sellers with better reputations were more likely to sell their items, they enjoyed no boost in price, at least for the two sets of items that the authors examined. Finally, there was a high correlation between buyer and seller feedback, suggesting that the players both reciprocate and retaliate.
Sellers in eBay auctions have the opportunity to choose both a public minimum bid amount and a secret reserve price. Katkar and Lucking-Reilly ask whether the seller is made better or worse off by setting a secret reserve above a low minimum bid, as opposed to making the reserve public by using it as the minimum bid. In a field experiment, they auction 50 matched pairs of Pokémon cards on eBay, half with secret reserves and half with equally high public minimum bids. The authors find that secret reserve prices make sellers worse off by reducing the probability of the auction resulting in a sale, deterring serious bidders from entering the auction, and lowering the expected transaction price of the auction. They also show that some sellers choose to use secret reserve prices for reasons other than increasing their expected auction prices.
Internet shopbots allow consumers almost effortlessly to compare prices and service levels of dozens of competing retailers. Brynjolfsson and Smith analyze the choices of 20,227 shopbot consumers who choose among 33 competing retailer offers for books over a sample period of 69 days. They find that consumers are remarkably sensitive to how the total price for goods is allocated among the item price, the shipping cost, and tax; consumers are also quite sensitive to the ranking of retailer offerings with respect to price. Even in this setting, brand is important; in particular, consumers appear to use brand as a proxy for a retailer's credibility in terms of noncontractible aspects of the product, such as shipping time.
Zettelmeyer, Morton, and Silva-Risso investigate the effect of Internet car referral services on dealer pricing of automobiles in California. Using data from J. D. Power and Associates and Autobytel.com, a major online auto referral service, they compare online transaction prices to regular "street" prices. They find that the average customer of this online service pays approximately 2 percent less for a car, which corresponds to about $450 for the average car. Fifteen percent of the savings is attributable to making the purchase at a low-price dealership affiliated with the web service. The remaining 85 percent of the savings seem to be a result of the bargaining power of the referral service and the lower cost of serving an online consumer. Consumers who indicate that they are ready to buy within two days of going online pay even lower prices. Consumers who use the web do better than at least 61 percent of offline consumers, the authors conclude.
Using data collected between August 1999 and January 2000 covering 399 books, including New York Times bestsellers, Clay, Krishnan, and Wolff examine pricing by 32 online bookstores. Over the sample period, they find no change in either price or price dispersion. The New York Times bestsellers have the lowest prices as a fraction of the publisher's suggested price, and the random books have the highest prices. The authors observe differentiation (or attempted differentiation) in selection in certain categories, overall selection, and price. A few online stores, usually branches of specialty physical stores, focus on depth of offering within a particular category such as computer books, children's books, or religious books. Other full-line stores focus on overall selection, carrying 90 percent or more of the books in the sample. For those offering low prices, most stores focus on marginally undercutting Amazon, usually by 10 cents or less. As of January 2001, some of these stores have gone out of business or changed their business model, though, and the surviving ones still appear to be at risk.
Using a new data source on computer purchases of almost 30,000 people, Goolsbee estimates the relative price sensitivity of individuals' choice of whether to buy a computer online or in a retail store. Goolsbee finds that the decision to buy online is sensitive to the relative price of computers in retail stores. Conditional on buying a computer, the elasticity of buying online with respect to retail store prices is about 1.5.