Data from the 2002 Supermarket Panel are used to estimate a supermarket production function with weekly gross
margin as the output measure and store selling area and total labor hours as variable inputs. The model also includes
productivity shifters describing format and service offerings, store ownership structure, unionization, and adoption of
new information technologies and related business practices. The null hypothesis of constant returns to scale cannot
be rejected. Increases in ownership-group size, warehouse and supercenter formats, unionization of the workforce,
and adoption of vendor-managed inventory and a frequent-shopper program are all associated with significantly higher
productivity.
The supermarket industry experienced profound changes during the 1990s. Changes in store characteristics were readily apparent to consumers.
Median store size grew from 31,000 square feet in 1990 to 44,600 square feet in 2000 (Food Marketing Institute 2002b). New formats also emerged, most notably the supercenter. There were less-visible but equally important changes in industry structure. In 1990, independent supermarkets operated by companies owning ten or fewer stores accounted for 22.2% of all grocery sales, while chain supermarkets operated by companies with eleven or more stores accounted for 51.5% of grocery sales (Progressive Grocer 1991). By 2000 the respective shares of independent and chain supermarkets were 14.3% and 63.6% of grocery sales (Progressive Grocer 2001). This prompted a shift away from distribution by independent wholesalers toward self-distributing systems with retail stores and primary distribution centers under common ownership. Between 1992 and 1997 the percentage of total retail grocery sales supplied by independent wholesalers fell from 42.3% to 37.3% (A.T. Kearney 1998, p. 8), and it is likely that this trend has continued New information and communications technologies have also had important impacts on business
operations, decision processes, and trading-partner relationships in food retailing. Widespread adoption
of scanning technology and the Uniform Product Code during the 1980s provided the technological foundation for the introduction of electronic transmission of order data, industry-supported mechanisms for sharing scanner data, and computer-based product-movement analysis at the store level. Information technology also was the basis for significant changes in warehouse operations, logistics systems, and manufacturing processes. (Walsh 1993, pp. 89- 106; King and Phumpiu 1996). In the mid-1990s the Efficient Consumer Response initiative brought together food retailers, wholesalers, brokers, and manufacturers in an industry-wide effort to foster adoption of new technologies and business practices
based on information technology (Kurt Salmon Associates, Inc. 1993). More recently, rapid development of Internet-based technologies has fostered new initiatives in electronic commerce; scan-based trading; and collaborative planning, forecasting, and replenishment (Kinsey 2000).
While the general impacts of larger stores and new formats, changing industry structure, and new operating practices and trading-partner relationships based on information technology have been described and discussed by many, relatively little is known about how these changes have affected productivity at the store level. In this study we use data from a unique national survey of supermarkets, the 2002 Supermarket Panel, to estimate a store-level production function that includes explanatory variables describing not only store and organizational characteristics but also the adoption of new information technologies and related business practices. The overall objective is to analyze empirically how changes in supermarket operations—reflected in
store size and format, relationships with key suppliers, and new operating practices and trading-partner relationships based on information technology—are affecting productivity at the store level. Size Economies, Business Organization, and Information Technology The supermarket emerged as the dominant business model for food retailing in the years immediately following World War II. This fundamentally transformed the way consumers purchase food, combining self-service concepts pioneered in the 1930s with significantly larger stores that carried a much wider range of product offerings. As the size of the typical supermarket grew, not only in the U.S. but also in Europe, economists began to focus attention on economies of size and scale in food retailing. The empirical evidence has been mixed. Using data collected in the mid 1970s from fiftyeight stores operated by a large retail firm, Marion et al. (1979, footnote, pp. 135–137) find no statistically significant relationship between store size and per-unit operating expenses. In a 1981 review of published analyses of economies of size in food retailing, Grinnell (1981) concludes that most evidence available at that time indicated scale economies at the store level. Citing statistics reported by Progressive Grocer for 1988 and analytical results presented by Nooteboom (1983), Oi (1992) also asserts that there are size economies in supermarket operations. Still more recently, for a study using data collected by the Economic Research Service of USDA (Kaufman and Handy 1989), Betancourt and Malanoski (1999) report constant marginal cost for their measure of supermarket output (a quantity index constructed by dividing sales by an index of price relatives) but declining marginal costs for their measure of distribution services (an index based on store offerings for twenty specific services). They conclude that this implies overall multiproduct scale
economies for supermarkets. Grinnell (1981) and Marion, Parker, and Handy (1986) also emphasize the importance of considering economies achieved through ownership of multiple stores and through vertical integration of retail and wholesale functions—i.e., through differences in the structure of the business organization that operates a particular supermarket. Multistore economies can be achieved through advertising, increased buying power enjoyed by high volume firms, and savings on administrative functions that can be centralized to corporate headquarters. Retail companies that own their own distribution facilities may realize added cost savings through improved coordination in logistics and product-assortment decisions. Both Grinnell (1981) and Marion, Parker, and Handy (1986) assert that there are significant economies associated with multistore ownership and self distribution, but it is not clear whether these would be observed at the store level. Betancourt and Malanoski (1999) find that stores belonging
to chains with more than ten stores enjoy statistically significant cost savings, but they are not able to separate the effects of multistore ownership and
self distribution. A large, wide-ranging literature on the relationship between information-technology investment and productivity has emerged since the late 1980s, when researchers puzzled over the apparent lack of productivity gains associated with rapidly growing investments in computer hardware and software. As Devaraj and Kohli (2000) note in their review of previous research, this relationship has been explored at three distinct levels: the overall economy, an industry or cross section of industries, and the individual firm or establishment within an industry. Since this study focuses on store-level productivity, we limit our review to firm-level studies. Hitt and Brynjolfsson (1996) note that inconclusive or inconsistent findings regarding the economic impacts of information technology can sometimes be attributed to differences in performance measures. They assert that information technology can affect productivity, profitability, and consumer surplus. They go on to argue that conceptual frameworks and analytical methods for assessing relationships between information technology and each of these performance dimensions can be quite different. Production-function analysis has been the most commonly used framework for investigating firm-level relationships between information-technology inputs and productivity, which is the focus of this study. Most recent firm-level studies have found statistically significant, positive relationships between
information-technology investments and productivity. Brynjolfsson and Hitt (1996) and Hitt and Brynjolfsson (1996) report results of Cobb-Douglas
production-function analyses using panel data on spending for information-system inputs by large firms. These data were collected through annual
