In preparation for other uses of the data, this is a technical report on the interlocking directorates among companies used to define three widely-used market indices of the American economy at the turn of the century: the S&P 500, S&P SmallCap 600, and S&P MidCap 400. I draw primarily on director and company data assembled by the Investor Responsibility Research Center (IRRC). The time interval described is 1999 through 2003 an interval spanning the two years preceding and two years following the first year of the century. I make six points in this note. With respect to the study population: (1) The IRRC data are not a panel through the five years so much as a sequence of cross sections. Companies are selected for the indices because their performance is believed to indicate broader market performance. I refer to a selected company as an "index company." About 1,500 companies are observed each year. The average board of directors contains nine or ten people, so about 15,000 directorships held by 12,000 individuals are observed each year (one individual can sit on more than one board). (2) The index companies overlap extensively with the familiar Fortune 1000 roster. Larger companies are more likely to appear in each year of the IRRC data.

With respect to personal connections between boards, let "single-seat" directors refer to people who sit on the board of a single index company during a year and let "multi-seat" or "interlocking" directors refer to people who sit on the boards of two or more index companies, creating a link between the boards on which they sit. Interlocking seems to be about social standing and information. Directors brought in from other companies are status-enhancing to the board and looking to learn things useful in their own situation. The CEO who invited them is looking for quotable advice and counsel from prestigious experts, expressed in a civil, sympathetic way.

There is structure to the interlocks: (3) The average board has an affiliated Chairman corroborating the CEO, single-seat independents serving on board committees, and interlocking directors (affiliated and independent) used on committees as channels to external information. (4) The more boards on which a director sits, the more likely the director is a woman, a minority, and over the age of 65. (5) Interlocks are more likely with larger companies and more likely in certain lines of business. Finance companies, contrary to their central role in interlocks through much of the twentieth century, stand out for their disproportionate number of single-seat directors at the end of the century (Davis and Mizruchi, 1999). At the other extreme, companies in the manufacture and distribution of durable goods have disproportionate numbers of interlocking directors.

(6) My final point is the regional pattern to the interlocks. When directors interlock company boards, they connect the geographic places in which company headquarters are located. There are three regional patterns to the interlocks: First, interlocks are concentrated in central cities (Dooley, 1969; Allen, 1974). Second, there is a strong preference for directors from one's own region (Kono et al., 1998). Local elites play more prominent board roles, the odds of an interlock decrease with geographic distance between two companies, and interlocks are concentrated within regional categories. Third, again holding constant regional differences in interlock volume, each region prefers directors from certain other regions such that there is a geographic interlock network (Table 11, Figure 8). The network reflects historical boundaries in the United States, with a cluster of connected areas in the former Confederate states, an array of connected areas in the former Union states, and a cluster of areas in the former Western territories. The Southern Cluster is largely segregated from the rest of the country except for a brokerage port through St. Louis. The Northern Cluster is anchored on a cohesive East Coast subcluster, itself anchored on New York City, surrounded by a balkanized Midwest. The Western Cluster is anchored on a cohesive West Coast subcluster, itself anchored on Los Angeles and the San Francisco Bay Area, with satellites Houston, Dallas Fort Worth, and the Mountain States. The regional clusters are held together by four network bridges: St. Louis is a port out of the Southern Cluster, a port anchored on strong connection between SBC Communications and Anheuser-Busch. New York City is a broadly-connected port to locations in the Northern and Western Clusters. The East and West Coasts are connected by a bridge of links between technology companies in Boston and San Francisco. Fourth, there is a network bridge between Los Angeles and the twin cities of Minneapolis and St. Paul anchored in industrial and consumer-goods companies.

My goal in this paper is to use the sociology of markets to better define industries for organization research. How an organization operates, and how well it operates, depends in large part on its fit to the market environment, the industry, in which it operates. Valid distinctions between industries are therefore a critical exogenous factor affecting the quality of organization research. The market boundaries around an industry are defined in theory by the network concept of structural equivalence: products are in the same market to the extent that their production involves purchases from the same supplier markets and sales to the same customer markets. I apply the structural equivalence criterion to detailed transaction data from the U.S. Department of Commerce and get three results: (1) The analysis yields a partition of the American economy into 123 industries (versus the 77 distinguished by Commerce before 1987, and the 88 thereafter), showing where product distinctions are negligible, and where distinctions between structurally nonequivalent kinds of products are needed. (2) Market boundaries are defined less by the homogeneity of transaction patterns within industries, than by differences between industries. Many industries contain products that are structurally nonequivalent with respect to an important supplier or customer transaction. (3) Nevertheless, the proposed partition into 123 industries promises stronger results for organization research because transaction patterns are more equivalent within the proposed industries (more reliable market boundaries), and performance differences are more between than within the proposed industries (market boundaries with higher construct validity). Structure-performance scores on several variables for the proposed industries are presented and can be downloaded from the internet.