This paper reports upon the profitability of firms that locate their headquarters in same-industry geographic concentrations or clusters and those that opt to maintain headquarters in other locations. While the preponderance of the theoretical and descriptive literature emphasizes the potential benefits associated with clustering, some papers suggest that clustering should not be beneficial, at least for particular types of firms in particular circumstances. This empirical study, which examines a sample of more than 4000 Compustat firms from 86 different industries, compares the profitability of firms in industry clusters and firms in other locations. The sample is partitioned into small and large firms to account for expected differences in profitability, in general, and the possible differential impact of geographic clustering. The results show that for smaller firms, the profitability of cluster members tends to be considerably lower than for firms that opt not to join clusters. For the subsample of larger firms, the results are mixed depending upon the measure of profitability. The results imply that smaller firms should carefully evaluate the decision to locate in industry clusters.
Numerous papers including Ellison and Glaeser (1997) [
Industries concentrate for a variety of reasons. Historically, firms concentrated to benefit from nearby natural resources such as oil and harbors. Other reasons for clusters include the availability of trained labor, the availability of specialized business services, and the enhanced ability to obtain information, including current and planned actions of competitors. Along these lines, Shilton and Stanley (1999) [
This paper investigates the association between cluster membership and financial performance for a large sample of US firms. The interest is in headquarters clusters rather than concentrations of operating units. The rationale for focusing upon corporate headquarters is that potential information spillover benefits should relate not only to technology, but also to corporate strategy, decisionmaking, and overall corporate culture at the levels of top management.
The extensive theoretical and descriptive literature that addresses industry concentration generally supports the notion that firms benefit through membership in geographic clusters. Extant empirical studies, which typically include only a single industry or small set of industries, however, do not suggest that cluster firms are more profitable than other firms. The results of the large sample examination reported here show that clustering may be detrimental in that smaller-size cluster firms are less profitable than firms that opt not to join clusters. These findings provide insights on the relative costs and benefits of cluster membership and as such, have implications for the financial and strategic management of firms.
The next section reviews pertinent theoretical and empirical literature. Following this we present the study design, data, and results.
Historically, a major reason for business concentration was the desire to benefit from access to nearby natural resources. Marshall (1920) [
Delgado, Porter, and Stern (2010) [
Davis and Henderson (2008) [
Krugman (1986) [
Spillover refers to the tendency for knowledge to “spill over” for exploitation by others. The potentially beneficial effects of informal information exchanges between technical and managerial employees of competing firms may well be the most significant advantage of clusters, particularly for technology-dependent businesses. Much of the recent literature emphasizes the role of knowledge spillover, the transmission of knowledge. Baptista and Swann (1998) [
We generally think of spillover in the context of R&D. However, information exchanges, even in non-technical businesses, can foster awareness of changes in market conditions, awareness of new developments in management strategy, and can facilitate the monitoring of competitors (Sorenson & Audia, 2000 [
Along similar lines, Porter (1998) [
The literature suggests that firms in industries with significant R&D requirements benefit relatively more from cluster membership than firms in other industries. However, businesses in industries such as transportation and retailing face continuous competitive pressure and also require innovation and adoptive behavior to remain viable. At a minimum, the spillover associated with industry proximity should better ensure top management awareness of changes in business practices and competitive situations.
The receivers of spillover should obviously benefit through cluster membership. However, it is less clear that firms in possession of proprietary information should participate in clusters, thereby contributing to the success of competitors.
Yang, Phelps, and Steensma (2010) [
Yang et al. test this originator-benefit hypothesis using a sample of patent data. Their results support the notion that spillover provides benefits to originating firms.
On a theoretical level, Tallman, Jenkins, Henry, and Pinch (2004) [
Other literature illustrates the development and success of specific clusters. The Pinch and Henry (1999) analysis of the British racing car industry was cited earlier. Along similar lines, Saxenian (1994) [
One may question the relationship of geographic location and spillover given instantaneous electronic communication and the seemingly unlimited availability of electronic data. However, Evers, et al. (2010) [
Bell (2005) [
In summary, membership in industry clusters would seem to be advantageous for a number of reasons. One is the potential for reduced transport costs. Another is the potentially favorable effect upon growth and entrepreneurial activity. Clusters also tend to include industryspecific specialists who can provide, in addition to services, information and advice based upon their knowledge of industry events and facilitates opportunities for face-to-face interaction. The literature suggests that the potential for spillover associated with geographic proximity may serve as the single most important advantage of cluster membership and that spillover should benefit both the firms that have developed technological advances and those that have not.
While the majority of the literature suggests that clustering should be beneficial, some papers suggest that clustering may adversely affect financial performance. Data, to be reported later, indicates that the majority of firms avoid locating in major industry concentrations. Both the Silicon Valley and Boston areas, for example, host concentrations of computer chip businesses. However, a number of large semi-conductor companies in this highly technology-dependent industry maintain headquarters in diverse locations including Portland, Oregon, and Phoenix, Arizona. If these firms believe that clustering is beneficial, we should expect them to join the primary industry clusters. In addition, entire industries including trucking, freight, grocery stores, and department stores show little evidence of headquarters concentrations. WalMart Stores, Inc., for example, is located in Bentonville, Arkansas, a relatively small community. The decision of most firms to avoid cluster memberships suggests that clustering should not enhance financial performance.
Geographically remote firms should be able to compensate, in part, for their non-central locations through increased reliance upon travel to trade shows, conferences, alumni reunions, and similar networking-oriented events and through the increased use of electronic communications. Some firms, such as Wal-Mart, work with nearby communities of specialized service providers. Other businesses compensate for relatively isolated locations by hiring innovators from other firms and through the formation of alliances (Rosenkopf & Almeida, 2003) [
The literature sets forth a number of reasons to expect that clustering may adversely affect performance. These include the loss of effectiveness over time, the potential for complacency and inward thinking, firm-specific considerations, and the expected costs associated with the possible loss of proprietary information and key employees.
A number of papers discuss the growth and decline of clusters and the reduced likelihood that firms will benefit through membership in declining clusters. Pouder and St. John (1996) [
Cluster firms may also be subject to complacency and cluster-oriented rather than total-industry thinking. Pouder and St. John (1996) [
It also seems likely that firms do not benefit equally through cluster membership. Shaver and Flyer, (2000) [
Wennberg and Lindqvist (2010) [
The desire to minimize costs may also be responsible for cluster avoidance on the part of most firms. Although firms may benefit through membership in clusters, the costs of maintaining headquarters in high wage and rent areas such as the Silicon Valley and New York City likely tend to keep some firms away from these same areas. Firms may also elect to maintain headquarters in remote locations in the belief that indirect costs will outweigh the benefits. Indirect costs include the inadvertent leakage of valuable proprietary information to competitors and the increased likelihood that competitors will tap top employees. While firms in clusters may find it easier to raid competitor’s employees, raiders are also more vulnerable. Similarly, locating in proximity to competitors increases the likelihood that these competitors will obtain trade secrets. Consequently, it isn’t clear that the net effect of clustering is positive, particularly for cluster members with valuable employees and proprietary information.
In summary, if cluster members outperform other firms, we should expect to observe that the majority of firms are members of industry clusters. This is not the case. The literature also suggests a number of reasons to expect that cluster members should not outperform other firms. These include the loss of cluster effectiveness over time, the potential for complacency and inward thinking, firm-specific considerations, and the expected costs associated with the possible loss of proprietary information and key employees.
A number of empirical papers examine associations between clustering and various indicators of business performance. These studies, which tend to focus upon patents and other measures of innovation, generally reveal that firms in clusters outperform other firms.
Jaffe (1986) [
Audretsch and Feldman (1996) [
Baptista and Swann (1998) [
Other papers focus upon patent citations. Jaffe et al. (1993) [
Another set of literature examines the impact of cluster size. Folta, Cooper, and Baik (2006) [
The empirical literature also addresses the association of clusters and entrepreneurship. Delgado, Porter, and Stern (2010) [
The Delgado, Porter, and Stern findings appear to contrast with those reported by Stuart and Sorenson (2003) [
Pirinsky and Wang (2006) [
A small group of studies examines various measures of financial success. Shaver and Flyer (2000) [
In summary, an extensive theoretical and descriptive literature addresses the potential advantages associated with clustering. The empirical evidence, however, is industry-specific, consists of relatively small samples of firms, and reports mixed results. Story and Westhead (2009) [
This study compares the profitability of firms that locate corporate headquarters in industry-specific concentrations and those that opt not to participate in clusters. Other studies have focused upon research locations and manufacturing locations. The interest in corporate headquarters is motivated by the potential benefits of clustering, and information exchanges, in particular, to all levels of management.
The initial sample includes all Compustat firms for the years 2004 through 2009. The observation period, the most recent available at the time of data collection, includes data representing both strong and weak economic conditions. Following the usual procedure in financial studies, we then excluded overseas firms, financials, utilities, companies with less than $1 million in assets, firms in industries with fewer than 10 companies, and firms missing necessary data. The final sample consists of 20,969 firm-years of data (total number of observations) in 86 industries. Over 4000 distinct firms are included.
Firms are classified by industry using the Global Industry Classification Standard (GICS), a system developed as a joint undertaking in 1999 by Standard & Poor’s and MSCI/Barra. Bhojraj et al. (2003) [
We should note that while the assignment of firms to industries on the basis of GICS (or SIC) codes) is objective and convenient, it does have limitations. One is that many firms operate in more than one industry. Boeing Corporation, for example, produces not only commercial airplanes, but also military aircraft, and space systems. Unrelated or loosely related sectors such as these typically use different suppliers, different technology, and vastly different marketing. Unfortunately, industry coding systems necessarily assign Boeing and other multiindustry firms to a single industry.
Adopting the procedure in Pirinsky and Wang (2006) [
The US Office of Management and Budget (OMB) assigns counties, and in some cases cities and towns, to 374 MSAs. OMB aggregates these into combined statistical areas (CMSAs) for the very largest metropolitan areas. The New York City CMSA, for example, includes parts of 4 states and its Year 2000 population exceeded 21 million people. At the other end of the distribution, fewer than 58,000 people live in the MSA that includes Enid, Oklahoma. This study assigns firms to CMSAs, where available; otherwise to MSAs. The assignments of firms to MSAs use Compustat county information. In cases of missing data, the location information is obtained from other sources.
The sample is comprised of firms in 86 distinct industries. These industries differ considerably in size. The smallest consists of only 52 firm-years and the largest includes 854 firm-years. On average, each industry includes 254 firm-years.
We refer to the metropolitan area with the largest number of firm-years in each industry as the primary cluster. All firms in the industry which are located in this metropolitan area are referred to as cluster members. As a number of industries appear to have large concentrations of firms in more than one location, the study also incorporates an alternative definition of cluster membership. Firms in those metropolitan areas which serve as headquarters for at least 5 other same-industry firms are considered to be members of “regional clusters.” For measurement purposes, our regional clusters require at least 25 firm-years of data in the same industry.
Our analysis separately considers small and large firms. The reason is that larger firms, which are less likely to include startups, tend to be more profitable. We define small firms as those with total assets less than the industry median. Large firms report assets greater than or equal to the industry median. Partitioning on the basis of medians, rather than means, results in equal size subsamples.
Panel B shows that cluster membership increases to 38% when we consider both primary and regional clusters. The data shows that smaller firms are also less likely than larger firms to be members of regional clusters. We should note that firms are included in the study only when their respective industries report at least 50 firmyears of data over the observation period. Consequently, the reported percentage of firms that are cluster members actually overstates the percentage for the population of firms, as a whole.
The data show that the extent of clustering varies considerably between industries. We observe that primary cluster sizes range from 14 to 313 firm-years. We also observe that primary clusters tend to be located in the very largest cities. The New York City metropolitan area, for example, hosts 36 of our 86 industries and San Francisco ranks second with 10 primary clusters. Houston follows with 8 clusters, all of which are engaged in various types of oil and gas businesses. These 3 cities host more than 60% of the primary clusters examined in the study. It may seem as though the largest numbers of clusters and the sizes of these clusters correspond to the sizes of cities, in which case, cluster membership would seem to be merely a proxy for the size of the city. This, however, is not necessarily the case. For example, although Philadelphia and Detroit rank as the sixth and eighth largest US metropolitan areas respectively, each of these cities hosts the primary concentration of only one of the 86 industries.
The data show that the largest primary clusters tend to be located in the New York City and San Francisco areas. We also observe that the percentage of firm-years comprising the largest clusters varies from 7% to 62%, a substantial spread. Not surprisingly, the restaurant industry shows a low propensity to cluster, likely due to the regional nature of some restaurant chains and the large variation of restaurant types. It is surprising, however, to observe that only 8% of the industrial machinery firms are located in that industry’s primary cluster. In common with the restaurant industry, this may reflect substantial variation in industry machinery products.
This section presents data concerning the association of business performance and headquarters location, the primary objective of this paper. Following the majority of predecessor financial studies, profitability is measured as the return on assets (ROA). Due to small reported asset numbers in a number of cases, ROAs are sometimes very large in absolute terms. Consequently, ROAs are truncated at 0.50, when necessary. We also present median
This table shows the distribution of the sample based upon firms that cluster (CONC) and those that do not cluster (Not).
return, a measure which is less sensitive to outliers.
These data reveal that smaller firms that cluster perform more poorly than firms which choose to maintain headquarters in locations not populated by competing firms. We conducted a two-sample t-test for differences between returns for small firms that cluster and for noncluster firms. This test reveals that ROAs differ between the two groups at the 1% significance level. The magnitudes of differences between cluster and non-cluster members suggest that the difference in profitability is also economically significant.
The fourth and fifth columns in Panel A show results for larger firms. While large firms that do not cluster are more profitable than large cluster firms, the performance differential is considerably lower than for the smaller firms. Although the difference between cluster and noncluster firms is statistically significant at the 1% level, the economic significance of the differences is subject to question, particularly when interpreting the relatively small difference in median values.
Panel B compares firms in regional clusters and other firms. While the general pattern of results is the same as in Panel A, the performance differences between cluster and non-cluster firms are more pronounced. In fact, both the mean and median measures of return on investment suggest an economically significant profitability advantage for non-cluster firms.
Previously we summarized a number of papers that relate to the differential effects associated with clusters. Shaver and Flyer [
This table shows mean and median returns on assets for small firms and large firms (20,969 observations). Two-sample t-tests show that all differences in returns for cluster and non-cluster firms are significant at the 1% level.
pliers, and other assets will gain little and suffer the most through spillover to competitors and suggest hypothesize that the best performing firms will choose not to cluster. In discussing their empirical examination of startup firms, Stuart and Sorenson (2003) [
We also examine industry-adjusted returns on assets in an effort to separate the effects of individual-firm performance and overall industry performance. The procedure involves determining industry median returns for each of the 86 industries. Industry-adjusted returns are then calculated as the individual firm return less the industry median return. Panel A in
Panel B shows industry-relative returns. For the smaller firms, the data generally reflect the same underlying pattern as observed for the unadjusted returns measure except that the difference between cluster and non-cluster firms is more pronounced. The results for the
This table shows industry adjusted mean and median returns on assets for small firms and large firms (20,969 observations). Two-sample t-tests show that all differences in returns for cluster and non-cluster firms are significant at the 1% level.
large firms located in regional clusters are different, however, from those reported in
Taken as a whole, Tables 3 and 4 data reveal that smaller firm members of both primary and regional clusters do not perform as well firms that avoid clusters. This conclusion contrasts with the situation for larger firms where the differences in returns between cluster and noncluster firms tend to be smaller and where the sign of the differences between returns depends upon the particular measure of profitability. Larger firms in clusters underperform based upon unadjusted ROA, but outperform larger non-cluster firms when returns are adjusted for industry returns. Given these results, it does seem safe to conclude that non-cluster smaller firms outperform those that locate in clusters and that these differences are economically significant. It is more difficult to reach conclusions for the subsample of larger firms given the mixed results obtained from alternative measures of profitability.
This paper reports upon the profitability of firms that locate in same-industry geographic clusters and those that opt to maintain headquarters in other locations. While the preponderance of the theoretical and descriptive literature emphasizes the potential benefits associated with clustering, some papers do suggest that clustering should not be beneficial, at least for certain types of firms in certain circumstances.
This empirical study examines a sample of more than 4000 Compustat firms from 86 different industries. Firms located in metropolitan areas which host the greatest numbers of firms in their particular industry are members of primary clusters. Approximately 22% of the sample firms fall into this category. Metropolitan areas which host more than about 10 firms (50 firm-years of observations) are considered to be regional clusters (which include primary clusters). About 38% of the sample firms participate in either primary or regional clusters.
Following the majority of financial studies, profitability is measured as the return on assets. In order to separate the effects of individual company profitability from that of industry profitability, we also calculate industry-adjusted returns on assets. Since the literature shows a relationship between asset size and profitability, we also partition the sample into those with smaller and larger asset size compared to the industry median amount.
The results show that for the set of smaller firms, profitability is considerably lower for firms that cluster compared to firms that opt not to join clusters. The data also show that this difference is both statistically and economically significant. For the sample of larger firms, the results depend upon the measure of profitability. Consequently, it is difficult to draw conclusions based upon these tests. These results imply that smaller firms should carefully evaluate the decision to locate in industry clusters.