!!! Measuring Environmental complexity.full[1]

296 Measuring Environmental Complexity A Theoretical and Empirical Assessment Alan R. Cannon University of Texas at Arlington Caron H. St. John Clemson University Researchers attempting to measure environmental complexity remain challenged by the lack of a theoretically compelling and empirically sound scheme for operationalizing this impor- tant construct. The authors’ synthesis of complexity’s supporting literatures leads them to con- clude that complexity is a multidimensional construct that has been operationalized overly narrowly in many cases. Exploratory and confirmatory factor analyses of industry-level data drawn from two periods support this conclusion and suggest that researchers should consider specifying a multidimensional measure of environmental complexity. Keywords: organizational environments; environmental complexity; dimensionality; confirmatory factor analysis For decades, it has been widely held among organizational researchers that organizationsmust adapt to their environments if they are to survive and succeed (Lawrence & Lorsch, 1967; Thompson, 1967). Since the seminal work of Emery and Trist (1965), researchers have attempted to define and describe the “components and relevant dimen- sions” (Duncan, 1972, p. 313) of environments. The purpose of these efforts has been to determine the effects of environments on management decision making, strategy choice, organization characteristics (e.g., structure or information processing), and individual, workgroup, and organizational performance (cf. Barney, 1991; Bourgeois, 1985; Child, 1972; Dess & Origer, 1987; Duncan, 1972; Flynn & Flynn, 1999; B. Gibbs, 1994; Lawrence & Lorsch; 1967; Miller, 1987, 1988; Sanchez, 1997; Sharfman & Dean, 1991a, 1991b; Tung 1979; Tushman & Nadler, 1978). Most researchers and theorists have identified complexity as one of the important char- acteristics of environments. Emery and Trist (1965) were among the first to recognize com- plexity versus simplicity in environments, followed soon thereafter by Lawrence and Lorsch (1967) and Thompson (1967), who described complexity as environmental hetero- geneity or diversity. In the years that followed, researchers developed several definitions and measures of environmental complexity. As described by Boyd and Fulk (1996), several “labels” have been used to capture complexity, “including effect uncertainty (Milliken, 1987), analyzability (Daft & Weick, 1984; Perrow, 1970), predictability (Duncan, 1972), and the utility of information in decision-making (Duncan, 1972; Perrow, 1970)” (p. 3). Organizational Research Methods Volume 10 Number 2 April 2007 296-321 © 2007 Sage Publications 10.1177/1094428106291058 http://orm.sagepub.com hosted at http://online.sagepub.com Cannon, St. John / Measuring Environmental Complexity 297 These different definitions, measurements, and interpretations have contributed to a lack of conceptual clarity. As described by Sharfman and Dean (1991a) in a discussion about environments: Neither a single set of constructs nor a single set of measures of the organizational environment is widely accepted, making it difficult to build a comprehensive literature on the impact of the environment on the firm. . . . Because neither a single approach to conceptualizing the environ- ment nor to measuring it has received widespread acceptance, we have been unable to build a comprehensive, coherent literature about the environment and its impact on the firm. (p. 681) Similar arguments were advanced by Boyd, Dess, and Rasheed (1993) in their review of objective/archival and perceptual measures of environments. In this article, we reexamine one of the characteristics of environments—complexity—both conceptually and empiri- cally. We first review the literature on environmental complexity and then apply factor analysis to identify dimensions of the complexity construct. We consider the effects of using single versus multiple measures in a typical inquiry, then conclude by discussing the implications of our findings for existing and future research. Theoretical Foundation Foundation Definitions of Environmental Complexity During the past 40 years, the construct of environmental complexity has been variously defined. In their seminal work on the causal texture of environments, Emery and Trist (1965) identified the environmental complexity-simplicity dimension as one of the charac- teristics of environments with which focal firms must contend. Drawing on a similar con- cept, Thompson (1967) classified industries along a continuum of homogeneity-heterogeneity, with more heterogeneous environments imposing greater constraints on organizations. Lawrence and Lorsch (1967) identified an analogous dimension in their environmental diversity scale. Adding more to the definition, Child (1972) described environmental complex- ity as heterogeneity and range in the activities that are highly relevant—that is, strategic— to an organization’s operations. Duncan (1972) was the first to specifically operationalize the complexity construct. Elaborating on the Emery and Trist (1965) “simple-complex” term, Duncan conceptualized complexity as: (a) the number of factors in the decision environment and (b) the dissimi- larity or heterogeneity among them. Furthermore, Duncan distinguished between com- plexity in the internal and external environments of the organization, with the external environment characterized as including customers, suppliers, competitors, technology, and sociocultural components. In an extension to that work, Tung (1979), defining complexity as the number and heterogeneity of factors and components with which managers must contend, explained that increasing complexity limited “the CEO’s cognitive abilities to grasp and comprehend the relationships that exist among them” (p. 675). Drawing from resource dependency theory, Aldrich (1979) argued for a multidimen- sional interpretation of environmental complexity as: (a) the number of units with which interaction is required and (b) the extent to which an organization requires specialized or sophisticated knowledge to cope with complexity. According to the interpretation of Sharfman and Dean (1991a, 1991b), Mintzberg (1979) used the term market diversity to mean Thompson’s (1967) heterogeneity and complexity to capture the need for sophisti- cated technical knowledge, similar to Aldrich’s second dimension. Sharfman and Dean interpret the market diversity and complexity terms offered by Mintzberg to mean the breadth and depth of knowledge needed for effective interaction in the environment. Similarly, within manufacturing environments, complexity has been conceptualized in terms of the level of technological complexity observed in the industry. Singh (1997) defined a complex technology as “an applied system whose components have multiple interactions and constitute a nondecomposable whole” (p. 340). Kotha and Orne (1989) distinguished between complexity as it applies to the product line and as it applies to the manufacturing processes used to produce the product. Product-line complexity is charac- terized by the number of different products produced, complexity of the products (number of subcomponents), and range of product volumes. As described, product-line complexity is similar to measures offered by Dess and Beard (1984) in that it captures product and sup- plier diversity and implies market diversity. Process structure complexity, however, is different from previously discussed ideas. Process structure complexity is determined by the levels of mechanization, systemiza- tion, and interconnectedness within and among manufacturing processes (Kotha & Orne, 1989). Using the Kotha and Orne (1989) typology, industries with large, mixed- model assembly plants (e.g., automobile) and integrated, continuous flow productions operations (e.g., petroleum refining, paper processing) face high process structure com- plexity. Capital-intensive operations, which are generally characterized by processes that are more automated and integrated, would exhibit higher process structure com- plexity. As described in various treatments of product-process evolution in industries (Abernathy & Utterback, 1978; Hayes & Wheelwright, 1979a, 1979b), product com- plexity is often higher in the early stage of industry evolution, declining over time, whereas process complexity may increase over time as more technologically sophisti- cated (automated, fully integrated) processes are implemented to produce a more nar- row set of commodity products. Other researchers have focused on the increases in information requirements associated with increased numbers of external constituents or complexity. Decision making amid com- plexity requires a greater understanding of the environment; managers in complex envi- ronments must know and consider more than those in relatively straightforward ones (Sharfman & Dean, 1991a, 1991b). Increases in information requirements can result either from the breadth of organizational activities or linkages that must be considered (Pfeffer & Salancik, 1978) or from the level of intellectual and/or technical sophistication required for comprehension (B. Gibbs, 1994). The industrial organization (IO) literature has taken a similar but narrower approach to conceptualizing complexity—primarily treating it as a manifestation of competitive intensity. At the heart of these models is a general character- ization of firms as output restrictors, seeking monopoly rents via direct or indirect collu- sions with other firms (Conner, 1991). Collusive actions are difficult to maintain under conditions of uncertainty, however, and uncertainty grows as the number of competitors increases and their relative market shares converge (Stigler, 1964). IO researchers infer oligopolistic behavior (i.e., increasing clarity with respect to competitors, their potential 298 Organizational Research Methods tactics, and valid responses thereto) when a small number of firms control a large portion of industry output (Scherer, 1980). The preceding literature, taken as a whole, leads to an understanding that the primary components of the environment that give rise to complexity are suppliers, customers, com- petitors, technology, and sociocultural factors (Duncan, 1972). And it is the nature of some or all of these components that provides three subdimensions to a definition of complexity. First, complexity is a function of the number of environmental components with which the firm must interact. Second, given some number of environmental components, complexity is a function of heterogeneity, dissimilarity, or diffusion among them. Third, given the pres- ence of particular environmental components, complexity is a function of the sophisticated or technical knowledge required to interact effectively with them. These distinct and poten- tially independent subdimensions are reflected in efforts to measure complexity during the past two decades, which we review in the next section. Measurements of Environmental Complexity Measurements of environmental complexity are generally of two types: (a) perceptual and (b) objective or archival. Perceptual measures, such as those proposed by Duncan (1972) and Tung (1979), make use of surveys, with organizational members as respondents. These perceptual measures are often used to capture the level of perceived environmental uncertainty and its effect on managerial decision making (Boyd et al., 1993; Boyd & Fulk, 1996). Objective measures, on the other hand, are calculated using industry-level data and allow the environmental characteristics of one industry to be compared to another (e.g., Lawless & Finch, 1989). As noted by Boyd and Fulk (1996), these measures can be repli- cated and compared because they are developed from readily available archival data sources. Such has been the case with the work of Dess and Beard (1984), which combined two subdimensions of complexity—the number of and heterogeneity among organizational requirements or constituents—and served as a standard reference for a wide variety of works involving objective or archival measures of organizational environments at the indus- try level (Boyd et al., 1993; Lawless & Finch, 1989). Dess and Beard (1984) used data at the 4-digit Standard Industrial Classification (SIC) level to extend the work of Aldrich (1979) and frame complexity as a function of homogeneity-heterogeneity and concentration-dispersion. As shown in Table 1, the authors considered eight measures of industry complexity, as introduced by resources (inputs), cus- tomers (outputs), breadth of product line, competitors, and geographical concentration. As can be seen in Table 2, which chronologically summarizes many subsequent studies consid- ering complexity, much later work reflects the Dess and Beard approach either directly— through adoption of the authors’ original measures—or indirectly—through emphasis on component preponderance and heterogeneity. This is true even for research grounded by IO theory, given that stream’s strong interest in a particular class of environmental components— competitors. Component heterogeneity. Early extensions to the work of Dess and Beard (1984) rested heavily on IO theory and its depiction of complexity. The fundamental logic of this depic- tion is that because industries with few firms holding substantial market power are more Cannon, St. John / Measuring Environmental Complexity 299 300 Organizational Research Methods strategically tractable, measures of competitive concentration are reasonable proxies for heterogeneity among competitors, often the most relevant environmental components. Measures of competitive concentration include n-firm ratios (i.e., the market share con- trolled by the n largest firms) and the Herfindahl (H) index (Shughart, 1990), the sum of the squared market shares of all industry incumbents. The earliest work after Dess and Beard (1984) that reflected this view was that of Keats and Hitt (1988), in which environ- mental complexity was measured using a dynamic measure of competitive concentration developed by Grossack (1965). Their measurement captured the trend toward or away from dominance by large firms during a 5-year period, indicating the increase or decrease in monopoly power. Boyd (1990) continued this emphasis on component heterogeneity by using the H-index as his measure of competitive diversity. Although they maintained attention on component heterogeneity by measuring product diversity and geographical concentration, Sharfman and Dean (1991a, 1991b) criticized the overemphasis on this one source or dimension. Left unmeasured, they argued, was the chal- lenge of managing in environments requiring high levels of technical or scientific sophisti- cation. Their arguments reflected those of Aldrich (1979) and Mintzberg (1979), who both proposed that complexity was higher in environments requiring sophisticated scientific or technical knowledge. Thus, to their component heterogeneity measures, Sharfman and Dean added one intended to tap technical intricacy, operationalized as the percentage of sci- entists and engineers in an industry’s workforce. Table 1 Dess and Beard’s (1984) Measures of Environmental Complexity Dimension Definition Measurement Heterogeneity-homogeneity Concentration of industry inputs Function of the dollar volume of (sources of supply) inputs and the number of industries supplying the inputs Concentration of industry outputs Function of the dollar volume of (customer groups) outputs and the number of industries to which outputs are supplied Diversity of industry products Function of the number of product (breadth of product line, in codes and dollar volumes different industries) Specialization ratio (degree to Ratio of primary product shipments which firms in the industry are to total product shipments for all concentrated in one industry) establishments classified in the industry Concentration-dispersion Geographical concentration of sales Function of dollar volume of industry sales and number of census divisions Geographical concentration of value Function of dollar volume of industry added by manufacturers value added and number of census divisions Geographical concentration of total Function of total industry employment employment and number of census divisions Geographical concentration of total Function of total establishments and establishments the number of census divisions Cannon, St. John / Measuring Environmental Complexity 301 Continuing the emphasis on component heterogeneity was the work of Wiersema and Bantel (1993), in which environmental complexity was assessed via industry specialization ratios (i.e., the proportion of an industry’s shipment accounted for by primary products). This was followed by Kotha and Nair’s (1995) choice of a modified version of the H-index in which only the four largest shareholders in an industry were considered. Other published works using similar measures include Dean and Snell’s (1996) examination of the strategic Table 2 Complexity Measures Used Since Dess and Beard (1984) Primary Subdimension Tapped Complexity Measure Component Component Required Study or Measures Preponderance Heterogeneity Knowledge Keats and Hitts (1988) Grossack’s ratio X Boyd (1990) Herfindahl index X Sharfman and Dean Diversity of product categories X (1991a) Workforce’s % scientists/ X engineers Workforce’s geographical X concentration Competitors’ geographical X concentration Wiersema and Bantel Specialization ratio X (1993) Kotha and Nair (1995) Modified Herfindahl index X Dean and Snell (1996) Schmalensee index X Miller and Chen (1996) No. of geographic markets X No. of competitors faced X Jarley, Fiorito, and Avg. employees per firm X Delaney (1997) Kostova and Zaheer No. of regulatory domainsa X (1999) No. of cognitive domainsa X No. of countries in which multinational X enterprise (MNE) operatesa Variety of countries in which X MNE operatesa Note: Regression coefficient for industry firms’ end-of-year market shares regressed onto beginning-of-year shares. Sum of the squared market shares for all firms in an industry. Total number of product categories (7-digit Standard Industrial Classification [SIC] codes) within an industry. Measured at 3-digit SIC code level. Sum of squared industry employee totals divided by the squared sum of employee totals across census divi- sions. Sum of squared industry establishment totals divided by the squared sum of establishment totals across census divisions. Ratio of primary product shipments to total (excluding miscellaneous) product shipments. Sum of the squared market shares for largest 4 firms in an industry. Herfindahl approximation via weighting of market shares of largest 4 and second-largest 4 firms in an industry. Number of airports served by sampled air- lines. Number of competing airlines faced by sampled airlines. Measured at 2-digit SIC code level over a 5-year period (1983-1987). a. Proposed use of integrated manufacturing, in which the authors operationalized complexity using Schmalensee’s (1977) approximation of the H-index. Component preponderance. Later published works tended to emphasize the view of environmental complexity as primarily a function of scale, with particular attention paid to the number of environmental components faced by the typical industry member. Such a scale can result from the exercise of strategic choice, as in Miller and Chen’s (1996) work, in which the authors measured complexity as simply the number of markets served and the number of competitors faced by individual organizations. Similarly, some have argued for the use of firm size as a measure of complexity, making reference to the increased numbers of products and markets that are often associated with organizational scale. In their study of U.S. national unions, for