We think that research on organizational diversity must attend to dynamics and must analyze the entire range of variation within types of organizations. The remainder of this chapter summarizes these themes and previews the approach that we use in empirical analyses reported in Part III.
The diversity of organizations in society depends on both the number of organizational forms and the distribution of organizations over forms. The diversity of a collection of organizations in a society can increase either because new forms are created or adopted or because the distribution of organizations over forms becomes more uniform. Social forces can affect diversity either by affecting the number of forms in use or by affecting the relative abundances of the various forms. The two dimensions of diversity are related at the extreme: when the number of organizations with a certain form is zero, the form does not exist in the system. However, ecological processes may differ in strength as they affect the number of forms and the relative abundances of forms. Thus it is useful to keep these two dimensions of diversity conceptually and analytically distinct.
Because organizational diversity has these two dimensions, analysis of the dynamics of diversity must consider a number of basic processes. First, there is the process of creating new organizational forms, including both inventing forms and borrowing forms from other systems. This process can be analyzed with reference to the establishment of abstract designs such as the rational-legal bureaucracy (Weber 1968) or the multidivisional form (Chandler 1977). It can also be analyzed with reference to more concrete, institutionalized forms such as the newspaper wire service or the comprehensive high school (Tyack 1974).
Second, there is the growth of numbers of organizations within existing forms. Interesting examples include the spread of colleges in the United States with the expanding frontier in the nineteenth century (Tewksbury 1932) and the proliferation of specific types of manufacturing firms such as semiconductor producers.
Third, some organizations adapt their core features, that is, they change from one organizational form to another. A classic example is the wholesale switch of single-gender colleges to the two-gender form during the late 1960s and 1970s. Similar changes are frequently reported in the industrial world (see Miles 1982, for example).
Fourth, there is the demise of organizations within populations. There are two quite different processes by which organizations cease to exist as corporate actors: disbanding and merger. Examples of disbanding are routine business failures (Wedervang 1965), closings of public institutions like prisons and mental hospitals, and the withering away of protest movements (Gamson 1975). Examples of merger include cases in which manufacturing firms purchase suppliers and distributors (see Chandler 1977; Williamson 1975) and the amalgamation of political parties.
Finally, there is the disappearance of entire organizational forms, which occurs when the number of organizations using the form reaches zero. Historically interesting examples in the United States include the demise of proprietary medical schools, utopian labor unions (Commons et al. 1927), and the party press (Schudson 1978).
Given our theoretical objectives, should we emphasize the creation of new forms and the demise of old ones, or rather the growth and decline of established sets of forms? In our empirical research we attempt to collect information on all the relevant processes. However, our theoretical work has made the most progress in pursuing the narrower questions of variations in the components of net mortality, the equivalent in organization theory of population ecology. Although we discuss the environmental factors affecting the rate at which new forms are created, this analysis is speculative.
Our approach emphasizes dynamics, the detailed processes of change in organizational populations. Almost all empirical work and most theoretical work on organizations deal with statics, with the relationships between variables in systems assumed to be at rest. Static analysis is appropriate when the systems under study are almost always in or near equilibrium. Such conditions are likely to obtain when environments are certain or stable or when the speed of adjustment of the systems is rapid relative to the rate of environmental change (Tuma and Hannan 1984). We think that both of these conditions are unlikely to hold for most kinds ot organizations.
We argue that organizational selection processes favor organizations with relatively inert structures, organizations that cannot change strategy and structure as quickly as their environments can change. If inertial pressures on organizational structures are indeed strong, it is highly unlikely that a great deal can be learned from static analysis. Instead one must study the actual time paths of change in organizational populations, looking in detail at the rates at which organizations enter and leave populations and the rates at which structures are changed in response to environmental threats and opportunities.
We have followed such a strategy. The chapters in Part II outline the research designs we used to trace patterns of change over time in organizational populations and our approach to estimating causal effects using the data produced by our research designs.
2. Comparative Analysis
Most writing and research on organizations concentrate on the largest and most successful ones of each type. There are a number of reasons for this emphasis. One motivation for studying the largest organizations is a concern with politics in the conventional sense. The biggest and most successful organizations presumably have the strongest impacts on the larger social structure. Given limited time and resources, one might expect to learn more about the impact of organizations on society by studying the “Fortune 500” than by studying more representative samples of firms.
A second motivation is practical: more information is usually available on the largest and most successful organizations. It is very difficult to gather information systematically on organizations with brief lifetimes. A third motivation is normative. Some analysts, especially those who want to promulgate a certain organizational form, want to direct attention to the most successful instances of the form as models for others. For example, Peters and Waterman (1982) draw conclusions about appropriate ways to organize from intense study of 62 of America’s most successful firms. Likewise, Kanter (1983) draws lessons about success in a changing technical environment from cases on a few of the largest and most solvent firms in the country. Thus organizational analysts, researchers, and prophets are drawn to the study of the biggest and longest-lived organizations.
Although it may seem reasonable to concentrate effort on the current winners, this is almost always a mistake if one wants to learn about the processes that create success and failure. The problem is that all key comparisons—between winners and losers—are implicit. Winners may have a certain structural feature; but how does one know that the losers did not have it as well? This is an instance of a very general methodological problem called sampling on the dependent variable or sample selection bias (see, for example, Heckman 1979; Berk 1983). The general result is that choosing a sample on the basis of values of the dependent variable causes estimates of causal effects to be biased. Thus restricting a study of profitability to the most profitable firms distorts inferences about the conditions producing success; it biases estimates of the effects of all independent variables on the rate of profit. Similarly, it is a mistake to make inferences about risk of business failure from studies of long- lived organizations.
The popularized literature on organizations typically glorifies one form of organization as best. Recent examples include Ouchi’s (1981) “theory Z” and Ranter’s (1983) “organic” organization. From an ecological perspective, the search for the one best organizational form is misguided. Whether a given form is superior to another depends upon the structure of the relevant environments. Just because some form has been successful in one country or in one industry at some time does not mean that it would be useful the next year under other conditions. In fact, ecological-evolutionary theory suggests that uncertain, volatile environments will support diverse organizational forms and that the apparent winners will fluctuate from time to time as conditions change. A potential benefit of a fully developed ecology of organizations would be the setting of limits on the claims that a given structure is best.
Our research designs make an effort to avoid sample selection bias. We chose to study populations of organizations for which it was possible to collect life histories on both large and long-lived and small and short-lived organizations. The use of such data to estimate parameters of dynamic models is a distinguishing characteristic of the empirical work reported in Part III.
Source: Hannan Michael T., Freeman John (1993), Organizational Ecology, Harvard University Press; Reprint edition.