The unit of observation in the studies we report is the individual organization. We follow the life histories of individual members of organizational populations, studying events such as founding, disbanding, and merger. The formal statistical analyses applied to such histories may look little different from analyses designed to test propositions at the level of the individual organization. In what sense do the analyses address questions stated at the population level?
Consider the case of organizational mortality, which plays a prominent role in ecological research. Social scientists want to understand determinants of organizational longevity for various practical and theoretical reasons (see the review by Carroll 1984). To this end, they collect life histories on samples (or whole populations) of organizations and estimate the effects of various covariates on disbanding rates. They tend to find, for example, that age has a powerful effect on disbanding rates for many kinds of organizations. The reasons advanced to explain this fact typically adduce explanations at the level of the individual organization, such as collective learning or development of trust within work teams (Stinchcombe 1965).
Recent analyses of mortality processes from a population ecology per- spective also specify dependence on age and use many of the standard covariates. Does the structure of these analyses imply that the unit of analysis is really the individual organization? Admittedly population ecologists seek to understand demographic regularities in the rates at which vital events occur to organizations. Yet our interest does not lie in explaining organizational mortality per se. Rather we want to understand the dynamics of organizational diversity, how social changes affect the mix of organizations in society and vice versa. Analyses of mortality patterns in populations provide only one ingredient for an explanation of the dynamics of diversity. Information about mortality patterns must be combined with similar information about other vital rates in order to make valid inferences about diversity.
Diversity of forms is a property of a population or a community of organizations. In seeking to explain the dynamics of diversity, population ecologists develop propositions about processes holding at the populati level. It is in this sense that the population is the unit of analysis.
Explanation of organizational variability and change with re erence o selection processes is confusing to some because it seems to turn the usua logic of causal analysis on its head. Causal analysis at the organ,zabona level examines whether the presence (or level) of some organizational characteristic (such as size) affects the likelihood that some »‘her charac- teristic (or level) of another variable (such as structural differentiation) will occur. Selection analysis explores whether the prevalence of orgamza- tional characteristics and their joint distributions are governed by their links to rates of foundings, disbandings, mergers, and changes m forms In other words, selection analysis asks whether these vital rates contro the relation between the distribution of attributes at one time and that at a later time.
In the simplest case, an organizational characteristic may affect one or more of the rates. Such variables are treated in the statistical analysis as the causal factors affecting the rates. Yet at the same time they are outcomes to be explained—the fact that an organizational characteristic a fects the rates explains variations in the prevalence of that characteristic. Ecologists are generally less interested in mortality processes per se than in their implications for distributions of organizational characteristics.
Reference to selection processes can also explain the preva ence o characteristics that do not affect vital rates directly but are associated with characteristics that do. Consider Stinchcombe’s (1965) observation that organizational forms tend to incorporate and retain packages of characteristics that are common when the form emerges. Presumably only some elements in the package affect the vital rates. Analysts attempting to uncover static causal relations among the elements in the package may be badly misled if they do not realize that the items covary because they are linked in a selection process. If Stinchcombe is right, the covana ton of organizational characteristics may often reflect the vagaries of selection processes rather than causal relations among the variables themselves. That is a set of structural characteristics may covary today only beca they happened to be associated at an earlier time with a character.stic that confers strong survival advantages. In this sense, selection analysis can be a useful supplement for—and sometimes corrective to—static causal ex-planations at the level of individual organizations.
Source: Hannan Michael T., Freeman John (1993), Organizational Ecology, Harvard University Press; Reprint edition.