Beschreibung:
<jats:p> Theories that relate organization-level variables to one another frequently contain just three variables. “Formalization is negatively associated with success in a turbulent environment” and “Technology is a determinant of span of control at lower organizational levels” are examples. Theories limited to two or three variables tend to have low predictive validity and consequently are of limited use to anyone attempting to predict or interpret relationships among organizational variables. One purpose of the study reported here was to develop a more encompassing and more valid theory about a specific relationship—the relationship between centralization and effectiveness. </jats:p><jats:p> A second purpose of the study was to set forth and test the efficacy of a general approach for developing more encompassing and more valid theories about organizations. The successful application of this approach resulted in a six-variable theory: the relationship between the two variables centralization and effectiveness is a function of (1) the average size of the units of analysis, (2) the effectiveness subconstruct considered, (3) the extent of professionalization in the organizations, and (4) whether the organizations produce primarily goods or services. Specifically, 57% of the variance in the linear association between centralization and effectiveness was shown to be explainable by a linear function of the four variables just noted. This theory enables identification of the conditions where centralization will be related to effectiveness most positively (i.e. r = 0.49) and the conditions where it will be related to effectiveness most negatively (i.e., r = −0.56). </jats:p><jats:p> Besides these four predictor variables, four others were tested as possible predictors of the centralization-effectiveness relationship, but were rejected. Rejection of two of these latter four variables contradicted commonly accepted beliefs—decentralization was not more positively related to effectiveness in turbulent environments and decentralization was not more positively related to effectiveness in larger organizations. </jats:p><jats:p> Two concerns regarding the organization science literature resulted from this research. The first follows from our observations that most reports of studies include dysfunctionally sparse descriptions of the organizational contexts in which the data were collected and that in many studies the magnitudes of both the variables of primary interest and also the contextual variables are reported on coarse scales (e.g., on nominal scales with few levels, such as “small” and “large”). As a consequence of these two practices, researchers synthesizing the literature and developing theories from it—as was done in the research reported here—necessarily group together studies that should be distinguished from one another on the basis of their attributes. This unwanted grouping results in theories that are less predictive than they could be. </jats:p><jats:p> The second concern is that the subset of the organization science literature dealing with relationships between organizational design variables (such as centralization) and organizational effectiveness contains few studies demonstrating the causal directions of observed relationships. As a result, because various levels of effectiveness may lead to the adoption or evolution of different levels of “design variables,” or may be related to certain levels of a design variable because both are related to a third variable researchers and administrators observing high correlations between a design variable and effectiveness must be cautious in inferring that effectiveness can be enhanced by changing, the level of the respective design variable. </jats:p>