Organizational Properties from Aggregate Data: Separating Individual and Structural Effects
论证聚合数据可作为组织属性的有效指标，并提出用Hauser的路径分析模型分离个体与组织层面的因果效应，通过对20家社会服务机构的调查数据验证了该方法能揭示不同层面的矛盾过程。
In this paper, we argue that aggregate measures (means computed on distributions of individual scores) may be valid indicators of organizational properties which also enable an investigator to determine whether statistical relations among organizational measures arise from organization-level, as opposed to individual-level, causal processes. We review certain statistical aggregation issues as these pertain to organizational analysis, and we propose Hauser's path analytic model of analysis of covariance as a device for separating individual and structural effects. These methods are then applied to data gathered in a survey of 20 social service organizations. We specify and estimate a causal model wherein administrative intensity, lateral communications, and decentralization of decision making are endogenous variables. We find that a number of total effects on these properties mask quite different-and in some cases contradictory-processes at individual and organization levels. Among other inferences, we suggest that the organizationaland individual-level influences we observe on decentralization raise questions regarding certain widely accepted interpretations of this property. We recommend that analysts working with aggregate measures adopt similar procedures in order to fully exploit their data for the insights that may be gained into multilevel organizational processes.