Bridging Disciplinary Boundaries (January 11 - 14, 2007)

Purpose: Observational studies, defined as empiric investigations of events and processes and their effects in which the investigator cannot control the assignment of treatments to subjects, face two major problems. First, because such studies lack experimental controls, they confront the counterfactual problem, which qualifies confidence in model specification and limits strong causal inference thereby undermining the authoritativeness of research conclusions. Second, conventional approaches for dealing with missing data, a ubiquitous problem for observational studies, have been criticized as potentially responsible for producing misleading or spurious research findings because of systematic bias in subsequent analyses.

The purpose of this study is to illustrate how these problems can be addressed using propensity matching score analysis to address the counterfactual problem, and multiple imputation methodology to address the missing data problem. The substantive focus is on how the transformation of human service organizations from nonprofit to for-profit status affects the subsequent survival chances of such organizations. Structural inertia theory provides the theoretical basis for the study, predicting that such transformations will not be adaptive but instead will increase the risk of organizational death.

Methods: The design is longitudinal using retrospective data from the Census Bureau's Longitudinal Business Database (LBD). The observation period covers 1970 to 1999. The initial study population numbered 1.5 million and was comprised of all organizations in the United States listed in the LBD that came into existence since1970 to provide social and rehabilitation services to persons with social or personal problems who require special services. Multiple imputation procedures were used to assign values when data were missing for founding date, thereby permitting the full calculation of organizational age. Subsequently, propensity matching score analysis was used to generate treatment and control groups; the former was comprised of organizations subject to the stimulus of transforming from nonprofit to for-profit status and numbered 2,385. The control group numbered 5,374, and was comprised of nonprofit organizations with matching propensity scores to organizations in the treatment group except they were not subject to the stimulus of status change. Analyses were carried out using logistic regression, tests on difference between means, and event history analysis.

Findings: The findings strongly support predictions from structural inertia theory. Compared to organizations in the control group, organizations in the treatment group had significantly higher death rates. Furthermore, this significantly higher propensity to die persisted over the remaining life span of the organizations in the treatment group.

Implications: The macro-organization theory paradigm, which originated in sociology and economics and gives the theoretical context for structural inertia theory, has been criticized for producing neither authoritative nor socially relevant findings. Through the use of propensity score matching and multiple imputation procedures on a national sample of organizations, this research shows that it is possible to address the authoritative findings criticism. At the same time, by focusing on human services, its potential for producing socially relevant analysis is illustrated in terms of significant policy questions it raises about the effects on vulnerable populations of government-induced change processes such as transformation and death.