From Association to Causation via a Potential Outcomes Approach

Despite the importance of causal analysis in building a valid knowledge base and in answering managerial questions, the issue of causality rarely receives the attention it deserves in information systems (IS) and management research that uses observational data. In this paper, we discuss a potential outcomes framework for estimating causal effects and illustrate the application of the framework in the context of a phenomenon that is also of substantive interest to IS researchers. We use a matching technique based on propensity scores to estimate the causal effect of an MBA on information technology (IT) professionals' salary in the United States. We demonstrate the utility of this counterfactual or potential outcomes–based framework in providing an estimate of the sensitivity of the estimated causal effects because of selection on unobservables. We also discuss issues related to the heterogeneity of treatment effects that typically do not receive as much attention in alternative methods of estimation, and show how the potential outcomes approach can provide several new insights into who benefits the most from the interventions and treatments that are likely to be of interest to IS researchers. We discuss the usefulness of the matching technique in IS and management research and provide directions to move from establishing association to assessing causation.

• 저널Information Systems Research · 20(2) · 295–313
• 토픽Advanced Causal Inference Techniques · Statistics and Probability
• DOI10.1287/isre.1080.0184
• 저자Sunil Mithas, Mayuram S. Krishnan