Pre-Conference Talk by HOANG Ai Phuong Emmy | Extending Propensity Score Matching to Capture Censored Observations for Causal Explanation

Causal inference based on observational data has become a key pillar in empirical studies, when natural experimental research designs are not available. Similar to data used in the Marketing, Labor Economics and Medical Epidemiology disciplines, the data used in IS research is prone to being unavailable and unrecoverable. Historical data for consumers, employers, patients, and users are rarely available in complete form. Left-censoring arises when the events of interest occurred before the study period; right-censoring refers to events that might or might not have occurred after the period of observation ended. Censored data create a roadblock for establishing a solid foundation for causal inference. This should not be viewed as a shortcoming but as an opportunity for a methodological advance to support the causal explanation. We propose a reconsideration and repurposing of propensity score matching (PSM) to address data censoring so observations can be preserved, when the events of interest recur during the study period. Thus, it will be possible to improve the completeness of the observational data for causal inference. We demonstrate the use of this approach to recover censored cable TV viewing and purchasing activities for series dramas. We briefly discuss on how this method can be evaluated based on (1) the extent to which it can recover unobserved data; and (2) the value of information gained, on the basis of the Bayesian inference framework.

This is a pre-conference talk for the Workshop on Information Systems and Economics (WISE 2017) in Seoul, Korea. The proposed method will also be discussed in the 38th International Conference on Information Systems (ICIS 2017) Doctoral Consortium.