Causal Inference Techniques in Disaster Recovery Research: New Kids on the Corner

Disaster management has been claimed as being “evidenced-based” through rigorous disaster research practices. Evidence in the most scientific sense is defined as “a causal relation between a treatment and its outcome.” In real-life disaster research endeavors, however, researchers often encounter the situations where treatment and outcome variables are both affected by the confounding factors, which could lead to wrong conclusions about the causal relations. One such example is an issue of selection bias since it is almost practically and/or ethically impossible to assign subjects randomly into experimental or control groups. The other examples include the use of essentially correlational data and making inferences about the causations. This paper reviews recent use of new causal inference techniques in the disaster research field (e.g., long-term recovery) and advocates their wider utilization. Case examples are used to demonstrate such techniques as 1) adjusting selection bias using propensity score matching in the study that examined the effects of different types of temporary housing programs upon life (individual) recovery (Tatsuki, 2007) from the Great East Japan Earthquake (GEJE), 2) making causal inferences about the longitudinal impacts of critical life recovery facilitation factors over within-subject variabilities on life recovery scores using panel data analysis, and 3) identifying pre-disaster characteristics/conditions that enabled faster housing recovery among the GEJE impacted survivors using Kaplan-Meier survival curve analysis.