Published: September 2017
This report reviews, summarises and gives guidance for the choice and application of causal inference techniques to ex-post evaluation of the impacts of interventions in the transport sector.
Such techniques seek to establish robust counterfactual outcomes against which post-intervention outcomes can be compared, while accounting for the effects of other factors aside from the intervention itself.
We illustrate the application of key techniques using two New Zealand case studies, and give recommendations to improve the robustness of ex-post evaluations.
Keywords: causal inference, cross-sectional regression, difference-in-differences, empirical Bayes, ex-post analysis, instrumental variables (IV) regression, interrupted time series (ITS), panel regression, propensity score matching (PSM), regression discontinuity (RD)