Advanced level · Module 1: Causal Inference and Programme Evaluation · Lesson 4 of 4
Other Designs and Sensitivity
Recognise matching and regression discontinuity designs and test how sensitive conclusions are to assumptions.
In this lesson you will
- Describe matching and regression discontinuity designs
- Carry out simple sensitivity analyses
Key concepts
propensity score matchingregression discontinuitysensitivity analysisexternal validity
The lesson
Matching compares participants with similar non-participants on observed characteristics; it cannot fix unobserved differences. Regression discontinuity uses eligibility cut-offs — such as a vulnerability score — comparing units just above and below.
Sensitivity analysis asks how conclusions change under different assumptions, comparison groups or specifications. Robust findings survive reasonable changes.
Even a credible estimate applies to the setting studied; results from one county may not transfer to another (external validity).
Step by step
- Identify an eligibility cut-off suitable for regression discontinuity.
- Re-estimate the DiD with a different comparison group.
- Exclude one district and re-estimate.
- Summarise how stable the result is.
Practice activity
Write the sensitivity section of the feeding evaluation.
Expected result: A sensitivity analysis with a clear judgement of the result's robustness.
Check your understanding
-
What can matching not correct?
Show answer
Differences in unobserved characteristics.
-
What does regression discontinuity exploit?
Show answer
An eligibility cut-off, comparing units just either side of it.
Taught live by an instructor — online or in person. Free course lessons are for self-study and do not lead to a certificate — see registration options.