Lesson 2 · Quasi-Experimental Designs · Step 5 of 6
Regression discontinuity and HPV vaccination in Ontario
When a program switches on at a threshold, people just either side of it are nearly identical, so comparing them estimates the effect almost as a trial would. The design needs a running variable, a known cutoff and a jump in treatment there. Its diagnostics matter: no pile-up at the cutoff, no effect at placebo cutoffs, and covariates that stay smooth. Ontario's free school HPV vaccine started with girls born on or after January 1, 1994. Among 260,493 girls, eligibility caused no jump in pregnancy or other STIs (RR 0.96, 0.81 to 1.14).
Also covered: running variable and cutoff · sharp and fuzzy designs · manipulation checks · placebo cutoffs
Alongside this step
- Slides and transcript — the whole video
- Read the chapter — Quasi-Experimental Designs
Lesson 2 · 6 steps
1What is your identification strategy?6:192Comparing groups without baseline data4:193Difference-in-differences and maternal health vouchers in Bangladesh4:434Interrupted time series and the kidney transplant wait-time policy6:025Regression discontinuity and HPV vaccination in Ontario5:366Instrumental variables, and choosing a design8:51
Research Designs · lesson 2 of 5