In Simple Words
Imagine a medical trial where hospital patients are assigned a new doctor every six months. If a patient switches from a doctor participating in a new drug trial to a non-participating doctor, the patient's exposure to the experimental treatment gets interrupted, requiring researchers to track doctor-switching events closely.
Swiss jobseekers in the SAPS trial
- Caseworker has no access
- 12% matched the top recommendation
- Caseworker has access
- 12% matched the top recommendation
2005 trial, excluding Geneva. First programme within 90 days. The same rounded rates held when caseworker switchers were excluded.
See source →Research question
Public employment agencies often enforce mandatory caseworker rotation policies after a jobseeker reaches a specific threshold of unemployment duration to introduce fresh job search perspectives. In field experiments randomized at the caseworker level, this causes long-term jobseekers to rotate between treatment and control staff, potentially contaminating experimental estimation.
Participants
N = 22,758 stock jobseekers and N = 12,632 flow jobseekers across pilot Swiss employment offices (excluding Geneva).
Study Design
Sensitivity analysis comparing full intent-to-treat sample against a time-constant non-switcher subsample (excluding 2,263 jobseekers whose caseworker status changed between treatment and control).
Results
Caseworker switching occurred for 7.1% of stock jobseekers and 5.1% of flow jobseekers; empirical estimates of compliance and software impact remained identical with or without switchers (12% Definition 1 compliance across both samples).
Takeaway
Define experimental treatment status using initial caseworker assignment while running sensitivity robustness checks on non-switcher cohorts to verify that administrative rotation rules do not bias impact estimates.