Multi-Method Approach
When endogeneity concerns are serious but neither credible instruments nor plausible copula assumptions are available, a pragmatic alternative is triangulation: combining observational analyses with controlled studies.
The Logic
The idea is straightforward: if a focal predictor cannot be cleanly identified in the observational data, manipulating it in a complementary experiment can help assess whether the observational conclusions hold up. Triangulation serves two purposes:
- Triangulation tests the robustness of the estimated effect
- Triangulation sharpens insight into the mechanisms behind it.
Important Limitation
Multimethod designs do not substitute for internal validity within the observational component. Assembling multiple data sources does not, by itself, eliminate endogeneity. The observational analysis must still address endogeneity rigorously — experimental add-ons cannot legitimize weak identification in the observational study.