Frontdoor Adjustment (Frontdoor Criterion)
Also known as: frontdoor criterion, Pearl's frontdoor adjustment, frontdoor formula, Ön Kapı Düzenlemesi (Frontdoor Adjustment)
Frontdoor adjustment is Judea Pearl's graphical identification strategy, introduced in 1995, that recovers the causal effect of a treatment on an outcome through a fully mediating variable even when an unobserved confounder sits between the treatment and the outcome. It is the go-to tool when the backdoor criterion cannot be satisfied because the confounder is unmeasured.
Read the full method
Sign in with a free account to read this section.
Method map
The neighbourhood of related methods — select a node to explore.
When to use it
Use frontdoor adjustment with observational data when you suspect an unobserved confounder between treatment and outcome, so the backdoor criterion is out of reach, but you can identify a variable that fully mediates the treatment's effect. The required conditions are strict: there must be no unobserved confounding on the treatment-to-mediator path, the mediator-to-outcome path must be free of treatment confounding (blocked by the backdoor over the treatment), the entire treatment effect must travel through the mediator (full mediation), and the frontdoor criterion must be verified on a credible DAG. A sample of at least about 100 observations is recommended for the two-stage estimation to be stable.
Strengths & limitations
- Identifies the causal effect even when the treatment-outcome confounder is unobserved, where the backdoor criterion fails.
- Rests on an explicit, checkable graphical criterion rather than an untestable ignorability assumption.
- Decomposes the problem into two confounder-free regressions that are straightforward to estimate and interpret.
- Demands a genuine full mediator that carries the entire treatment effect — a strong and often unrealistic structural requirement.
- Invalid if any unobserved confounder touches the treatment-to-mediator path or confounds the mediator-to-outcome relationship.
- The two-stage estimate becomes unstable in small samples (n < 100).
Frequently asked
How is frontdoor adjustment different from backdoor adjustment?
Backdoor adjustment blocks confounding by conditioning on observed common causes of treatment and outcome. Frontdoor adjustment is used when those confounders are unobserved: instead of conditioning on the confounder, it identifies the effect through a fully mediating variable, so it works precisely in cases where the backdoor criterion cannot be met.
What does the 'full mediation' assumption require?
It requires that the entire causal effect of the treatment on the outcome travels through the mediator, with no direct treatment-outcome path bypassing it. If any of the effect leaks around the mediator, the frontdoor formula no longer identifies the true effect.
How large a sample do I need?
At least about 100 observations are recommended. Below that the two-stage estimation underlying the frontdoor formula becomes unstable, and a DAG identification check or a different identification strategy is safer.
What if the assumptions fail?
If full mediation cannot be justified, the frontdoor criterion is invalid and you should consider a backdoor strategy or an instrumental-variables approach such as two-stage least squares. Always verify the criterion on a credible causal diagram first.
Sources
- Pearl, J. (1995). Causal Diagrams for Empirical Research. Biometrika, 82(4), 669-688. DOI: 10.1093/biomet/82.4.669 ↗
- Pearl, J. (2009). Causality: Models, Reasoning, and Inference (2nd ed.). Cambridge University Press. ISBN: 978-0521895606
How to cite this page
ScholarGate. (2026, June 1). Frontdoor Adjustment (Frontdoor Criterion). ScholarGate. https://scholargate.app/en/causal-inference/frontdoor-adjustment
Which method?
Set this method beside its closest kin and read them side by side — the library lays the books on the table; the choice is yours.
- Causal Discovery AlgorithmsCausal inference↔ compare
- DAG Causal IdentificationCausal inference↔ compare
- Regression DiscontinuityCausal inference↔ compare
- Two-Stage Least Squares (2SLS)Causal inference↔ compare