Double Selection in Causal Inference

Christian Aleman, Christopher Busch, Santaeulà lia-Llopis, Raül and Tanya Surovtseva

No 22017, CEPR Discussion Papers from Centre for Economic Policy Research

Abstract: We introduce a unified double selection framework accommodating group-specific additive and multiplicative shifters of an unobserved common time trend (CTT). Under double selection, the CTT does not manifest as a parallel trend; the resulting non-constant selection bias is a linear combination nesting standard difference-in-differences (DiD) in levels and logs. Opposing shifters can yield patterns structurally impossible under standard DiDs, such as a ranking reversal of the groups over time. We establish that in the presence of double selection bias treatment effects are non-parametrically identified by the change of the ratio of within-group outcome changes over time. The double selection structure induces a stable linear relationship in the outcome-on-outcome space, grounding a simple two-step estimator operating in levels. Revisiting Card (1992), double selection uncovers valid counterfactuals where standard DiD fails.

Keywords: Double Selection; Difference-in-differences; Causal inference; Nonparametric identification (search for similar items in EconPapers)
JEL-codes: C14 C21 J38 (search for similar items in EconPapers)
Date: 2026-10
References: Add references at CitEc
Citations:

Downloads: (external link)
https://cepr.org/publications/DP22017 (application/pdf)

Related works:
This item may be available elsewhere in EconPapers: Search for items with the same title.

Export reference: BibTeX RIS (EndNote, ProCite, RefMan) HTML/Text

Persistent link: https://EconPapers.repec.org/RePEc:cpr:ceprdp:22017

Ordering information: This working paper can be ordered from
https://cepr.org/publications/DP22017

Access Statistics for this paper

More papers in CEPR Discussion Papers from Centre for Economic Policy Research 33 Great Sutton Street, London EC1V 0DX, UK.
Bibliographic data for series maintained by CEPR ().