{"paper":{"title":"Estimating Dynamic Marginal Policy Effects under Sequential Unconfoundedness","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"Dynamic marginal policy effects can be identified via tractable reduced-form expressions and estimated with a doubly robust estimator under sequential unconfoundedness.","cross_cats":[],"primary_cat":"stat.ME","authors_text":"I-han Lai, Stefan Wager","submitted_at":"2026-04-07T09:41:11Z","abstract_excerpt":"We develop methods for estimating how infinitesimal policy changes affect long-term outcomes in dynamic systems. We show that dynamic marginal policy effects (MPEs) can be identified via tractable reduced-form expressions, and can be estimated under a general sequential unconfoundedness assumption. We also propose a doubly robust estimator for dynamic MPEs. Our approach does not require observing full dynamic state information (as is typically assumed for off-policy evaluation in Markov decision processes), and does not incur an exponential curse of horizon (as is typical in non-Markovian off-"},"claims":{"count":4,"items":[{"kind":"strongest_claim","text":"Dynamic marginal policy effects can be identified via tractable reduced-form expressions and estimated under sequential unconfoundedness with a doubly robust estimator that does not require full dynamic state information and avoids the exponential curse of horizon.","source":"verdict.strongest_claim","status":"machine_extracted","claim_id":"C1","attestation":"unclaimed"},{"kind":"weakest_assumption","text":"The sequential unconfoundedness assumption holds, allowing identification of dynamic MPEs from observed data without full state information.","source":"verdict.weakest_assumption","status":"machine_extracted","claim_id":"C2","attestation":"unclaimed"},{"kind":"one_line_summary","text":"Develops tractable reduced-form identification and a doubly robust estimator for dynamic marginal policy effects that avoids full state observation and exponential horizon curse.","source":"verdict.one_line_summary","status":"machine_extracted","claim_id":"C3","attestation":"unclaimed"},{"kind":"headline","text":"Dynamic marginal policy effects can be identified via tractable reduced-form expressions and estimated with a doubly robust estimator under sequential unconfoundedness.","source":"verdict.pith_extraction.headline","status":"machine_extracted","claim_id":"C4","attestation":"unclaimed"}],"snapshot_sha256":"3e436a7e97964c73dc7e47e61b3127a842b338d6ca077972cb7439234320f98d"},"source":{"id":"2604.05639","kind":"arxiv","version":2},"verdict":{"id":"681906c9-af77-403c-94e6-62efdbd845e3","model_set":{"reader":"grok-4.3"},"created_at":"2026-05-19T16:52:56.681214Z","strongest_claim":"Dynamic marginal policy effects can be identified via tractable reduced-form expressions and estimated under sequential unconfoundedness with a doubly robust estimator that does not require full dynamic state information and avoids the exponential curse of horizon.","one_line_summary":"Develops tractable reduced-form identification and a doubly robust estimator for dynamic marginal policy effects that avoids full state observation and exponential horizon curse.","pipeline_version":"pith-pipeline@v0.9.0","weakest_assumption":"The sequential unconfoundedness assumption holds, allowing identification of dynamic MPEs from observed data without full state information.","pith_extraction_headline":"Dynamic marginal policy effects can be identified via tractable reduced-form expressions and estimated with a doubly robust estimator under sequential unconfoundedness."},"integrity":{"clean":true,"summary":{"advisory":0,"critical":0,"by_detector":{},"informational":0},"endpoint":"/pith/2604.05639/integrity.json","findings":[],"available":true,"detectors_run":[],"snapshot_sha256":"c28c3603d3b5d939e8dc4c7e95fa8dfce3d595e45f758748cecf8e644a296938"},"references":{"count":4,"sample":[{"doi":"","year":null,"title":"Non-parametric causal inference in dynamic thresholding designs","work_id":"4777ba7d-54fe-409a-a59c-3f02af73af28","ref_index":1,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"","year":null,"title":"Switchback experiments under geometric mixing.arXiv preprint arXiv:2209.00197","work_id":"422299fc-5ed6-4ea3-9a5d-9ab055c3c775","ref_index":2,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"","year":null,"title":"Estimation of treatment effects under nonstation- arity via the truncated policy gradient estimator.arXiv preprint arXiv:2506.05308","work_id":"5746d186-1e2b-48a1-b9d9-9130141f891c","ref_index":3,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"10.1111/biom.13859","year":null,"title":"Yuya Sasaki and Takuya Ura","work_id":"74609e00-e75a-4c40-ae97-7ebda99c1031","ref_index":4,"cited_arxiv_id":"","is_internal_anchor":false}],"resolved_work":4,"snapshot_sha256":"1345672107af94ce575c84541f36fc6f971aaceeedbc8fc35bf6af7baa300010","internal_anchors":0},"formal_canon":{"evidence_count":2,"snapshot_sha256":"7dc1fff9362d38ca335f903cb12f4b72ef504d7c0f20006c8fc34dccd3e2a2b0"},"author_claims":{"count":0,"strong_count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"builder_version":"pith-number-builder-2026-05-17-v1"}