docs(methodology): Deaner & Ku (2026) causal duration DiD paper review (PR-A)

ROADMAP.md

@@ -16,6 +16,10 @@ Queued work, ordered by expected leverage. Each item is its own PR. Ordering is
 
 - **Context-aware `practitioner_next_steps()`.** Substitutes actual column names from fitted results instead of generic placeholders, so next-step guidance is executable rather than illustrative. (Standalone follow-up to the `BusinessReport` / `DiagnosticReport` layer; tracked under the AI-Agent Track too.)
 
+### New estimators
+
+- **`DurationDiD` - causal duration DiD for binary absorbing-state outcomes** (Deaner & Ku, arXiv:2405.05220v2, 2026; revise-and-resubmit at *Quantitative Economics*). Parallel trends in mean outcomes generally fails when the outcome is an absorbing state (the paper's wording, pp. 1-2; exception: identical group means, footnote 1); the method imposes linear restrictions on counterfactual hazard rates (common dynamics / proportional hazards, Section 2) and avoids parametric hazard-model estimation (p. 3). Maintainer-approved first scope (2026-09): two-group, common timing, both specifications, individual block-bootstrap inference with pointwise and uniform bands, pre-treatment diagnostics; covariate adjustment, K-group restrictions, censoring input and staggered adoption deferred to follow-on PRs. Methodology review on file: `docs/methodology/papers/deaner-ku-2026-review.md`.
+
 ### Practitioner tutorials
 
 - **dCDH comprehensive tutorial.** One notebook covering reversible treatment, dynamic event study, covariates, trends, HonestDiD on placebos, and survey. Favara-Imbs (2015) banking-deregulation replication as the headline application.
@@ -81,12 +85,6 @@ Long-running program, framed as "building toward" rather than with discrete ship
 
 Frontier methods that may graduate to Under Consideration given time and research signals.
 
-### Causal Duration Analysis with DiD
-
-Extends DiD to duration / survival outcomes where standard methods fail (hazard rates, time-to-event). Duration analogue of parallel trends; avoids distributional and hazard-function assumptions.
-
-**Reference**: Deaner & Ku (2025), *AEA Conference Paper*.
-
 ### CATT Meta-Learner for Heterogeneous Effects
 
 ML-powered conditional ATT, using a doubly robust meta-learner to discover which units benefit most from treatment.

changelog.d/20260905-duration-did-paper-review.md

@@ -0,0 +1,21 @@
+### Documentation
+- **Deaner & Ku (2026) "Causal Duration Analysis with Diff-in-Diff" paper review on
+  file (PR-A).** Added `docs/methodology/papers/deaner-ku-2026-review.md`, a
+  paper-sourced fidelity review of the arXiv preprint 2405.05220v2 (SHA-256 pinned;
+  revise-and-resubmit at *Quantitative Economics* per the authors' pages) - the Step-1
+  artifact for a prospective `DurationDiD` estimator for binary absorbing-state
+  outcomes. Transcribes the absorbing-state and no-anticipation assumptions, the
+  common-dynamics / proportional-hazards / general-linear restrictions on counterfactual
+  hazards (Eqs. 2.1-2.5), the negative-log-survival identification results (Theorems
+  1-3, Remarks 1-4), the plug-in estimators (Eqs. 3.1-3.8), the individual block
+  bootstrap with pointwise and uniform bands and the pre-treatment specification test
+  (Algorithms 1-2), the covariate-balancing, staggered-adoption and semiparametric
+  extensions (Appendices A.2-A.4), the Appendix C simulation design with Table 2 in
+  full as a future test oracle, the Austrian unemployment-insurance application
+  numbers, and the proofs (Appendix E). Records the unlicensed Stata / MATLAB reference
+  implementation (commit `202e92ef`; black-box reference only, no source port) with an
+  eleven-item paper-versus-code fidelity list, and states the maintainer-approved
+  first-estimator scope (two-group, common timing, both specifications, bootstrap
+  inference, pre-treatment diagnostics; covariate adjustment and staggered adoption
+  deferred). Moves the ROADMAP entry to Shipping Next. Docs-only; no code change. This
+  is a deliberate exception to the published-source rule.