T23 SpilloverDiD tutorial: TVA-style ~40% understatement walkthrough

TODO.md

@@ -159,7 +159,6 @@ Deferred items from PR reviews that were not addressed before merge.
 | `SpilloverDiD(vcov_type="conley", conley_lag_cutoff > 0, survey_design=...)` no-effective-PSU serial Bartlett HAC. Wave E.2 follow-up ships the panel-block composition when an effective PSU exists (explicit `survey_design.psu` OR injected via `cluster=<col>` per `_inject_cluster_as_psu`). Weights-only / strata-only survey designs WITHOUT a cluster fallback raise `NotImplementedError` at `SpilloverDiD.fit` post-resolution because under the pseudo-PSU = obs-index fallback each pseudo-PSU appears in exactly one period — the per-PSU serial cross-period loop would silently contribute zero. Fix would either derive a unit-level serial fallback for no-PSU designs (mixes IF allocators with the pseudo-PSU spatial term — needs methodology work) or route the serial loop through `conley_unit` with explicit documentation of the IF-allocator asymmetry. Regression goldens vs the effective-PSU shipped path. | `spillover.py::SpilloverDiD.fit`, `two_stage.py::_compute_stratified_serial_bartlett_meat` | follow-up (Wave E.2 follow-up tail) | Low |
 | `SpilloverDiD(ring_method="count")` extension. Currently only the nearest-treated-ring specification is exposed. Count-of-treated-in-ring (paper Section 3.2 end) is methodologically supported by Butts but re-introduces functional-form dependence; expose with an explicit kwarg gate and documentation warning. | `spillover.py::SpilloverDiD.fit` | follow-up | Low |
 | `SpilloverDiD` data-driven `d_bar` selection (Butts 2021b / Butts 2023 JUE Insight cross-validation). | `spillover.py::SpilloverDiD` | follow-up | Low |
-| `SpilloverDiD` T22 TVA tutorial (`docs/tutorials/22_spillover_did.ipynb`): synthetic TVA-style DGP reproducing Butts (2021) Section 4 Table 1 Panel A bias-correction direction (~40% understatement). Split from the methodology PR per user-confirmed scope split (2026-05-15). | `docs/tutorials/`, `tests/test_t22_*_drift.py` | follow-up (Wave B) | Medium |
 | Extend `TwoStageDiD` with Conley vcov as a first-class feature (mirrors Wave A's TWFE/MPD/DiD extension). Currently `TwoStageDiD.__init__` lacks `vcov_type` / `conley_*` kwargs; `SpilloverDiD` works around this by threading Conley directly via `solve_ols` at stage 2. Promoting Conley to TwoStageDiD's API removes the workaround and lets non-spillover users access Conley + Gardner two-stage. | `diff_diff/two_stage.py` | follow-up | Medium |
 | `SpilloverDiD` sparse cKDTree path for the staggered nearest-treated-distance helper (mirrors the static helper's sparse branch). Currently `_compute_nearest_treated_distance_staggered` always builds dense `(n_units, n_treated_by_onset)` pairwise distance matrices per cohort; on large staggered panels with many cohorts this is avoidable memory/runtime. Add a sparse k-d-tree branch analogous to `_compute_nearest_treated_distance_sparse`, gated on `n > _CONLEY_SPARSE_N_THRESHOLD`. | `spillover.py::_compute_nearest_treated_distance_staggered` | follow-up (Wave B) | Low |
 | `SpilloverDiDResults` in `DiagnosticReport` dispatch tables. Wave C event-study emits a TwoStageDiD-compatible `event_study_effects: Dict[int, Dict]` alias that `plot_event_study` consumes via the new `reference_period` attribute fallback in `_extract_plot_data`, but `SpilloverDiDResults` is NOT registered in `DiagnosticReport`'s `_APPLICABILITY` / `_PT_METHOD` tables — so `DiagnosticReport(spillover_result)` doesn't currently route to event-study diagnostics. Registering requires (a) deciding which diagnostics apply (parallel trends, pre-trends power, heterogeneity, design-effect) AND (b) adding an end-to-end test. | `diff_diff/diagnostic_report.py::_APPLICABILITY`, `_PT_METHOD` | follow-up (Wave C) | Low |

docs/index.rst

@@ -83,6 +83,7 @@ Quick Links
    tutorials/20_had_brand_campaign
    tutorials/21_had_pretest_workflow
    tutorials/22_had_survey_design
+   tutorials/23_spillover_tva
 
 .. toctree::
    :maxdepth: 1

docs/methodology/papers/butts-2021-review.md

@@ -254,7 +254,7 @@ where `D^k_{it} := D_i * 1{K_it = k}` and `K_it` is years since treatment turned
   - Agriculture employment: standard DiD = -5.1% per decade. With spillovers controlled, total effect = -7.4% per decade. **Spillover bias = +2.3 pp; canonical DiD UNDERESTIMATED the agricultural decline by ~40%.** Spillover-on-control coefficients are negative (-3.7%, -1.6%, -3.0%, -1.6% per decade across the 4 distance bins) — consistent with farm-worker out-migration to higher-paying TVA manufacturing jobs.
   - Manufacturing employment: standard DiD = +5.6% per decade. With spillovers controlled, total effect = +3.5% per decade. **Spillover bias = +2.1 pp; canonical DiD OVERESTIMATED the manufacturing gain by ~40%.** Spillover coefficients are negative (-2.0%, -2.5%, -3.3%, -3.0%), consistent with "urban shadow" effects whereby firms relocate INTO the TVA from neighboring areas.
 - Interpretation (page 21): "the long-run spillovers cause the original estimates to be about 40 percent too small for agriculture employment and 40 percent too large for manufacturing employment."
-- Useful design choice for diff-diff T22 tutorial DGP: a TVA-style bias-correction percentage of ~40% is large enough to be visible without being implausible.
+- Useful design choice for diff-diff T23 tutorial DGP: a TVA-style bias-correction percentage of ~40% is large enough to be visible without being implausible. (Tutorial slot 22 went to `22_had_survey_design.ipynb`; SpilloverDiD landed in slot 23 — `docs/tutorials/23_spillover_tva.ipynb`.)
 
 ### Empirical illustration — Opportunity Zones (Appendix B)
 - Application: revisits Chen, Glaeser, and Wessel (2021) on the 2017 federal Opportunity Zone program's effect on home prices.

docs/references.rst

@@ -188,7 +188,11 @@ Multi-Period and Staggered Adoption
 
 - **Butts, K. (2023).** "Difference-in-Differences with Spatial Spillovers." *arXiv:2105.03737v3* (originally posted 2021). https://arxiv.org/abs/2105.03737
 
-  Identifies the ring-indicator estimator implemented in our ``SpilloverDiD`` class. Section 2-3 covers non-staggered timing (Equations 5/6/8); Section 5 covers staggered timing via two-stage Gardner (Table 2). Section 3.1 (page 13) recommends Conley spatial-HAC for inference with cutoff = ``d_bar``.
+  Identifies the ring-indicator estimator implemented in our ``SpilloverDiD`` class. Section 2-3 covers non-staggered timing (Equations 5/6/8); Section 5 covers staggered timing via two-stage Gardner (Table 2). Section 3.1 (page 13) recommends Conley spatial-HAC for inference with cutoff = ``d_bar``. Section 4 Table 1 Panel A (TVA agriculture) is the empirical anchor for tutorial ``docs/tutorials/23_spillover_tva.ipynb`` (~40% understatement direction).
+
+- **Kline, P., & Moretti, E. (2014).** "Local Economic Development, Agglomeration Economies, and the Big Push: 100 Years of Evidence from the Tennessee Valley Authority." *The Quarterly Journal of Economics*, 129(1), 275-331. https://doi.org/10.1093/qje/qjt034
+
+  Original Tennessee Valley Authority application that Butts (2021) §4 revisits to illustrate spillover bias in the canonical place-based-intervention DiD. Cited as the motivating real-world example in ``docs/tutorials/23_spillover_tva.ipynb``.
 
 - **Conley, T. G. (1999).** "GMM Estimation with Cross Sectional Dependence." *Journal of Econometrics*, 92(1), 1-45. https://doi.org/10.1016/S0304-4076(98)00084-0