Interpretation
TROP’s time-localization factors are exponential in distance to the center of the post-treatment block. Normalizing the pre-period factors permits a visual comparison with SDID’s simplex weights. Uniform or increasing TROP profiles are imposed by that kernel family; its low-rank adjustment also contributes to the treatment estimate. These factors are not the full implicit outcome weights of the adjusted estimator.
All samples are the same 27 previously selected SDID examples. No new paper, outcome, donor restriction or time window was chosen based on the TROP results. The second-tranche sample-retention caveats and original design labels remain.
Findings
11 selected TROP profiles are uniform and 16 increase toward treatment. 9 select the shortest half-life offered by the grid, so their chosen decay should be read as a boundary result. 11 retain a nonzero low-rank adjustment. All 27 final fits meet the relative primal–dual gap target of 10⁻⁸.
Median weight in the final fifth of pre-periods is 51.6% for TROP and 69.7% for SDID defaults, across these 27 preselected examples. The selection is exploratory and not representative of the full study corpus.
Coordinate tuning reached a fixed triplet in 27/27 cases. 0 papers have at least one numerically invalid candidate; the candidate/fold exports identify these explicitly. There are 10 complete-panel author-code comparisons and 16 independent weighted-DiD identity checks.
Weight profiles
The main SDID series is always the default. Its optimized reference is added only when the preceding display rule finds a meaningful discrepancy: maximum period change above 0.001 or total variation above 0.005. TROP is always shown separately. Every zero and small weight is retained in the downloads.
Tuning and estimates
The reported TROP effects are in each outcome’s original units, on the retained sample. No confidence intervals are computed. The three-fold grid search is described in the numerical analysis, including scaling, coordinate updates and solver diagnostics. Changing its grid or folds can change the selected kernel.
| 0 |
ARRA and voting |
0.120563 |
0.000000 |
inf |
0.0 |
inf |
0 |
0.000000e+00 |
| 1 |
Birth of a Nation |
-0.214987 |
0.308065 |
2.25 |
0.5 |
2.110219 |
20 |
4.470972e-09 |
| 2 |
Cap-and-trade |
-0.032873 |
0.866434 |
0.80 |
0.0 |
inf |
0 |
0.000000e+00 |
| 3 |
Collectivization |
-0.097463 |
0.000000 |
inf |
0.0 |
inf |
0 |
0.000000e+00 |
| 4 |
Credit Markets |
0.205342 |
0.000000 |
inf |
0.5 |
24.021653 |
3 |
3.626482e-09 |
| 5 |
Curriculum and Ideology |
0.854635 |
0.000000 |
inf |
2.0 |
0.398524 |
2 |
8.335255e-10 |
| 6 |
E-governance rollout |
-0.012702 |
3.465736 |
0.20 |
0.0 |
inf |
0 |
0.000000e+00 |
| 7 |
Fiscal capacity (mixed-design) |
0.018691 |
0.000000 |
inf |
0.0 |
inf |
0 |
0.000000e+00 |
| 8 |
Government audits (randomized) |
0.018170 |
0.000000 |
inf |
0.0 |
1.426084 |
11 |
5.266517e-09 |
| 9 |
Hospitals / Gift of a Lifetime |
0.463049 |
0.000000 |
inf |
0.0 |
inf |
0 |
0.000000e+00 |
| 10 |
Ideology and public organizations |
-0.012473 |
0.216608 |
3.20 |
0.0 |
51.135885 |
15 |
7.178571e-09 |
| 11 |
Italian opera |
-0.368293 |
0.247553 |
2.80 |
2.0 |
inf |
0 |
0.000000e+00 |
| 12 |
Market integration |
-0.020448 |
0.000000 |
inf |
0.0 |
inf |
0 |
0.000000e+00 |
| 13 |
Medical implants |
-0.148706 |
2.772589 |
0.25 |
0.0 |
inf |
0 |
0.000000e+00 |
| 14 |
Nigerian oil |
-0.165624 |
0.577623 |
1.20 |
0.0 |
0.990919 |
6 |
5.732746e-09 |
| 15 |
Powerful politicians |
0.075790 |
0.729629 |
0.95 |
0.0 |
0.452652 |
6 |
7.114257e-09 |
| 16 |
Powerful politicians: reply |
-0.019443 |
0.034657 |
20.00 |
0.0 |
0.558158 |
25 |
6.641581e-09 |
| 17 |
Procurement auctions |
1.709056 |
0.231049 |
3.00 |
2.0 |
0.841325 |
1 |
1.302438e-09 |
| 18 |
Rebel on the Canal |
0.095199 |
0.000000 |
inf |
0.0 |
inf |
0 |
0.000000e+00 |
| 19 |
Religion and conflict |
0.160881 |
6.931472 |
0.10 |
0.0 |
inf |
0 |
0.000000e+00 |
| 20 |
Retail pharmacies |
0.043035 |
0.338121 |
2.05 |
2.0 |
2.662378 |
15 |
6.205204e-09 |
| 21 |
Right to counsel |
0.066787 |
0.094306 |
7.35 |
0.0 |
inf |
0 |
0.000000e+00 |
| 22 |
School construction |
0.386783 |
6.931472 |
0.10 |
0.0 |
inf |
0 |
0.000000e+00 |
| 23 |
Trading nontradables |
0.014468 |
0.346574 |
2.00 |
2.0 |
0.028211 |
2 |
9.035623e-09 |
| 24 |
Vertical integration |
-0.140087 |
0.000000 |
inf |
2.0 |
inf |
0 |
0.000000e+00 |
| 25 |
Village officials |
0.384284 |
0.000000 |
inf |
2.0 |
inf |
0 |
0.000000e+00 |
| 26 |
Women's political voice |
0.007241 |
0.385082 |
1.80 |
2.0 |
inf |
0 |
0.000000e+00 |
Downloads
Original results, the original 22-paper collection, eligibility decisions and the earlier SDID files remain unchanged.
