| Unit | Mean post-period gap | Max abs. gap |
|---|---|---|
| XLK (treated) | +20.40% | 31.56% |
| XLE | +12.17% | 30.73% |
| XLF | +11.13% | 15.41% |
| XLI | +11.07% | 17.86% |
| XLY | -10.15% | 17.04% |
| XLP | -6.30% | 8.49% |
| XLU | -6.07% | 15.69% |
| XLB | -6.04% | 13.14% |
| XLC | -6.03% | 12.41% |
| XLV | -0.87% | 11.72% |
| XLRE | -0.09% | 3.71% |
Caution on fit quality: the fit-window RMSE is 0.60× this unit's own fit-window standard deviation, a loose fit. Treat the gap below as suggestive at best, not a clean read.
XLK's post-period gap ranks #1 of 11 units by absolute size in the placebo test (larger rank = more likely a real, unit-specific effect rather than noise the method would produce for any random unit).
Caution: the confound-window gap (+13.17) is already a substantial fraction of the post-period gap (+34.35); this divergence looks like it was emerging before the treatment date, so the treatment can't cleanly take full credit for the post-period gap.
Read: this looks like a real, unit-specific effect; the gap this method finds for the treated unit is larger than what it produces when re-run on most other units in the donor pool.