The 2022 Debrief · post-mortem · permanent archive
How did the polling industry do this cycle?
The 2022 industry MAE landed at 4.6pt — graded across 426 historical races with a known outcome — and 438 polls were analyzed across 102 pollsters. Signed bias +0.35pt: polls landed close to the mean — no systematic over-predict on either side.
INDUSTRY MAE
4.6pt
vs midterm avg
SIGNED BIAS
+0.35pt
R-leaning
POLLS · POLLSTERS
438
102 firms
RACES SCORED
426
w/ known outcome
Industry mean absolute error · 2012–2022
PoliAgg model · self-grade · 2022
How did our own production model do this cycle? The model issues one prediction per race; the polling industry above is averaged across many polls per race. Metrics are not directly comparable but published side-by-side for transparency.
Mean error per race
7.00pp
poliagg-v8 ensemble vs final margin
Call accuracy
93.2%
fraction with right winner called
Races graded
426
outcomes used in the grade
Per-pollster scorecard · 2022 · sorted by MAE
Signed bias · ±4.2pt
MAE · 0–4.2pt
Biggest-miss races · 2022
| Race | Expected | Actual | Miss | Direction |
|---|---|---|---|---|
| us-senate-special-2022-ok | D+11.50 | D+26.53 | -15.0pt | R-margin underestimate |
| us-senate-2022-ok | D+18.00 | D+32.20 | -14.2pt | R-margin underestimate |
| us-governor-2022-ar | D+14.50 | D+27.76 | -13.3pt | R-margin underestimate |
| us-governor-2022-ok | D+2.00 | D+13.66 | -11.7pt | R-margin underestimate |
| us-senate-2022-wa | R+4.50 | R+14.52 | 10.0pt | D-margin underestimate |
| us-senate-2022-fl | D+7.50 | D+16.41 | -8.9pt | R-margin underestimate |
| us-governor-2022-fl | D+11.50 | D+19.40 | -7.9pt | R-margin underestimate |
| us-governor-2022-co | R+12.00 | R+19.35 | 7.3pt | D-margin underestimate |