FM3 Research
Graded against the actual result across 25 races (from 31 polls, through 2022).
Head-to-head vs VotePredictor Elections
The fair, apples-to-apples test: on the 16 races FM3 Research actually polled, how its final poll's margin compared to what VotePredictor Elections predicted for those same races.
| Model | Avg miss (pts) | Called right |
|---|---|---|
| FM3 Research | 6.28 | 69% |
| VotePredictor Elections | 1.96 | 94% |
VotePredictor Elections aggregates all the pollsters, so it's expected to beat any single one on margin — that's the value of averaging. The honest comparison among forecasters is on the combined board.
Every race (23)
Each race FM3 Research polled, scored on its final poll — the call right before the vote — against the actual Dem−Rep result. Click a race for its full detail.
| Race | Their call | Result | Miss | Call |
|---|---|---|---|---|
| 2022 OR Governor | D+2.0 | D+3.4 | 1.4 | ✓ |
| 2020 NE-2 House | D+2.0 | R+4.6 | 6.6 | ✗ |
| 2020 N2 President | D+11.0 | D+6.5 | 4.5 | ✓ |
| 2020 AL Senate | D+1.0 | R+20.4 | 21.4 | ✗ |
| 2018 WA-5 House | R+3.0 | R+9.5 | 6.5 | ✓ |
| 2014 UT-4 House | R+3.0 | R+5.1 | 2.1 | ✓ |
| 2012 CA-21 House | R+4.0 | R+15.5 | 11.5 | ✓ |
| 2012 WI-7 House | R+3.0 | R+12.3 | 9.3 | ✓ |
| 2010 WI Governor | EVEN | R+5.8 | 5.8 | ✗ |
| 2010 WA-8 House | R+4.0 | R+4.1 | 0.1 | ✓ |
| 2010 WI Senate | EVEN | R+4.8 | 4.8 | ✗ |
| 2006 WI-8 House | D+8.0 | D+2.1 | 5.9 | ✓ |
| 2004 CO-3 House | D+17.0 | D+4.0 | 13.0 | ✓ |
| 2004 CO President | EVEN | R+4.7 | 4.7 | ✗ |
| 2004 CO Senate | D+5.0 | D+4.8 | 0.2 | ✓ |
| 2004 WA Senate | D+15.0 | D+12.2 | 2.8 | ✓ |
| 2002 TX Governor | R+3.0 | R+17.8 | 14.8 | ✓ |
| 2002 TX Senate | D+1.0 | R+12.0 | 13.0 | ✗ |
| 2000 CA-20 House | D+11.0 | D+6.9 | 4.1 | ✓ |
| 2000 CA President | D+13.0 | D+11.8 | 1.2 | ✓ |
| 1998 NV Senate | D+5.0 | D+0.1 | 4.9 | ✓ |
| 1998 WA Senate | D+13.0 | D+16.8 | 3.8 | ✓ |
| 1998 WI Senate | D+1.0 | D+2.1 | 1.1 | ✓ |
Accuracy by time to election
Lower is better. Time to election runs right (election week) to left (~2 months out).
By the numbers
| Time to election | Polls | Avg miss | vs field | Called right |
|---|---|---|---|---|
| 1–3 wk | 8 | 6.56 | +1.49 | 75% |
| 3–6 wk | 18 | 7.64 | +1.90 | 67% |
| 6–9 wk | 5 | 6.08 | -0.07 | 100% |
vs field is this pollster's average miss minus all pollsters' at the same lead time — green beats the field, redtrails it. Our historical polls reach ~2 months out; earlier polling isn't in the record.
Track record by cycle — getting better?
| Year | Polls | Avg miss | Lean (house effect) |
|---|---|---|---|
| 1998 | 4 | 4.2 | D+1.7 |
| 2002 | 5 | 14.5 | D+14.5 |
| 2004 | 5 | 4.3 | D+4.3 |
| 2010 | 4 | 3.6 | D+3.6 |
| 2012 | 3 | 8.4 | D+10.4 |
| 2020 | 3 | 10.8 | D+10.8 |
Do we credit a pollster for fixing its bias? Each cycle, the model re-estimates every pollster's lean from all its earlier polls (walk-forward) and subtracts it before using the poll. We tested weighting recent cycles more — it doesn't help: a pollster's lean in one cycle barely predicts the next (correlation 0.28), so the swings above are mostly noise, and averaging over more history beats chasing the latest cycle. The all-time estimate we use came out within ~0.5% of the best option.