VotePredictor
← All pollsters

Impact Research

Graded against the actual result across 75 races (from 97 polls, through 2024).

Races polled
75
Polls
97
Avg miss
6.42 pts
Most recent
2024

Head-to-head vs the VotePredictor model

The fair, apples-to-apples test: on the 75 races Impact Research actually polled, how its final poll's margin compared to what the VotePredictor model predicted for those same races.

ModelAvg miss (pts)Called right
Impact Research6.4765%
VotePredictor4.7181%

VotePredictor 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 (75)

Each race Impact 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.

RaceTheir callResultMissCall
2024 MT-1 HouseR+1.0R+7.76.7
2022 FL-28 HouseR+10.0R+27.417.4
2022 IA-3 HouseEVENR+0.70.7
2022 MT-1 HouseR+2.0R+3.21.2
2022 OH-1 HouseD+3.0D+5.52.5
2020 NC GovernorD+10.0D+4.55.5
2020 AR-2 HouseEVENR+10.710.7
2020 MI-3 HouseD+2.0R+5.97.9
2020 NJ-2 HouseD+4.0R+5.89.8
2020 MI PresidentD+8.0D+2.85.2
2020 NC PresidentD+3.0R+1.34.3
2020 NV PresidentD+4.0D+2.41.6
2020 NC SenateD+12.0R+1.713.7
2020 SC SenateEVENR+10.310.3
2019 LA GovernorD+16.0D+2.713.3
2019 NC-9 HouseEVENR+2.02.0
2018 IL GovernorD+16.0D+15.70.3
2018 MI GovernorD+11.0D+9.61.4
2018 SD GovernorD+3.0R+3.46.4
2018 FL-27 HouseD+10.0D+6.04.0
2018 MI-11 HouseD+10.0D+6.73.3
2018 MN-7 HouseD+18.0D+4.313.7
2018 TX-31 HouseR+4.0R+2.91.1
2017 GA-6 HouseD+1.0R+3.64.6
2017 SC-5 HouseR+10.0R+3.16.9
2016 CA-10 HouseD+5.0R+3.48.4
2016 FL-13 HouseD+11.0D+3.87.2
2016 MI-1 HouseD+1.0R+14.815.8
2016 UT-4 HouseR+10.0R+12.52.5
2016 VA-5 HouseR+6.0R+16.710.7
2016 OH PresidentD+2.0R+8.110.1
2016 OH SenateR+12.0R+20.98.9
2015 LA GovernorD+12.0D+12.20.2
2014 FL-2 HouseD+3.0D+1.11.9
2014 KS-2 HouseR+10.0R+18.48.4
2012 AZ-9 HouseD+5.0D+4.10.9
2012 FL-22 HouseD+10.0D+9.30.7
2012 IL-13 HouseD+9.0R+0.39.3
2012 IL-17 HouseD+1.0D+6.65.6
2012 IL-8 HouseD+15.0D+9.55.5
2012 NC-8 HouseD+1.0R+7.88.8
2012 OH-6 HouseD+6.0R+6.512.5
2012 PA-12 HouseD+11.0R+3.514.5
2012 TX-23 HouseD+5.0D+4.80.2
2012 AZ SenateD+4.0R+3.07.0
2012 IN SenateD+7.0D+5.81.2
2010 AR-1 HouseD+2.0R+8.310.3
2010 FL-22 HouseD+8.0R+8.716.7
2010 IA-3 HouseD+9.0D+4.24.8
2010 IL-11 HouseR+4.0R+14.710.7
2010 NM-2 HouseD+7.0R+10.817.8
2010 PA-12 HouseD+2.0D+7.75.7
2010 SC-2 HouseR+7.0R+9.72.7
2010 WV-3 HouseD+19.0D+12.16.9
2010 IL SenateD+2.0R+1.63.6
2008 AL-2 HouseD+1.0D+0.60.4
2008 AL-5 HouseD+8.0D+3.64.4
2008 AZ-3 HouseD+1.0R+12.013.0
2008 FL-10 HouseR+17.0R+21.44.4
2008 IL-11 HouseD+19.0D+23.94.9
2008 LA-6 HouseD+7.0D+2.94.1
2008 MS-1 HouseD+1.0D+7.66.6
2008 NE-2 HouseR+1.0R+3.92.9
2008 NH-2 HouseD+18.0D+15.03.0
2008 NV-3 HouseD+9.0D+5.13.9
2008 PA-4 HouseD+12.0D+11.70.3
2008 N2 PresidentR+4.0D+1.25.2
2008 NC SenateD+5.0D+8.53.5
2006 CO-3 HouseD+27.0D+25.02.0
2006 FL-22 HouseD+6.0D+3.82.2
2006 PA-4 HouseR+4.0D+3.97.9
2004 CO-3 HouseD+15.0D+4.011.0
2004 FL-13 HouseR+6.0R+10.64.6
2004 LA-3 HouseD+4.0D+0.53.5
2004 UT-2 HouseD+23.0D+11.511.5

Accuracy by time to election

24686–9 wk3–6 wk1–3 wk≤1 wkavg miss (pts)
Impact ResearchAll pollsters (field average)

Lower is better. Time to election runs right (election week) to left (~2 months out).

By the numbers

Time to electionPollsAvg missvs fieldCalled right
1–3 wk135.35+0.2862%
3–6 wk516.30+0.5675%
6–9 wk327.14+0.9959%

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?

YearPollsAvg missLean (house effect)
200447.6D+7.6
200643.7R+1.6
2008185.7D+1.3
2010138.7D+7.8
2012135.8D+4.9
201437.9D+7.9
2016810.1D+10.1
2018123.2D+2.1
2020117.9D+7.9
202254.8D+2.8

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.