Normington, Petts & Associates
Graded against the actual result across 49 races (from 52 polls, through 2020).
Head-to-head vs the VotePredictor model
The fair, apples-to-apples test: on the 47 races Normington, Petts & Associates actually polled, how its final poll's margin compared to what the VotePredictor model predicted for those same races.
| Model | Avg miss (pts) | Called right |
|---|---|---|
| Normington, Petts & Associates | 7.98 | 55% |
| VotePredictor | 5.53 | 79% |
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 (49)
Each race Normington, Petts & Associates 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 |
|---|---|---|---|---|
| 2020 CA-25 House | D+6.0 | R+0.1 | 6.1 | ✗ |
| 2020 MN-2 House | D+18.0 | D+2.3 | 15.7 | ✓ |
| 2020 MO-2 House | EVEN | R+6.4 | 6.4 | ✗ |
| 2018 CO-6 House | D+11.0 | D+11.2 | 0.2 | ✓ |
| 2018 VA-7 House | EVEN | D+1.9 | 1.9 | ✗ |
| 2016 IA-1 House | R+1.0 | R+7.7 | 6.7 | ✓ |
| 2016 IL-10 House | D+5.0 | D+5.2 | 0.2 | ✓ |
| 2016 IN-9 House | R+2.0 | R+13.7 | 11.7 | ✓ |
| 2016 ME-2 House | D+1.0 | R+9.6 | 10.6 | ✗ |
| 2016 NH-1 House | D+10.0 | D+1.3 | 8.7 | ✓ |
| 2016 IL President | D+17.0 | D+17.1 | 0.1 | ✓ |
| 2016 M2 President | D+4.0 | R+10.3 | 14.3 | ✗ |
| 2016 IL Senate | D+9.0 | D+15.1 | 6.1 | ✓ |
| 2014 ME Governor | D+2.0 | R+4.8 | 6.8 | ✗ |
| 2014 MA-6 House | D+12.0 | D+13.8 | 1.8 | ✓ |
| 2014 ME-2 House | D+8.0 | R+5.2 | 13.2 | ✗ |
| 2014 NH-1 House | D+2.0 | R+3.6 | 5.6 | ✗ |
| 2014 NH-2 House | D+6.0 | D+10.0 | 4.0 | ✓ |
| 2012 ME-2 House | D+25.0 | D+16.4 | 8.6 | ✓ |
| 2012 WI-8 House | R+6.0 | R+12.0 | 6.0 | ✓ |
| 2010 AZ-5 House | D+3.0 | R+8.8 | 11.8 | ✗ |
| 2010 CA-20 House | D+6.0 | D+3.4 | 2.6 | ✓ |
| 2010 CO-4 House | EVEN | R+11.1 | 11.1 | ✗ |
| 2010 IA-3 House | D+8.0 | D+4.2 | 3.8 | ✓ |
| 2010 SD-1 House | D+13.0 | R+2.2 | 15.2 | ✗ |
| 2010 TX-17 House | R+4.0 | R+25.2 | 21.2 | ✓ |
| 2010 KY Senate | D+2.0 | R+11.5 | 13.5 | ✗ |
| 2010 PA Senate | R+1.0 | R+2.0 | 1.0 | ✓ |
| 2008 IL-10 House | R+6.0 | R+5.1 | 0.9 | ✓ |
| 2008 IL-11 House | D+21.0 | D+23.9 | 2.9 | ✓ |
| 2008 LA-6 House | D+17.0 | R+7.8 | 24.8 | ✗ |
| 2008 MN-3 House | D+5.0 | R+7.6 | 12.6 | ✗ |
| 2008 VA-2 House | R+5.0 | D+4.9 | 9.9 | ✗ |
| 2008 WA-8 House | D+5.0 | R+5.6 | 10.6 | ✗ |
| 2006 AZ-5 House | R+1.0 | D+4.0 | 5.0 | ✗ |
| 2006 AZ-8 House | D+25.0 | D+12.2 | 12.8 | ✓ |
| 2006 CT-4 House | D+3.0 | R+3.4 | 6.4 | ✗ |
| 2006 IA-1 House | D+11.0 | D+11.9 | 0.9 | ✓ |
| 2006 IL-6 House | D+1.0 | R+2.7 | 3.7 | ✗ |
| 2006 IL-8 House | D+16.0 | D+6.9 | 9.1 | ✓ |
| 2006 PA-10 House | D+14.0 | D+5.9 | 8.1 | ✓ |
| 2006 TX-17 House | D+21.0 | D+17.8 | 3.2 | ✓ |
| 2006 MO Senate | D+5.0 | D+2.3 | 2.7 | ✓ |
| 2006 NJ Senate | D+9.0 | D+16.0 | 7.0 | ✓ |
| 2004 IL-8 House | R+2.0 | D+3.4 | 5.4 | ✗ |
| 2004 TX-1 House | D+4.0 | R+23.8 | 27.8 | ✗ |
| 2004 TX-17 House | D+10.0 | D+3.8 | 6.2 | ✓ |
| 2002 ME-2 House | D+12.0 | D+4.0 | 8.0 | ✓ |
| 2002 TX Senate | R+2.0 | R+12.0 | 10.0 | ✓ |
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 | 9 | 6.53 | +1.46 | 56% |
| 3–6 wk | 31 | 8.37 | +2.63 | 58% |
| 6–9 wk | 12 | 8.37 | +2.22 | 50% |
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) |
|---|---|---|---|
| 2004 | 3 | 13.1 | D+9.5 |
| 2006 | 10 | 5.9 | D+3.3 |
| 2008 | 7 | 10.2 | D+3.5 |
| 2010 | 8 | 10.0 | D+10.0 |
| 2014 | 5 | 6.3 | D+4.0 |
| 2016 | 10 | 7.4 | D+6.1 |
| 2020 | 3 | 9.4 | D+9.4 |
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.