National Research
Graded against the actual result across 30 races (from 40 polls, through 2021).
Head-to-head vs the VotePredictor model
The fair, apples-to-apples test: on the 30 races National Research 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 |
|---|---|---|
| National Research | 5.75 | 67% |
| VotePredictor | 3.80 | 70% |
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 (30)
Each race National 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 |
|---|---|---|---|---|
| 2021 NJ Governor | D+3.0 | D+3.2 | 0.2 | ✓ |
| 2020 MI-3 House | R+7.0 | R+5.9 | 1.1 | ✓ |
| 2020 GA President | D+4.0 | D+0.2 | 3.8 | ✓ |
| 2020 IA President | D+2.0 | R+8.2 | 10.2 | ✗ |
| 2020 MI President | D+9.0 | D+2.8 | 6.2 | ✓ |
| 2020 GA Senate | D+1.0 | D+2.1 | 1.1 | ✓ |
| 2020 IA Senate | R+1.0 | R+6.6 | 5.6 | ✓ |
| 2020 MI Senate | D+8.0 | D+1.7 | 6.3 | ✓ |
| 2018 NJ-11 House | D+3.0 | D+14.6 | 11.6 | ✓ |
| 2018 NJ-3 House | R+4.0 | D+1.3 | 5.3 | ✗ |
| 2018 NJ Senate | D+2.0 | D+11.2 | 9.2 | ✓ |
| 2016 NC Governor | R+4.0 | D+0.2 | 4.2 | ✗ |
| 2016 NC President | D+2.0 | R+3.7 | 5.7 | ✗ |
| 2016 NC Senate | R+7.0 | R+5.7 | 1.3 | ✓ |
| 2014 FL-2 House | R+6.0 | D+1.1 | 7.1 | ✗ |
| 2014 NJ-3 House | R+7.0 | R+9.6 | 2.6 | ✓ |
| 2014 NC Senate | EVEN | R+1.6 | 1.6 | ✗ |
| 2012 NC Governor | R+11.0 | R+11.4 | 0.4 | ✓ |
| 2012 CT-5 House | R+6.0 | D+3.1 | 9.1 | ✗ |
| 2012 NC President | R+1.0 | R+2.0 | 1.0 | ✓ |
| 2012 CT Senate | R+3.0 | D+12.5 | 15.5 | ✗ |
| 2010 AZ-5 House | R+2.0 | R+8.8 | 6.8 | ✓ |
| 2010 CT-4 House | R+4.0 | D+4.0 | 8.0 | ✗ |
| 2010 FL-2 House | R+16.0 | R+12.2 | 3.8 | ✓ |
| 2010 MI-7 House | R+13.0 | R+4.8 | 8.2 | ✓ |
| 2010 NJ-6 House | D+1.0 | D+11.0 | 10.0 | ✓ |
| 2010 AK Senate | R+6.0 | R+12.1 | 6.1 | ✓ |
| 2010 NC Senate | R+20.0 | R+11.8 | 8.2 | ✓ |
| 2008 MI-7 House | R+10.0 | D+2.3 | 12.3 | ✗ |
| 2008 NJ-7 House | R+9.0 | R+8.0 | 1.0 | ✓ |
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 | 8 | 4.18 | -0.89 | 63% |
| 3–6 wk | 21 | 6.59 | +0.85 | 62% |
| 6–9 wk | 8 | 4.22 | -1.93 | 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) |
|---|---|---|---|
| 2010 | 8 | 6.9 | R+3.7 |
| 2012 | 5 | 6.4 | R+3.4 |
| 2014 | 5 | 3.7 | D+0.8 |
| 2016 | 6 | 3.0 | D+0.4 |
| 2018 | 4 | 8.8 | R+8.8 |
| 2020 | 9 | 4.0 | D+3.5 |
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.