Research & Polling Inc.
Graded against the actual result across 45 races (from 82 polls, through 2022).
Head-to-head vs the VotePredictor model
The fair, apples-to-apples test: on the 38 races Research & Polling Inc. 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 |
|---|---|---|
| Research & Polling Inc. | 4.23 | 92% |
| VotePredictor | 4.84 | 90% |
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 (45)
Each race Research & Polling Inc. 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 NM Governor | D+8.0 | D+6.4 | 1.6 | ✓ |
| 2022 NM-1 House | D+6.0 | D+11.5 | 5.5 | ✓ |
| 2022 NM-2 House | D+2.0 | D+0.7 | 1.3 | ✓ |
| 2022 NM-3 House | D+18.0 | D+16.3 | 1.7 | ✓ |
| 2020 NM-1 House | D+21.0 | D+16.4 | 4.6 | ✓ |
| 2020 NM-2 House | R+2.0 | R+7.4 | 5.4 | ✓ |
| 2020 NM-3 House | D+23.0 | D+17.4 | 5.6 | ✓ |
| 2020 NM President | D+12.0 | D+10.8 | 1.2 | ✓ |
| 2020 NM Senate | D+8.0 | D+6.1 | 1.9 | ✓ |
| 2018 NM Governor | D+10.0 | D+14.4 | 4.4 | ✓ |
| 2018 NM-1 House | D+12.0 | D+22.8 | 10.8 | ✓ |
| 2018 NM-2 House | R+1.0 | D+1.9 | 2.9 | ✗ |
| 2018 NM Senate | D+20.0 | D+23.6 | 3.6 | ✓ |
| 2016 NM President | D+5.0 | D+8.2 | 3.2 | ✓ |
| 2014 NM Governor | R+15.0 | R+14.4 | 0.6 | ✓ |
| 2014 NM Senate | D+7.0 | D+11.1 | 4.1 | ✓ |
| 2012 NM-1 House | D+15.0 | D+18.3 | 3.3 | ✓ |
| 2012 NM-2 House | R+18.0 | R+18.2 | 0.2 | ✓ |
| 2012 NM-3 House | D+18.0 | D+26.2 | 8.2 | ✓ |
| 2012 NM President | D+9.0 | D+10.1 | 1.1 | ✓ |
| 2012 NM Senate | D+8.0 | D+5.7 | 2.3 | ✓ |
| 2010 NM Governor | R+10.0 | R+6.7 | 3.3 | ✓ |
| 2010 NM-1 House | R+3.0 | D+3.6 | 6.6 | ✗ |
| 2010 NM-2 House | R+3.0 | R+10.8 | 7.8 | ✓ |
| 2008 NM-1 House | D+4.0 | D+11.3 | 7.3 | ✓ |
| 2008 NM-2 House | D+4.0 | D+11.9 | 7.9 | ✓ |
| 2008 NM-3 House | D+28.0 | D+26.3 | 1.7 | ✓ |
| 2008 NM President | D+8.0 | D+15.1 | 7.1 | ✓ |
| 2008 NM Senate | D+14.0 | D+22.7 | 8.7 | ✓ |
| 2006 NM Governor | D+32.0 | D+37.6 | 5.6 | ✓ |
| 2006 NM-1 House | D+4.0 | R+0.4 | 4.4 | ✗ |
| 2006 NM-2 House | R+26.0 | R+18.9 | 7.1 | ✓ |
| 2006 NM-3 House | D+51.0 | D+49.3 | 1.7 | ✓ |
| 2006 NM Senate | D+42.0 | D+41.3 | 0.7 | ✓ |
| 2004 NM-1 House | R+8.0 | R+8.9 | 0.9 | ✓ |
| 2004 NM-2 House | R+12.0 | R+20.4 | 8.4 | ✓ |
| 2004 NM-3 House | D+43.0 | D+37.4 | 5.6 | ✓ |
| 2004 NM President | R+3.0 | R+0.8 | 2.2 | ✓ |
| 2002 NM Governor | D+18.0 | D+16.4 | 1.6 | ✓ |
| 2002 NM-1 House | R+6.0 | R+10.7 | 4.7 | ✓ |
| 2002 NM-2 House | R+5.0 | R+12.5 | 7.5 | ✓ |
| 2002 NM Senate | R+38.0 | R+30.1 | 7.9 | ✓ |
| 2000 NM President | R+3.0 | D+0.1 | 3.1 | ✗ |
| 2000 NM Senate | D+29.0 | D+23.4 | 5.6 | ✓ |
| 1998 NM Governor | R+10.0 | R+9.1 | 0.9 | ✓ |
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 wk | 21 | 4.94 | +0.66 | 86% |
| 1–3 wk | 22 | 3.40 | -1.67 | 91% |
| 3–6 wk | 26 | 5.54 | -0.20 | 92% |
| 6–9 wk | 13 | 6.61 | +0.46 | 77% |
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) |
|---|---|---|---|
| 2000 | 3 | 3.2 | D+0.5 |
| 2002 | 12 | 6.8 | R+0.4 |
| 2004 | 7 | 5.9 | D+5.3 |
| 2006 | 10 | 4.2 | R+1.1 |
| 2008 | 9 | 7.0 | R+6.6 |
| 2010 | 6 | 5.2 | D+1.8 |
| 2012 | 10 | 3.1 | R+1.9 |
| 2014 | 4 | 2.5 | R+1.6 |
| 2018 | 8 | 6.9 | R+6.9 |
| 2020 | 5 | 3.8 | D+3.8 |
| 2022 | 4 | 2.5 | R+0.2 |
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.