MRG Research
Graded against the actual result across 24 races (from 42 polls, through 2020).
Head-to-head vs the VotePredictor model
The fair, apples-to-apples test: on the 24 races MRG 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 |
|---|---|---|
| MRG Research | 8.32 | 79% |
| VotePredictor | 5.85 | 92% |
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 (24)
Each race MRG 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 |
|---|---|---|---|---|
| 2020 HI President | D+33.0 | D+29.5 | 3.5 | ✓ |
| 2018 HI Governor | D+21.0 | D+29.0 | 8.0 | ✓ |
| 2014 HI Governor | D+6.0 | D+12.4 | 6.4 | ✓ |
| 2014 HI-1 House | EVEN | D+3.9 | 3.9 | ✗ |
| 2014 HI-2 House | D+50.0 | D+60.1 | 10.1 | ✓ |
| 2014 HI Senate | D+26.0 | D+42.1 | 16.1 | ✓ |
| 2012 HI-1 House | D+11.0 | D+9.2 | 1.8 | ✓ |
| 2012 HI-2 House | D+58.0 | D+61.1 | 3.1 | ✓ |
| 2012 HI President | D+27.0 | D+42.7 | 15.7 | ✓ |
| 2012 US President | D+3.0 | D+3.9 | 0.9 | ✓ |
| 2012 HI Senate | D+15.0 | D+25.2 | 10.2 | ✓ |
| 2010 CT Governor | EVEN | R+1.7 | 1.7 | ✗ |
| 2010 HI Governor | D+5.0 | D+17.1 | 12.1 | ✓ |
| 2010 CT-1 House | D+18.0 | D+20.6 | 2.6 | ✓ |
| 2010 CT-2 House | D+19.0 | D+18.3 | 0.7 | ✓ |
| 2010 CT-3 House | D+18.1 | D+27.4 | 9.3 | ✓ |
| 2010 CT-4 House | R+2.0 | D+4.0 | 6.0 | ✗ |
| 2010 CT-5 House | R+8.0 | D+7.1 | 15.1 | ✗ |
| 2010 HI-1 House | D+5.0 | D+6.5 | 1.5 | ✓ |
| 2010 HI-2 House | D+23.5 | D+46.9 | 23.4 | ✓ |
| 2010 WA-2 House | D+18.0 | D+2.1 | 15.9 | ✓ |
| 2010 CT Senate | D+8.0 | D+9.3 | 1.3 | ✓ |
| 2010 HI Senate | D+32.1 | D+53.2 | 21.1 | ✓ |
| 2010 MA Senate | R+9.6 | R+4.8 | 4.8 | ✓ |
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 | 22 | 8.72 | +3.65 | 82% |
| 3–6 wk | 14 | 6.92 | +1.18 | 86% |
| 6–9 wk | 4 | 7.10 | +0.95 | 75% |
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 | 23 | 8.2 | R+6.2 |
| 2012 | 9 | 7.9 | R+7.5 |
| 2014 | 8 | 8.1 | R+8.1 |
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.