Dan Jones & Associates
Graded against the actual result across 43 races (from 73 polls, through 2022).
Head-to-head vs the VotePredictor model
The fair, apples-to-apples test: on the 37 races Dan Jones & 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 |
|---|---|---|
| Dan Jones & Associates | 4.94 | 97% |
| VotePredictor | 4.38 | 97% |
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 (42)
Each race Dan Jones & 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 |
|---|---|---|---|---|
| 2017 UT-3 House | R+27.0 | R+32.4 | 5.4 | ✓ |
| 2016 UT Governor | R+39.0 | R+38.0 | 1.0 | ✓ |
| 2016 ID-1 House | R+20.0 | R+36.4 | 16.4 | ✓ |
| 2016 ID-2 House | R+26.0 | R+33.5 | 7.5 | ✓ |
| 2016 UT-4 House | R+13.0 | R+12.5 | 0.5 | ✓ |
| 2016 ID President | R+10.0 | R+31.8 | 21.8 | ✓ |
| 2016 UT President | R+8.0 | R+18.1 | 10.1 | ✓ |
| 2014 UT-1 House | R+32.0 | R+36.8 | 4.8 | ✓ |
| 2014 UT-2 House | R+32.0 | R+28.3 | 3.7 | ✓ |
| 2014 UT-3 House | R+51.0 | R+49.7 | 1.3 | ✓ |
| 2014 UT-4 House | R+5.0 | R+5.1 | 0.1 | ✓ |
| 2012 UT Governor | R+45.0 | R+40.8 | 4.2 | ✓ |
| 2012 UT-1 House | R+57.0 | R+46.8 | 10.2 | ✓ |
| 2012 UT-2 House | R+16.0 | R+28.7 | 12.7 | ✓ |
| 2012 UT-3 House | R+50.0 | R+53.2 | 3.2 | ✓ |
| 2012 UT-4 House | R+5.0 | D+0.3 | 5.3 | ✗ |
| 2012 UT President | R+43.0 | R+48.0 | 5.0 | ✓ |
| 2012 UT Senate | R+37.0 | R+35.3 | 1.7 | ✓ |
| 2010 UT Governor | R+34.0 | R+32.2 | 1.8 | ✓ |
| 2010 UT-1 House | R+45.0 | R+45.3 | 0.3 | ✓ |
| 2010 UT-2 House | D+12.0 | D+4.4 | 7.6 | ✓ |
| 2010 UT-3 House | R+51.0 | R+49.4 | 1.6 | ✓ |
| 2010 UT Senate | R+27.0 | R+28.8 | 1.8 | ✓ |
| 2008 UT Governor | R+61.0 | R+57.9 | 3.1 | ✓ |
| 2008 UT-1 House | R+33.0 | R+34.4 | 1.4 | ✓ |
| 2008 UT-2 House | D+39.0 | D+28.9 | 10.1 | ✓ |
| 2008 UT-3 House | R+34.0 | R+37.3 | 3.3 | ✓ |
| 2008 UT President | R+25.0 | R+28.2 | 3.2 | ✓ |
| 2006 UT-2 House | D+36.0 | D+21.7 | 14.3 | ✓ |
| 2006 UT-3 House | R+29.0 | R+25.5 | 3.5 | ✓ |
| 2006 UT Senate | R+37.0 | R+31.3 | 5.7 | ✓ |
| 2004 UT Governor | R+16.0 | R+16.4 | 0.4 | ✓ |
| 2004 UT-1 House | R+43.0 | R+38.8 | 4.2 | ✓ |
| 2004 UT-2 House | D+10.0 | D+11.5 | 1.5 | ✓ |
| 2004 UT-3 House | R+30.0 | R+30.9 | 0.9 | ✓ |
| 2004 UT President | R+45.0 | R+45.5 | 0.5 | ✓ |
| 2004 UT Senate | R+43.0 | R+40.3 | 2.7 | ✓ |
| 2002 UT-2 House | D+6.0 | D+0.7 | 5.3 | ✓ |
| 2000 UT Governor | R+15.0 | R+13.5 | 1.5 | ✓ |
| 2000 UT President | R+32.0 | R+40.5 | 8.5 | ✓ |
| 2000 UT Senate | R+40.0 | R+34.1 | 5.9 | ✓ |
| 1998 UT Senate | R+38.0 | R+31.0 | 7.0 | ✓ |
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 wk | 5 | 2.62 | -1.66 | 100% |
| 1–3 wk | 27 | 4.31 | -0.76 | 96% |
| 3–6 wk | 22 | 8.03 | +2.29 | 96% |
| 6–9 wk | 19 | 6.71 | +0.56 | 100% |
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 | 5 | 3.9 | R+0.1 |
| 2004 | 15 | 6.0 | D+4.9 |
| 2006 | 3 | 7.8 | D+1.7 |
| 2008 | 10 | 5.4 | R+0.6 |
| 2010 | 12 | 5.3 | D+4.7 |
| 2012 | 8 | 6.1 | R+0.8 |
| 2014 | 5 | 2.8 | R+0.8 |
| 2016 | 9 | 9.4 | D+7.9 |
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.