Digital Research
Graded against the actual result across 37 races (from 58 polls, through 2020).
Head-to-head vs the VotePredictor model
The fair, apples-to-apples test: on the 32 races Digital 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 |
|---|---|---|
| Digital Research | 7.36 | 91% |
| VotePredictor | 5.00 | 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 (37)
Each race Digital 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 ME-1 House | D+34.0 | D+24.3 | 9.7 | ✓ |
| 2020 ME-2 House | D+18.0 | D+6.1 | 11.9 | ✓ |
| 2020 M1 President | D+32.0 | D+23.1 | 8.9 | ✓ |
| 2020 M2 President | R+8.0 | R+7.4 | 0.6 | ✓ |
| 2020 ME President | D+11.0 | D+9.1 | 1.9 | ✓ |
| 2020 ME Senate | D+1.0 | R+8.6 | 9.6 | ✗ |
| 2018 ME Senate | R+20.0 | R+24.8 | 4.8 | ✓ |
| 2014 ME Governor | R+3.0 | R+4.8 | 1.8 | ✓ |
| 2014 ME-1 House | D+34.0 | D+29.8 | 4.2 | ✓ |
| 2014 ME-2 House | R+5.0 | R+5.2 | 0.2 | ✓ |
| 2012 ME-1 House | D+19.0 | D+29.6 | 10.6 | ✓ |
| 2012 ME-2 House | D+19.0 | D+16.4 | 2.6 | ✓ |
| 2012 M1 President | D+6.0 | D+21.4 | 15.4 | ✓ |
| 2012 M2 President | D+7.0 | D+8.6 | 1.6 | ✓ |
| 2012 ME President | D+7.0 | D+15.3 | 8.3 | ✓ |
| 2012 ME Senate | R+22.0 | R+17.5 | 4.5 | ✓ |
| 2010 ME Governor | R+12.0 | R+26.7 | 14.7 | ✓ |
| 2010 PA Governor | R+1.0 | R+9.0 | 8.0 | ✓ |
| 2010 ME-1 House | R+4.0 | D+13.6 | 17.6 | ✗ |
| 2010 ME-2 House | D+4.0 | D+10.3 | 6.3 | ✓ |
| 2010 PA-10 House | R+9.0 | R+10.4 | 1.4 | ✓ |
| 2010 PA-11 House | D+8.0 | R+9.4 | 17.4 | ✗ |
| 2010 PA Senate | R+4.0 | R+2.0 | 2.0 | ✓ |
| 2008 ME-1 House | D+21.0 | D+9.8 | 11.2 | ✓ |
| 2008 ME-2 House | D+41.0 | D+34.9 | 6.1 | ✓ |
| 2008 ME President | D+21.0 | D+17.3 | 3.7 | ✓ |
| 2008 ME Senate | R+12.0 | R+22.7 | 10.7 | ✓ |
| 2006 ME Governor | D+17.0 | D+7.9 | 9.1 | ✓ |
| 2006 ME-1 House | D+44.0 | D+29.5 | 14.5 | ✓ |
| 2006 ME-2 House | D+37.0 | D+41.0 | 4.0 | ✓ |
| 2006 ME Senate | R+60.0 | R+53.7 | 6.3 | ✓ |
| 2004 ME President | D+3.0 | D+9.0 | 6.0 | ✓ |
| 2002 ME Governor | D+19.0 | D+5.7 | 13.3 | ✓ |
| 2002 ME-2 House | D+9.0 | D+4.0 | 5.0 | ✓ |
| 2002 ME Senate | R+29.0 | R+16.9 | 12.1 | ✓ |
| 2000 ME President | R+5.0 | D+5.1 | 10.1 | ✗ |
| 2000 ME Senate | R+56.0 | R+37.9 | 18.1 | ✓ |
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 | 19 | 8.39 | +3.32 | 95% |
| 3–6 wk | 22 | 7.78 | +2.04 | 86% |
| 6–9 wk | 15 | 6.36 | +0.21 | 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) |
|---|---|---|---|
| 2002 | 3 | 10.2 | D+2.1 |
| 2006 | 4 | 8.5 | D+3.3 |
| 2008 | 4 | 7.9 | D+7.9 |
| 2010 | 22 | 9.4 | D+6.2 |
| 2012 | 12 | 4.8 | R+2.7 |
| 2014 | 3 | 2.1 | D+2.1 |
| 2020 | 6 | 7.1 | D+6.9 |
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.