Princeton Survey Research Associates International
Graded against the actual result across 17 races (from 48 polls, through 2017).
Head-to-head vs VotePredictor Elections
The fair, apples-to-apples test: on the 14 races Princeton Survey Research Associates International actually polled, how its final poll's margin compared to what VotePredictor Elections predicted for those same races.
| Model | Avg miss (pts) | Called right |
|---|---|---|
| Princeton Survey Research Associates International | 3.77 | 100% |
| VotePredictor Elections | 3.51 | 93% |
VotePredictor Elections 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 (17)
Each race Princeton Survey Research Associates International 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 VA Governor | D+5.0 | D+8.9 | 3.9 | ✓ |
| 2016 VA President | D+3.0 | D+5.3 | 2.3 | ✓ |
| 2014 VA Senate | D+10.0 | D+0.8 | 9.2 | ✓ |
| 2013 VA Governor | D+7.0 | D+2.5 | 4.5 | ✓ |
| 2012 WI Governor | R+8.0 | R+6.8 | 1.2 | ✓ |
| 2012 CA President | D+15.0 | D+23.1 | 8.1 | ✓ |
| 2012 US President | D+5.0 | D+3.9 | 1.1 | ✓ |
| 2012 CA Senate | D+26.0 | D+25.0 | 1.0 | ✓ |
| 2010 MN Governor | D+7.0 | D+0.4 | 6.6 | ✓ |
| 2008 MN President | D+11.0 | D+10.2 | 0.8 | ✓ |
| 2008 SC President | R+11.0 | R+9.0 | 2.0 | ✓ |
| 2008 US President | D+12.0 | D+7.4 | 4.6 | ✓ |
| 2008 MN Senate | D+4.0 | EVEN | 4.0 | ✓ |
| 2004 US President | R+6.0 | R+2.4 | 3.6 | ✓ |
| 2000 NY President | D+13.0 | D+24.2 | 11.2 | ✓ |
| 2000 US President | R+2.0 | D+0.5 | 2.5 | ✗ |
| 2000 NY Senate | D+1.0 | D+12.0 | 11.0 | ✓ |
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 | 4 | 2.70 | -1.58 | 75% |
| 1–3 wk | 14 | 4.71 | -0.36 | 79% |
| 3–6 wk | 18 | 5.58 | -0.16 | 78% |
| 6–9 wk | 12 | 5.84 | -0.31 | 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) |
|---|---|---|---|
| 2000 | 17 | 6.1 | R+1.6 |
| 2004 | 6 | 4.2 | R+1.9 |
| 2008 | 12 | 4.8 | D+0.8 |
| 2012 | 7 | 3.8 | R+2.3 |
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.