Harstad Strategic Research
Graded against the actual result across 25 races (from 33 polls, through 2020).
Head-to-head vs the VotePredictor model
The fair, apples-to-apples test: on the 24 races Harstad Strategic 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 |
|---|---|---|
| Harstad Strategic Research | 6.40 | 67% |
| VotePredictor | 3.20 | 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 (25)
Each race Harstad Strategic 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 AK President | R+1.0 | R+10.1 | 9.1 | ✓ |
| 2020 AK Senate | D+1.0 | R+12.7 | 13.7 | ✗ |
| 2016 MT-1 House | R+3.0 | R+15.6 | 12.6 | ✓ |
| 2014 AK Senate | EVEN | R+2.1 | 2.1 | ✗ |
| 2014 CO Senate | D+1.0 | R+1.9 | 2.9 | ✗ |
| 2014 IA Senate | EVEN | R+8.3 | 8.3 | ✗ |
| 2012 AZ Senate | D+4.0 | R+3.0 | 7.0 | ✗ |
| 2012 MA Senate | D+6.0 | D+7.5 | 1.5 | ✓ |
| 2012 MO Senate | D+12.0 | D+15.7 | 3.7 | ✓ |
| 2012 PA Senate | D+12.0 | D+9.1 | 2.9 | ✓ |
| 2010 AZ-5 House | D+1.0 | R+8.8 | 9.8 | ✗ |
| 2010 FL-22 House | D+5.0 | R+8.7 | 13.7 | ✗ |
| 2010 HI-1 House | R+16.0 | R+8.7 | 7.3 | ✓ |
| 2010 PA-8 House | D+3.0 | R+7.0 | 10.0 | ✗ |
| 2010 CO Senate | D+3.0 | D+1.6 | 1.4 | ✓ |
| 2008 ID-1 House | D+7.0 | D+1.2 | 5.8 | ✓ |
| 2008 ID-2 House | R+44.0 | R+42.0 | 2.0 | ✓ |
| 2008 ID President | R+23.0 | R+25.4 | 2.4 | ✓ |
| 2008 CO Senate | D+11.0 | D+10.3 | 0.7 | ✓ |
| 2008 ID Senate | R+12.0 | R+23.5 | 11.5 | ✓ |
| 2006 MT Senate | D+6.0 | D+0.9 | 5.1 | ✓ |
| 2004 TX-2 House | R+4.0 | R+12.6 | 8.6 | ✓ |
| 2004 IA President | R+3.0 | R+0.7 | 2.3 | ✓ |
| 2004 CO Senate | D+6.0 | D+4.8 | 1.2 | ✓ |
| 2002 CO-7 House | D+1.0 | R+0.1 | 1.1 | ✗ |
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 | 9 | 4.01 | -1.06 | 78% |
| 3–6 wk | 18 | 6.71 | +0.97 | 56% |
| 6–9 wk | 6 | 5.88 | -0.27 | 67% |
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) |
|---|---|---|---|
| 2004 | 4 | 3.4 | D+2.2 |
| 2008 | 7 | 4.4 | D+3.9 |
| 2010 | 6 | 7.7 | D+5.2 |
| 2012 | 4 | 3.8 | D+1.2 |
| 2014 | 4 | 5.1 | D+5.1 |
| 2020 | 4 | 11.8 | D+11.8 |
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.