RKM Research and Communications Inc.
Graded against the actual result across 21 races (from 34 polls, through 2016).
Every race (21)
Each race RKM Research and Communications Inc. 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 |
|---|---|---|---|---|
| 2016 US President | D+3.2 | D+2.2 | 1.0 | ✓ |
| 2004 NE-1 House | R+4.0 | R+11.2 | 7.2 | ✓ |
| 2004 NE-2 House | R+30.0 | R+24.9 | 5.1 | ✓ |
| 2004 NE-3 House | R+78.0 | R+76.9 | 1.1 | ✓ |
| 2004 NE President | R+29.0 | R+33.2 | 4.2 | ✓ |
| 2002 MA Governor | D+1.0 | R+4.8 | 5.8 | ✗ |
| 2002 ME Governor | D+10.0 | D+5.7 | 4.3 | ✓ |
| 2002 ME-2 House | D+1.0 | D+4.0 | 3.0 | ✓ |
| 2002 ME Senate | R+28.0 | R+16.9 | 11.1 | ✓ |
| 2000 MA President | D+24.0 | D+27.0 | 3.0 | ✓ |
| 2000 ME President | EVEN | D+5.1 | 5.1 | ✗ |
| 2000 NE President | R+25.0 | R+29.0 | 4.0 | ✓ |
| 2000 NH President | D+1.0 | R+1.3 | 2.3 | ✗ |
| 2000 MA Senate | D+49.0 | D+56.9 | 7.9 | ✓ |
| 2000 ME Senate | R+51.0 | R+37.9 | 13.1 | ✓ |
| 2000 NE Senate | D+12.0 | D+2.2 | 9.8 | ✓ |
| 1998 MA Governor | R+9.0 | R+3.4 | 5.6 | ✓ |
| 1998 NE Governor | R+14.0 | R+7.9 | 6.1 | ✓ |
| 1998 NH Governor | D+41.0 | D+35.2 | 5.8 | ✓ |
| 1998 NH-2 House | R+15.0 | R+8.4 | 6.6 | ✓ |
| 1998 NH Senate | R+45.0 | R+39.6 | 5.4 | ✓ |
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 | 17 | 5.48 | +0.41 | 82% |
| 3–6 wk | 4 | 7.05 | +1.31 | 75% |
| 6–9 wk | 10 | 9.65 | +3.50 | 90% |
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) |
|---|---|---|---|
| 1998 | 7 | 5.5 | R+3.9 |
| 2000 | 10 | 8.1 | D+2.3 |
| 2002 | 10 | 8.9 | D+3.1 |
| 2004 | 4 | 4.4 | D+1.3 |
| 2016 | 3 | 1.3 | D+1.2 |
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.