United Kingdom, 1997: the 5 final polls missed each party by 1.3 points on average. The biggest miss was Lab, +2.3 points; the closest pollster was ICM.
1.27
Poll average: average miss per party
3 parties at 5%+; all 3 measured: 1.27
2.19
A typical single poll
points a party, same parties
+2.3
Biggest miss: Lab
polls 46.6, result 44.3
ICM
Closest to the result
1.20 points a party; furthest: NOP 4.13
Each final poll against the result
The line is the result. Each dot is one pollster's final poll, placed by how far it was from the result; the ring is the poll average. Right of the line, the polls had the party too high.
← too lowtoo high →polls / result
Lab
+2.346.6 / 44.3
Con
−0.730.8 / 31.5
Lib Dem
−0.816.4 / 17.2
resultone pollster's final pollaverage of the final polls
The poll average against the result
| Party | Result | Poll average | Miss | Lowest–highest poll | Polls |
|---|---|---|---|---|---|
| Lab | 44.3 | 46.6 | +2.3 | 43.0–50.0 | 5 |
| Con | 31.5 | 30.8 | −0.7 | 28.0–33.0 | 5 |
| Lib Dem | 17.2 | 16.4 | −0.8 | 14.0–19.0 | 5 |
Each pollster's final poll
Closest first. Under each figure, the miss: + too high, − too low.
| Pollster | Lab | Con | Lib Dem | Avg miss |
|---|---|---|---|---|
| ICM 30 Apr 1997 | 43.0−1.3 | 33.0+1.5 | 18.0+0.8 | 1.20 |
| Gallup 30 Apr 1997 | 46.0+1.7 | 33.0+1.5 | 16.0−1.2 | 1.47 |
| Ipsos 30 Apr 1997 | 46.0+1.7 | 29.0−2.5 | 19.0+1.8 | 2.00 |
| Harris 29 Apr 1997 | 48.0+3.7 | 31.0−0.5 | 15.0−2.2 | 2.13 |
| NOP 29 Apr 1997 | 50.0+5.7 | 28.0−3.5 | 14.0−3.2 | 4.13 |
| Average of the final polls | 46.6+2.3 | 30.8−0.7 | 16.4−0.8 | 1.27 |
| Result | 44.3 | 31.5 | 17.2 |
How this is measured
- Parties counted in the average miss: Lab, Con, Lib Dem. A party counts if it won 5% or more and at least half the final polls reported it; "others" never count.
- British figures are each company's final Great Britain poll and the GB result, from Mark Pack's PollBase (used with credit). Greens and Reform (UKIP, the Brexit Party) are compared from 2015, the first election PollBase lists them in every poll. The forecast itself uses England figures from 2015 on: see the UK track record.