It’s no secret that in the 2017 General Elections the opinion polls missed the final result by a not insignificant margin. A number of different websites used data from opinion polls to try and forecast the outcome of the election and whilst all missed the Hung Parliament, some did better than others.
Wikipedia has a handy guide to forecasts last time which can help us see how Forecast UK did alongside other websites and organisations. In the table below you can see the forecast for each website together with the actual result and the sum of the errors.
| Election Forecast | Electoral Calculus | Lord Ashcroft | Elections Etc. | Britain Elects | Scenari Politici.com | Forecast UK | 2017 Result | |
| Con | 371 | 361 | 357 | 360 | 353 | 365 | 347.5 | 317 |
| Lab | 199 | 215 | 222 | 210 | 219 | 208 | 224.5 | 262 |
| SNP | 50 | 48 | 45 | 48 | 46 | 49 | 48 | 35 |
| Lib Dem | 7 | 4 | 4 | 9 | 9 | 6 | 6 | 12 |
| Plaid | 2 | 3 | 1 | 3 | 3 | 3 | 3 | 4 |
| Green | 1 | 1 | 0 | 1 | 1 | 1 | 1 | 1 |
| UKIP | 1 | 0 | 0 | 0 | 0 | 0 | 0 | 0 |
| Difference | 140 | 113 | 102 | 112 | 94 | 123 | 88 | |
In the table above the New Statesman forecast has been removed from the Wikipedia table as it does not include the minor parties. The YouGov model is also omitted as it is an MRP model with access to greater constituency data granularity than the other forecast methods.
Of all the forecasts, Forecast UK was the most accurate, though Britain Elects came close. This might be viewed as random chance, but we use a particular methodology, back-tested on previous elections in the UK and abroad, to analyse and project along the trend of the polling. This is one reason why, despite predicting a Conservative majority like every other forecast, we were more cautious on the size of the Conservative majority.
We expect polling to be more accurate in this election and we continue to use our standard methodology to try and forecast the outcome of the election. We will continue to produce a Nowcast until around the 8th of December, switching to a Forwardcast the weekend before polling day.