Simon Hix and Nick Vivyan discuss polling predictions and the likelihood of a hung parliament. Please note that these predictions are based on data gathered up to 2 March. This will be updated early next week.
We have developed a model for predicting the outcome of the next British General Election. This works in three steps.
The first step combines information from national polling data, to gain an estimate of national vote-share for each party. To do this we adapt a “pooling the polls” method developed by Simon Jackman, a political scientist at Stanford University in California. This method combines recent polling data from each British polling agency, taking into account how far each agency tends to deviate from the overall average.
The second step involves translating these polled estimates of national-level party vote shares into constituency-level predictions. In doing this we take into account information on how marginal seats in different regions in the country deviate from national-level vote-swings between parties. For this we use the most fine-grained polling data from marginal seats that is currently available.
The third step involves calculating the probability of likely election outcomes, in terms of seats in the House of Commons. We do this by generating a large number of simulations of national vote-share for each party, based on the pooling the polls model. Then for each of these simulations we calculate constituency-level predictions and aggregate these up to House of Commons seats.
Based on the opinion polls up to 2 March, our model predicts the following outcome:Conservatives 6 seats short of a majority.
| 2005 Election Result | Hix-Vivyan Prediction | |
|---|---|---|
| Conservatives | 208 | 320 |
| Labour | 346 | 255 |
| Liberal Democrat | 67 | 43 |
| Nationalists | 8 | 12 |
| Other Parties | 21 | 20 |
In terms of the probability of certain outcomes, our model currently suggests that there is a 67 per cent chance of a hung parliament, a 33 per cent chance of Conservative majority, and a 0 per cent chance of a Labour majority.
Simon Hix is a Professor of European and Comparative Politics in the Department of Government at the LSE
Nick Vivyan is a PhD student in the Department of Government at the LSE.
Thanks for this description of the Hix-Vivyan method. Has this method been used to predict other elections or has tested for accuracy using back-projections on data from previous elections? In other words, how does it stand up to other predictive methods? Thanks.