This post originally appeared on Politics in Spires.
As the dust settles on the results of the 2014 local and European elections several questions remain unanswered about what the results mean for the future of British politics: Who will win the next general election? How well will UKIP do? Are the Liberal Democrats doomed?
Predicting general elections from local election results
Although local and European elections are notionally concerned with who represents us at the local and European level, most media analysis and political commentary about the electoral results is more concerned with national politics (as indeed are the decisions of many voters). Two years ago I developed a simple statistical model that tries to predict the outcome of general elections from local election results. Although local election and general election results tend to be quite similar to begin with (the two are 90% correlated) a statistical prediction might offer a better indicator of what is likely to happen in the future then simply taking the results at face value. Local election results tend to differ from general elections in certain predicable ways – for example incumbent government parties have tended to perform worse on average at local elections than their eventual general election results by about 4.5%. Controlling for these factors improves the predictive power of local elections by about 40%: the vote share for each party at local elections is 4.4 percentage points different on average from their share at the next general election. The average difference between the vote shares fitted by the model and the eventual results is only 2.7.
Since my original specification of the model I’ve adjusted it slightly in order to try and get a prediction of the UKIP vote. This cannot be done in the same way as it is for the Conservative, Labour, or Liberal Democrat vote for the simple reason that the data does not exist – UKIP was not a real feature of local elections until very recently. We can however use the same procedure to predict the total ‘other’ vote and then make an informed guess as to how much of this is likely to be UKIP. The predictions below are based on the assumption that UKIP will get the same proportion of the other vote at the next general election as they did at this year’s local elections (63%), though of course other proportions are possible.
Inputting the 2014 results gives a forecast of:
- Conservative: 35.9% (± 1.3)
- Labour: 30.9% (± 3.4)
- Liberal Democrat: 16% (± 2.29)
- Total ‘Others’: 16.4% (± 5.3)
- Implied UKIP: 11.6% (± 3.9)
This forecast (which, incidentally, is almost identical to Steve Fisher’s latest forecast – based on polling data – for all parties except the Liberal Democrats) suggests that the Conservatives are likely to be the largest party at the next election. It also suggests that the Liberal Democrats will pick up another 3% or so from their performance at this year’s local elections (and substantially better than their European Parliament election performance). We should be cautious though in interpreting the forecast for the Liberal Democrats – for the simple reason that it has never happened before it is impossible to know what the effect of being in government is going to be on the performance of the Liberal Democrats at local elections relative to their eventual performance at the next general election. This forecast assumes a ‘normal’ pattern for the Liberal Democrat vote, which of course may well not be the case (the alternative of giving the Lib Dems the same incumbency effect as found for the other two main parties has them returning to around their 2010 vote share, which at this stage of the game seems unlikely). For the same reason we should take the UKIP forecast with a grain of salt – we are in uncharted territory – but an informed guess based on the assumptions of the historical performance of ‘others’ relative to the three main parties and of the performance of UKIP relative to the ‘others’ at recent elections is probably a more robust approach than the alternatives of reading tea leaves and gazing into a crystal ball.
Perhaps more interesting than a single forecast is comparing the forecasts generated by each of the four local elections that have been held under the current government, shown below. Two points stand out. 1) Each local election predicts that the Conservatives will win more of the vote than Labour at the next election, though the confidence intervals for the 2012 and 2013 elections are overlapping. 2) The rise of UKIP at the 2013 local elections dramatically alters the predicted UKIP vote share, lowers both the Conservative and Labour vote shares but also increases the margin of error around the predictions.
The uncertainty produced by the rise of UKIP is even more dramatically illustrated by a second model, which predicts the probability that the Conservatives or Labour will be the largest party at the next election. For the 2011 and 2012 local elections the model is fairly certain that the Conservatives will be the largest party at the next election. For the 2013 local elections however, the margin of error is so wide as to cover almost the entire range of probabilities – perhaps the statistical equivalent of shrugging your shoulders and admitting you haven’t got a clue. The 2014 local election produces a slightly narrower range of probabilities but the margin of error is still so wide as to make the prediction effectively useless.
Why this is the case is simple: the models base their prediction on the historical relationship between local election performance and general election performance. The vote share of the two major parties is unprecedentedly low: the combined Conservative and Labour vote share in 2013 is lower than any previous local election, and the 2014 vote share is only higher than two: 2009 and 2013. Although the forecast is muddy, the message is clear: the rise of UKIP has resulted in a high degree of uncertainty to the outcome of future British elections.
What do European elections tell us?
Local elections are not the only elections that can give us information about future electoral performance. There haven’t been enough European Parliament elections (particularly under the current electoral rules – which were only introduced in 1999) to conduct the same sort of analysis as I have done for the local election results, but they can still give us clues about what might happen in 2015.
An important thing to note is that European Parliament elections are much worse predictors of success at future general elections than local elections. On average the vote shares for each party from the 1999-2009 are 8.75 percentage points different from their vote shares at the next general election, almost double that of local elections. European Parliament elections also seem to be becoming worse predictors of general election results (the same is not true for local elections) – the difference between vote shares at European and general elections for the 1999 EP election was 7.5, 8.5 for 2004, and 10.3 in 2009.
We can however look to different data to tell us about how voters cast their votes in different elections. This could tell us a great deal, particularly about how well UKIP might do at the next election. Data from British Election Study panel surveys shows that the main problem UKIP has faced in translating its success from European Parliament elections to general elections has been retaining voters, whether because some UKIP voters only vote UKIP at European Parliament elections in protest and the return to their ‘normal’ party for general elections or because the nature of the British electoral system incentivises voters to cast their vote for one of the existing main parties rather than a new entrant. After the 2009 European Parliament elections, only 28% of UKIP voters in the BES survey said they were planning on voting UKIP again at the next general election. The 2010 wave of the survey reported an even smaller proportion in terms of actual vote choice, with only 21% of 2009 UKIP voters voting UKIP at the 2010 general election. Data from the first wave of the new BES, conducted in February and March of this year suggests that UKIP may be overcoming this problem with 58% of those saying they were likely to vote UKIP at the European Parliament election also saying they would vote for them at the next general election. UKIP voters are also more certain about their voting intentions with those saying they ‘don’t know’ who they will vote for dropping from 15.27% in 2009 to 7.83% in 2014. Although this is a substantial increase over past levels it still lags well behind the three main parties – Labour and the Conservatives both retain 94-95% of their European Parliament votes, whilst the Liberal Democrats retain about 80%. This suggests that UKIP look to do much better than their previous performances in general elections, but certainly not as well as they did in the European Parliament elections. When you take differing turnout into account these numbers give some credence to the earlier UKIP forecast of about 11.6%.
The four local and one European Parliament elections that have been held under the current government cannot give us exact answers about how the votes will fall at the next general election but they can give us clues. My model predicts that the Conservatives are likely to come out on top, though there is perhaps more uncertainty about the outcome than at any previous election. Although there are problems with predicting the Liberal Democrat vote the model suggests that they will be substantially down on their 2010 vote but will perhaps not face the routing they did at the European Parliament elections. UKIP look likely to do well at the next general election in terms of votes, though nowhere near as well as they did at the European Parliament elections. Perhaps the real question is whether UKIP can turn votes into seats and on that task they look to be facing an uphill battle. Other analysis suggests that UKIP would need to do very well – certainly much better than the results here suggest they will – in order to gain even a handful of seats in the next parliament. Regardless of whether they do though, one way or the other, the UKIP vote is likely to have serious consequences for the outcome of the next election.