Chris Hanley & James F. Downes
Introduction
First the Scottish referendum, then Israeli election and now the General Election in the UK. We are only six months in and it has already been a nightmare year for opinion polling companies. This paper discusses why voting behaviour is becoming harder and harder to predict and why seat share predictions should be taken with a pinch of salt.
What happened at the ballot box?
In May, the Conservative Party received 37% of the vote but received around 51% of the seats in parliament. Labour did slightly worse in terms of vote share (29%) but still did proportionally better in terms of the amount of seats they were able to claim. The Scottish National Party was the only other party to benefit from this electoral system with around 5% of the votes translating into around 9% of the seats. The Liberal Democrats, as expected declined considerably, trailing behind UKIP with only 8% of the votes and just a few parliamentary seats to show for it. All other parties, as usual, fell short of obtaining more than a handful of seats.
Table 1- Vote and Seat Share for Political Parties in the 2015 British General Election
A number of political commentators and key political scientists have outlined how the current First Past The Post (FPTP) is in inhibiting the vote share of smaller political parties, with Professor Matthew Goodwin , leading expert on UKIP outlining how ‘UKIP has done brilliantly. And totally failed. How can that be?’ Goodwin attributes this largely to the mechanics of the FPTP system.
First Past The Post: The Mechanics
This almost 'unfair' distribution of seats is not uncommon in modern liberal democracies. One only needs to look at the current state of gerrymandering in the United States to understand that. In the United Kingdom, we use the Single Member Plurality System by which each constituency, of around 70,000 residents, elects a person to represent them in Parliament. For that person to be elected, he or she just needs a plurality of the votes - which essentially means just more than anyone else.
For example, in a heavily contested constituency (known as a marginal) where the electorate are strictly split between Labour, Conservative and the Liberal Democrats, the winner would just need 34% of the vote (around 24,000 votes). 24,000 votes in the case equates to one seat. However, in a 'safe seat', where the electorate within a constituency are politically homogenous, an MP could be elected with more than twice as many votes as in the marginal. In this case, 48,000 votes would still only result in one seat.
This therefore underlines why it is so important for a political party to have concentrated electoral support in order for it to be successful. Parties like the Green Party often fall victim to such a system which reaps few rewards for those who have widespread support but are unable to retain a plurality in a given region. At the other end of the scale, this election marked a breakthrough for the Scottish National Party who were finally able to become a credible force by securing pluralities in most of the Scottish constituencies, benefiting from the combined disenchantment with the Scottish Labour and Liberal Democrat Party in Scotland through emphasising the nationalist-unionist cleavage.
Proponents of this system often defend this 'inequality' between vote share and seats with its ability to produce strong majorities and in turn strong governments. A by-product of this is that it also tends to heavily favour a two-party system. In political science, Duverger's law argues this since small parties have to merge with others in order to become politically relevant. It also posits that since plurality systems do not reward parties that consistently come third place, they tend to become extinct. Proportional systems on the other hand often result in multiparty politics and a fragmentation of power in the executive and legislature.
Historical precedence
There are two key phases in British politics that are used to define the patterns in voting behaviour as a result of this electoral system: the periods of alignment and dealignment. The period of alignment is said to encompass British elections up to the end of the 1970s, in which Peter Pulzer outlined that 'class is the basis of British party politics, all else is embellishment and detail'. In this period, the Labour Party had traditionally appealed to the working classes through the strengthening and support of the trade unions and social welfare system, whilst the Conservative Party would traditionally count on the votes of the middle and upper classes by keeping taxes low and safeguarding the status quo.
The era of dealignment was ushered in by the successful Thatcherite campaign in 1979. Social class was no longer such a defining characteristic of British elections as swathes of working class voters pledged their support to the Conservative party in this election and again in 1984. Pulzer's theoretical model began to give way to issue-based politics, whereby voters were assumed to cast their ballot one way or another based on the most important or salient issues of the day.
In political science, Inglehart's theory of post-materialism feeds into this notion whereby society has moved past the focus on basic needs and towards a focus on 'material luxuries' and consumerism. It follows that the electorate has become increasingly more educated and with the advancement of the technology, the mainstream media's reach has also never been so large. All this allows for more calculated decisions at the ballot box and has arguably led to an increase in the prevalence of short-term voting factors such as the economy, immigration, social welfare and party leadership. At the same time, recent academic scholarship has also outlined a growing disenchanted group in the United Kingdom that Goodwin and Ford refer to as ‘the left behind blue collar voters’ in British politics.
Arguably, a side effect of the disintegration of long-term voting behaviour is increased electoral volatility amongst the electorate which has been particularly evident in the last two general elections. This general pattern of electoral volatility can also be found in a number of contemporary European democracies, with parties of the traditional left, right, and centre under increasing threat from challenger and insurgent parties of the extreme right and extreme left alike. Political scientists such as Peter Mair have outlined the long-term disconnect in the representation between the voter-party representational link in a number of democracies across Western Europe. Figure 1 below shows the overall distribution of change in party performance for incumbent governing parties and the electoral increase of challenger parties in the onset of the 2007/2008 economic crisis in national parliamentary elections of 24 European Union member states. In sum, this Figure shows that traditional parties are under increasing threat from challenger parties.
Figure 1- Change in Percentage Party Vote by Incumbency
2015 Key Issues
As previously discussed in our recent Parliament Street and Ballot Box articles before the election, there seemed to be three core issues that dominated the political landscape: immigration, economy and party leadership. Looking back at the run-up rhetoric and the overall outcome, it is clear that all three issues tended to be important. Immigration, as well as the renegotiation of EU terms, were generally sidelined in the debates – both issues were only discussed when Farage or Sturgeon were on stage. Despite this, it was clearly a defining factor in this election - but also one that needed careful management. If Cameron moved too far right on the ideological spectrum, he would have risked losing key votes and the support of his closest allies; too close to the centre ground and UKIP would have likely been an even larger threat.
However, most political commentators focussed on the impact of the economy and more specifically, how many on-the-fence voters swung to the Conservative Party as a result on election day. On May 7th, all polls had Labour and the Conservative neck and neck in terms of vote share, but the actual result was starkly different. The first exit polls came in and the Conservative Party were not just edging ahead but were on track to achieve a clear majority.
The economy was the Tories' strongest hand. When Cameron entered into office there were little signs of improvement, but come election time, the UK was leading the economic recovery in Europe. In our previous article, we underlined the importance of personal economic expectations and that macroeconomic recovery is not necessarily a deal breaker. However, although many of those polled before the election said that they feel worse off than five years ago, the public tended to be much more confident in Osborne and Cameron's fiscal strategy than what the other parties had to offer.
A polling nightmare
The May election therefore marked one of the largest blips in polling history with virtually all polling companies and pundits hedging their bets on a hung parliament - some even put Labour ahead. The Guardian interviewed the Managing Directors of all the major polling companies in the UK and all said it was virtually too close to call. Even the renowned pollster Nate Silver said that no political party was likely to achieve a majority. On election day, with a 'larger than normal' sample size, ICM put Labour and the Conservatives at 50:50, and then later on in the day giving Labour a 1 point edge. Most significantly, seat share predictions were even further off.
So what went wrong?
Not many can confidently say they know the exact reason why the polls were so far off but many theories have been circulating. A lot of commentators have hinted towards a large proportion of 'shy-tories' in this election. Although these voters said they would vote for another party, on the day they opted for Cameron, partly resembling the political phenomena that took place in the 1992 British General Election. Whilst the Conservative party is still tainted with the idea of being out of touch and a resource for the rich, the reality is that nevertheless, the electorate still thought they would run the economy better than the others.
More importantly, alluding to something that was discussed earlier - voting behaviour is changing - and has been for some time in British politics. The hardest thing about predicting election results is being able to work out what will happen to the 'undecideds'. Arguably, as a result of an increasingly more educated, wealthy, consumer-driven society, the group of volatile voters is rapidly growing in size. And not only is the size of undecideds growing but as much as 5% are changing their minds in the week running up to the election. Fifty years ago, very few made up their mind on the eve of Election Day. Clearly, long-term voting factors were the dominant force - and polling companies were usually on the button with their predictions. It's definitely not as simple any more.
In recent elections, the undecideds have been polling at around 20% and its worth noting that most polling companies recalculate their vote share percentages by excluding these people. It makes sense to assume that there was a significant swing in this large group of people in the election run up.
A pinch of salt?
With the increasing difficulties of achieving representative samples (and thus larger margins of error), the shy Tory factor and the growing group of late deciders, it is possible to posit at least part of an explanation for why polling companies were so far off the mark. However, vote share was not what shocked people the most - it was the translation into seat share predictions. There are however inherent weaknesses with predicting seat shares and it is both a complex and difficult calculation for forecasters to make.
For example, an average constituency contains 70, 000 people. Providing that sample is not skewed, one would need to interview 400 respondents to achieve a 5% margin of error and thus obtain an accurate reflection of the voter landscape. If one was to interview 400 people in 650 constituencies, you would end up with a sample of size of 260,000 - practically a census. This, for obvious reasons, is not viable for polling companies. Inevitably, shortcuts are taken and algorithms are designed to compensate. ICM for example use constituency level polling when they can but not many companies carry this out for cost reasons. Where there is limited or old data, figures are adjusted to regional data or the national average. And when the national average is off, seats rapidly swing one way or the other as its effect is multiplied. With all the uncertainty surrounding seat projections, compounded with the issues raised above, it is clear that seat share predictions have to be taken with a pinch of salt and more robust statistical models will need to be developed to enhance the polling discipline.
