In Defence of Liberty

Driven by data; ridden with liberty.

In Defence of Polling

in-defence-of-polling-wikimedia

These are the polls between the general elections in 2010 and 2015. (Edited: Wikimedia Commons)

Despite most polls showing a very close race between the Conservatives and Labour, the Conservative Party ultimately triumphed in the 2015 General Election, and now lead a majority government in the House of Commons. The British Polling Council released its preliminary findings into why these errors occurred, before a full report is published in March 2016 [1].

In these inquisitions, it is worth recalling that many electoral trends were correctly identified by the pre-election polls: the collapse of the Liberal Democrats, the dominance of the Scottish National Party in Scotland, the limited nature of the Greens, and the middling performance of UKIP.

(Video: BBC News)

Severe underestimation

What is remembered is the severe underestimation of the Conservative lead. On the eve of the election, the BBC poll of polls suggested that 34% of voters were backing the Conservatives, whilst 33% were voting Labour. In reality, the Conservative vote share was 36.9%, whereas Labour took 30.5% of votes cast [2].

in-defence-of-polling-bbc

The Conservative vote share was underestimated. (Source: BBC)

These election results concern the whole of the United Kingdom, but polls typically only gather respondents from Great Britain, and not Northern Ireland. The difference between the vote percentages for Conservatives and Labour is slightly wider in Great Britain: these shares are 37.8% to 31.2%, respectively.

The preliminary findings are that the polling methods employed by polling companies yielded unrepresentative samples of the British public. Polling firms do not undertake random sampling of the whole population, as the Office for National Statistics or other organisations do. Polls often have certain demographic thresholds that have to be fulfilled, from contacting as many people as necessary, and then these results are weighted to reflect the general population.

Whether through online or telephone polls, Conservative voters were harder to get in contact with, as suggested by the research institute NatCen. This differential non-response meant that those who could be found quickly, as most polls only took two or three days to gather their data, did not represent the whole electorate.

The possibility of a ‘late swing’ remains open, but the findings suggest its effect was modest and inconsistent. The panel also could not rule out ‘herding’, where polling companies configured their results so they deviated less than would be expected from the sample sizes.

The recess

The exasperation of politicians and journalists at the failure of polls to suggest a substantial Conservative lead has heralded calls to either ban opinion polls in the election period, or to otherwise ignore them [3]. Either idea would be folly.

It is crucial, if we are understand how and why people vote the way they do, to know their votes and beliefs. In a country with a democratic government, public opinion is often interest to political parties, with discernable majorities used to buttress or undermine policy proposals.

In the absence of opinion polls, as was observed during the recent Oldham West and Royton by-election, the recess left by estimated opinions was filled by conversations with candidates, quotes from potential voters that the journalist just happened to meet, and ceaseless speculation of a UKIP victory [4]. Unbound by any data analysis on voting intentions within the constituency, the prognostications of journalists and commentators were heeded like ancient soothsayers.

Whilst polling companies should seek to improve their methods of data collection and aggregation, the best way to know which party is most popular is to measure it.

in-defence-of-polling-xkcd.jpeg

It is a constant surprise. (Source: XKCD)

References

[1] BBC, 2016. Election polling errors blamed on ‘unrepresentative’ samples. Available from: http://www.bbc.co.uk/news/uk-politics-35347948 [Accessed: 20th January 2016]

[2] BBC, 2015. Election 2015 Results. Available from: http://www.bbc.co.uk/news/election/2015/results [Accessed: 20th January 2016]

[3] BBC, 2016. Should polling be banned before an election? Available from: http://www.bbc.co.uk/news/uk-politics-35350419 [Accessed: 20th January 2016]

[4] Routledge, P., 2015. Why the Oldham West by-election is a battle that Labour HAS to win. Daily Mirror. Available from: http://www.mirror.co.uk/news/uk-news/oldham-west-election-battle-labour-6944091 [Accessed: 20th January 2016]

Advertisements

Information

This entry was posted on January 25, 2016 by in National Politics and tagged , , , , , .
%d bloggers like this: