Think you know which seats to watch on May 7? Think again

There is an old adage in politics that elections are won and lost at the margins. What this means is that small shifts in voting patterns between elections can significantly affect the outcome. For example, the 2000 US presidential election between George W Bush and Al Gore came down to swings in a few key states and was famously decided by only 537 votes in Florida. (OK, so there were allegations of electoral misconduct, lawsuits, recounts, and an eventual decision by the US Supreme Court, but I digress.)

As an election draws near, the media tends focus coverage on these marginal seats because it makes for exciting coverage. Let’s be honest. Who wants to tune into a seat where one party is ahead by 30 points in the polls? Likewise, the political parties strategically invest their resources into these marginal races to maximise potential gains.

So, how do the media, parties, and political pundits generally decide which seats deserve the marginal status and thus their undivided attention? That’s easy. They generally look at the returns from the previous election – 2010 in this case – to rank the constituencies based upon the difference in vote share between the winning candidate and second-place finisher.

Yet, a lot can change in five years. Identification with the major parties is down, and voters seem largely disaffected with what the Conservatives, Labour and the Liberal Democrats have to offer. The surge in support for the Scottish National Party (SNP), the rise of UKIP, and the collapsing base of the Liberal Democrats make it surprisingly difficult to use the 2010 results as a guide, as none of these huge changes to the political landscape were remotely visible then.

But all is not lost. One of the defining features of this parliament has been the explosion of constituency-level polling. Led by Lord Ashcroft, the former deputy chairman of the Conservative Party, constituency polling provides a close-up picture of what is happening on the ground in a quarter of the UK’s constituencies.

Ashcroft’s non-partisan polls are highly regarded not only because they are conducted by reputable companies but also because he discloses details about each poll for public scrutiny. This may seem like a minor achievement, but constituency polls are quite expensive – £5,000 to £7,000 per poll by some estimates – which discourages most media outlets from conducting them on a regular basis. Instead, they are often forced to use national polls with far fewer numbers of respondents in each constituency to make inferences about the outcome of these local elections.

To date, Lord Ashcroft has conducted 252 polls in 167 constituencies (there are 650 constituencies in the UK). Considering that each one of these local polls includes a sample of at least 1,000 respondents, Lord Ashcroft has surveyed more than 252,000 Britons. This is no mean feat.

Finding the real marginals

To determine where the real marginal seats are in the 2015 UK general election, I compiled the results from every Ashcroft constituency poll beginning in May 2014 to the present day. When multiple polls were available for a single constituency, I opted to include the latest results because we expect them to be a better approximation of the true election outcome. I then combined this local polling data with the 2010 UK general election returns for each constituency, so that we can compare the difference in predicted versus actual vote share by election year.

One way to test whether 2010 is a good benchmark for the upcoming 2015 election is to rank the constituencies by the difference in vote share among the top two parties. This rank, which I have labelled “marginality rank”, provides us with a measure for how close a given election is in 2015 and 2010, respectively.

The full ranking of marginal seats is available in this table. You can see the top 20 in the chart below.

Top 20 Marginal Seats in 2015 (Ashcroft Polls).

One of the first things that jumps out in the data is that for 74 of the 167 poll snapshots, or 44.3% of the total cases, the Ashcroft polls have a different party ahead compared to who won in the 2010 election. An important caveat is that these polls consist of the vote intentions of a random sample of potential UK voters, not the whole constituency.

As such, the estimates provided by the Ashcroft polls are subject to sampling error of plus or minus three percentage points. What this means is that in 95 out every 100 samples drawn, estimated proportions will be no more than three percentage points away from their true values in the population.

Given this margin of error, any differences between the two leading parties less than six percentage points in the Ashcroft polls could be due to random chance rather than true differences in vote intentions for a local race. But even if we only consider those poll results where the difference between the vote share between the two leading parties is greater than six, that leaves us with a different outcome in 40% of the races than we would have naturally expected given the 2010 returns.

More importantly, it is quite clear that the marginal seat rankings for 2015 are very different than those in 2010. In fact, there is not a single match of marginal seats between elections in the top 20 closest races in 2015. And, there are only seven seats within three percentage points in 2010 that fall within the top 50 closest races in 2015.

Some seats – Sheffield Hallam is a good example of a race that would never normally be considered marginal – now look like knife-edge battles in the Ashcroft polls. In fact, the Liberal Democrat Leader Nick Clegg won the 2010 election by nearly 30 points to the second-place finisher. Now he is down by one point to the Labour challenger. (My colleague Charles Pattie has written a an analysis of Nick Clegg’s chances.)

Another stark example is the race in Hampstead and Kilburn, which was extremely close in 2010. In that election, Labour’s Glenda Jackson won the seat by the narrowest of margins – just 42 votes of the 52,822 that were cast. This constituency ranked first in marginality status for 2010. Yet, recent data suggest an easy win for Labour – a poll conducted last August showed the party was ahead by 17 percentage points over the Conservatives in that race.

In short, the lessons of 2010 do not seem to provide us with a good idea of where the marginal seats are for 2015.

Mark Twain once quipped: “History doesn’t repeat itself, but it does rhyme.” What do you do when history doesn’t even rhyme? Make like Lord Ashcroft and collect better data.

Todd K. Hartman does not work for, consult to, own shares in or receive funding from any company or organisation that would benefit from this article, and has no relevant affiliations.