Copeland by Ward

Posted 13 February 2017

The by-election in Copeland will take place on 23 February 2017, following the resignation of Labour MP Jamie Reed to leave politics.

The Conservatives came a close second in the seat at the last general election, whilst UKIP was in a relatively distant third place. Much attention has focused on whether the Conservatives can gain the seat at the by-election.

Making predictions for by-elections is a fool's game, since they behave very differently from general elections. Previous general election results may not be a good guide to by-elections, and by-election results are famously a very poor guide to the next general election.

However, we can look at some political measures around Copeland to get a flavour of the political situation there.

The current Electoral Calculus prediction for the seat in a general election (not the by-election), as at 13 February 2017, is:

This has the Conservatives winning the seat by a thin margin of about 1pc. However this may not be a good guide to what will happen. By-elections are strange events and can easily produce unexpected results in one way or another.

On one hand, Labour are not doing well in the national polls and the local MP has resigned from politics. Both of these point to a less motivated Labour vote. On the other hand, governing parties are usually more unpopular than opposition parties during the mid-term, and rarely gain seats at by-elections. There have only been about a dozen instances of this in the last hundred years (see list).

To win, the Conservatives have to keep their vote solid, whilst working for a low turnout amongst Labour voters. It would also be useful for the Conservatives to squeeze UKIP for some votes "on loan".

To analyse this, we need to introduce some major new modelling of political behaviour at a local level which Electoral Calculus has performed.

Ward Level Analysis

Electoral Calculus has recently conducted a major new analysis of political attitudes and voting patterns at a detailed local level. The relevant level here is the local council ward, in which people cast votes every 4 years or so for their local district or borough councillors. There are around 8,700 wards in Britain (excluding Northern Ireland), with an average of 5,000 voters in each one. Wards are naturally sub-parts of parliamentary seats, so information gained at a ward level can be very helpful for the ultimate analysis of an entire seat, and also provides valuable campaign information.

Using "machine-learning" and big data techniques, well-known to mathematicians as linear regression and Bayesian analysis, the following calculations can be performed to get three very useful datasets.

First Dataset : General Election result by ward

From the 2015 General Election, recent local election results and the ward/seat geography, we can use the Electoral Calculus prediction model to allocate general election votes from the seat level to the ward level. This useful information is not published by the electoral authorities. But using our model means that we have an estimate for how people voted in each ward at the general election. This ward-level voting breakdown is shown below its map on each seat details page, such as Copeland.

Second Dataset : EU Referendum result by ward

The EU Referendum result on June 2016 was not just politically significant, but also contains a wealth of detailed political information. The relative vote shares of Remain and Leave across the country can help identify areas of UKIP strength and weakness, but also underlying political attitudes especially on the international / national axis of the Electoral Calculus 2D Political Map.

The EU Referendum reported results by District (Local Authority), which is the sum of all the wards within it. To allocate EU votes from the district down to its wards, we need a political/demographic model of voting behaviour. In this case, we used Electoral Calculus ward-level voting patterns plus ONS demographic data which was fed into a machine-learning algorithm to calculate estimates of each ward's EU referendum vote.

These data can be seen at the seat level on each seat details page, and at the ward level information shown on the results page of the 2D Plot My Position feature.

Third Dataset : 2D Political Attitudes by ward

Finally, we can estimate the 2D Political Attitude of each ward. This positions each ward on the Electoral Calculus two-dimensional map. The first axis is the economic left-wing/right-wing dimension, which runs from 100° Left (big government, pro-workers) to 100° Right (low-tax, pro-business). The second axis runs from internationalism (pro-immigration, EU single market) to nationalism (anti-immigrant, Britain first) denoted as 100° Int and 100° Nat respectively.

Using the valuable Social Market Foundation / Opinium "Dead Centre" survey on political attitudes, ONS demographic data, as well as the first and second datasets above, we can make another machine-learning algorithm to calculate estimates of each ward's 2D political position.

More details of this dataset and what it means nationally are the focus of the Electoral Calculus article "The hidden wiring behind modern politics and where it gives UKIP a chance to beat Labour".

Case Study: Copeland

We can bring all these data together in a practical way by looking at a single seat. Choosing Copeland, we can start with a list of all the ward (or part-wards) which lie within Copeland, and see their values from the three calculated datasets:

• General election result — summarised by the "winning" party which got the most votes in the ward
• EU referendum result — expressed as the percentage of voters in the ward who voted to leave
• 2D political position — given by left/right co-ordinate and international/national co-ordinate

The table below shows all the wards in the seat, along with these quantities (sorted by EU leave share):

Politically, the General Election "winner" in each ward is fairly balanced. The Conservatives have 16 wards, compared with 12 for Labour and only one for Ukip.

The EU Referendum vote is a medium-strength for Leave. Overall Copeland voted 60pc to leave, which puts it in the top-quarter of Leaving seats across the whole country. Across the constituency, only a couple of wards are significantly more extreme. The ward of Arlecdon has an estimated EU leave share of 75pc, and two others had shares of at least 70pc, which puts them all in the top tenth of wards by EU leave share.

The 2D Political positions show two things clearly. Firstly, that Copeland is fairly balanced between left and right. The average of the seat (3° Left) is only very slightly left of centre. Most wards lie in the centre ground between 10° Left and 10° Right. But some wards are more politically distinct with seven strong left-wing wards (such as Sandwith) and one right-wing ward (Bootle).

The international/national axis for is also fairly balanced. The average of the seat is 7° National, which is not an unusual position and puts Copeland only in the most nationalist third of all seats. Only one ward (Arlecdon) has a position more than 20° National. But the seat is Nationalist overall, with only five mildly internationalist wards, which will not help Labour.

Conclusions

Copeland is a Labour seat with a modest Labour majority. At a General Election, the seat would be very marginal.

The seat is relatively balanced politically, being fairly centrist economically and mildly nationalist. The Conservatives could gain the seat at the by-election, but they would have to

• Squeeze the other anti-Labour parties, especially UKIP
• Maintain the Conservative vote in naturally Conservative wards (16 wards out of 29)
• Encourage voter defection or at least absentions from Labour wards which have a nationalist position and are not too left-wing (Kells; Egremont South; Harbour; Egremont North; Moresby)

It is difficult to put a probability on this, because there are too many factors which can't be quantified. Politically, a Labour victory would be a helpful result for the Labour leadership under Jeremy Corbyn, without hurting the Conservatives much.

However, a Conservative victory would be bad news for Labour, since government gains at by-elections are so rare.

©2017 Electoral Calculus