Click to enlarge.
Last week in the UK's general election, Conservatives pulled off a surprise victory, gaining enough seats to form a government by themselves. With 650 seats and almost a dozen parties winning seats, there's a marvelously rich vein of data to tap. The best beginning may be a simple results maps for Britain, data courtesy of Drew Linzer. (Apologies to Northern Ireland, but I couldn't find the necessary file.) For background on the election in general, read Dave Beard's post.
The first thing you should notice after you click on the maps above is that the results have a basic and familiar urban/rural split. The colors may be reversed from the US, but the concept is the same: a sea of blue with red islands. The Conservative strength is in low density (with large land area) constituencies; Labour's map shows small red dots in a pink quilt along with larger metropolitan splotches. In the north of England, where the rural areas are red, the cities they surround are an even darker red.
We'll come back to this later; for now, join me below for more maps and graphs.
The map shows the Liberal Democrats (orange, upper right) were absolutely devastated through much of the country, with less than 10% of the vote. Their strongest remaining performance can be seen in Cornwall and Devon (South West), Wales, and Scotland.
The UK Independence Party, meanwhile, had a decent showing throughout large swaths of England and Wales, with the exception of many urban areas. A few pockets of regional strength weren't enough to get them more than one seat in parliament, however.
The Green Party shows low support throughout much of Southern England, generally faring better in urban areas.
A big story of the election was Labour's fall in Scotland to the Scottish National Party, above in yellow. A comparison of this year's election results to the Independence Referendum last year, here, shows the SNP appears to have outperformed the yes vote in most places. Plaid Cymru, of Wales, also showed some improvements.
Party Efficiency
UKIP mananged to win 12.6% of the votes, nationwide. But they only won a single constituency - that's 0.15% of the seats. This is approximately four million votes per seat. Liberal Democrats and Green votes were also spread ineffectively thin.
The regional SNP, on the other hand, was able to win 56 seats with just 4.7% of the vote, or 25,972 votes/seat. Conservatives were notably more efficient than Labour, but not as much as SNP. Below is a graph, showing all the parties that won seats.
Click to enlarge.
Change since 2010
The biggest changes are the implosion of Labour in Scotland, Liberal Democrats everywhere, and the rise of SNP and UKIP.
Below are the graphs for two regional parties, SNP and Plaid Cymru:
Click to enlarge.
In order to piece out the division between Labour and Conservatives a little better, the remainder of this post will look at the data for England only, where the lack of SNP and Plaid Cymru will make it a little easier to understand what's going on. Here's the change in vote share for the four main parties in England only:
Click to enlarge.
The degree to which LD support collapsed is immediately apparent, with decreases, typically severe, in every constituency. UKIP, likewise, is also clearly on the move, with increases in every constituency.
But Labour also looks like it has increases in an awful lot of constituencies, moreso than Conservatives. How can that be? Indeed, it's true: Labour increased vote share in 81% of constituencies, on average by 3.6 points. Conservatives increased vote share in 65% of constituencies, with an average of 1.3 points. Standard deviations of the vote share changes were 4.6 and 3.9, respectively.
Yet in the end, Conservatives gained 21 seats in England, while Labour gained just 15. This goes back in part to the earlier graph of Party Efficiency: Labour voters are not distributed in as advantageous a pattern as Conservative voters are.
Back to the maps
Let's take a closer look at the results in England. This time, we'll use a cartogram, with each constituency given approximately the same area, eliminating the 'sea of blue' effect. Also, the scale for each party is now comparable.
Click to enlarge.
The islands of red are now large urban bulges, making it much easier to see what's going on. UKIP, while not popular in London, has made some inroads in urban Northern areas. London itself has a wedge of Conservative support.
Next up, the change from 2010-2015:
Click to enlarge.
Liberal Democrats cratered, of course. UKIP gained support almost everywhere but London. Labour and Conservatives seemed to gain support where their support was already greatest, but it's a little hard to tell, since the changes are swamped by LD and UKIP. Here's a composite that tells us what happened to just Labour and Conservatives:
Click to enlarge.
Demographics
Why do we see the patterns we see above? Clearly, there's an urban/rural split, but there's regional differences as well. Here's some maps of demographic variables to suggest some relationships:
Click to enlarge.
All demographic variables shown have similar, but not exactly the same, urban/rural splits. The highest incomes in England are coincident with the wedge of Conservative support in London. Health in northern parts of England is not as good, where Labour support is generally greater.
There's typically good relationships between these variables and Labour's vote share within regions or counties. Here's an example for the West Midlands region, which on the map has an urban center surrounded by progressively larger constituencies:
The vote share of Labour has a good correlation with the proportion of the population over age 65. But notice the slope: -3. This is an obvious case where correlation is not causation; at least, not direct causation - not unless we think old people are each voting three times! So what is it then? One reasonable hypothesis is simply that younger people prefer to live in urban centers, where Labour voters are concentrated, and older people are more likely to be found outside urban centers. While some component of the correlation could be direct causation, overall it seems like it could serve as a useful stand-in for degree of urbanization. If this is true, we would expect to see this relationship repeated all over the country.
Here, then, are three more areas:
The first, Merseyside, was chosen because it was the strongest Labour enclave in the country. The second, Devon & Cornwall, was chosen because it is one of the more rural areas of England. The third, Berkshire and Buckinghamshire, was chosen as a bastion of Conservative support (Buckingham is excluded as it is the Speaker's seat).
We do indeed see that Labour's support is correlated to the share of the population over 65 in all three regions, indicating it may be a satisfactory substitute for degree of urbanization. With that in mind, here's a set of Labour support maps screened by age: on the left, only constituencies with %>65 that is higher than average are colored in; on the right, only those with %>65 that is lower than average are colored in:
Click to enlarge.
Within (arguably) most regions of the country, we see the darker red areas on the right, and their surroundings in lighter red on the left. There are some areas, however, where this is not the case, such as the region between Sheffield and Nottingham. Notably, this area also has a bit lower health and income than the urban centers next to it. Indeed, for all of England, a two-variable model with just health and age inputs does a fair job of predicting the vote share of Labour within 10 points of the actual results, although that's not terribly useful from a political standpoint!
There's a few more points I'm still looking at with regards to the UK election. In coming days, should they prove interesting, I'll put a few more graphs on twitter, where you can find me at @donnermaps.
Tue May 19, 2015 at 9:18 AM PT: With almost all races having three, four, or more candidates getting sizable chunks of the vote, I took a look at the distribution of the vote. The graph below shows, on the x-axis, the combined share of the vote of the top two candidates. (100 minus this value, of course, is the share of the remaining candidates.)
On average, the top two candidates received 76% of the vote in 2015 - unchanged from 2010, despite the crash of the Liberal Democrats, and the rise of UKIP and SNP.
Note this is quite different from the US House elections of 2014, where 93% of elections saw the top two candidates receiving a combined 95% or more of the vote.