Monday, 7 March 2011

The Geography of the Welsh Referendum


Harry has already covered the Welsh referendum result but I thought I’d look into the geography of the result. Whilst the referendum passed it did not pass in a uniform way.

Firstly, it is fairly clear that a big factor in the strength of the yes vote was Welsh language. I threw the 2001 census data for each council area’s percentage of people who had ‘one or more skills in Welsh’ into ace statistical program SPSS and attempted to correlate it with the Yes Vote. I got a scatter graph that looked like this:

Now for the statistically minded amongst you this graph has a Pearson’s r of 0.402 and a significance of 0.063, which, in English, means that it isn’t statistically significant. Nonetheless, it is close (a significance of less than 0.05 would be significant), and this is a blog, not an academic journal, and I can do what the hell I like. I am immediately struck by several thoughts, however:

  • Obviously Wales and the Welsh are more complex than a yes vote being linked to one factor.
  • If I went to a more localised level, say council wards, for the data, I’d probably have a much better link.
  • The census is almost ten years out of date. When the 2011 census is released it may provide a better correlation.
Now for our purposes those councils 'on' or near the line basically behave as expected considering their Welsh language population. Those 'above the line' voted less in favour than would be expected considering their Welsh language population and those 'below the line' voted more in favour.

Those on the line mostly make sense. Gwynedd is the most North-Westerly of the mainland council areas and demonstrates the highest levels of Welsh language take up. The council area has been a centre for controversy and protest over people from outside the region moving into second homes and demonstrates a startlingly strong Welsh identity. Gwynedd’s mirror opposite is Monmouthshire, the only council area to vote no (albeit by 320 votes). Monmouthshire has the lowest Welsh speaking population and has long been a source of discussion over whether it is Welsh or English. Some have suggested a Monmouthshire referendum on the subject, though I don’t really think 320 votes justifies it.

Anglesey demonstrates the second highest level of Welsh language, but a relatively low yes vote. Anglesey has an extremely strong local identity, because it is a remote island. Politics in Anglesey is notoriously localised (for a look into what we mean Britain-Votes recommends ‘The Druid’ blog). The Welsh Assembly, has, at times, been accused of centralising powers away from councils into its own hands. With Anglesey off the North West coast of Wales, too, Cardiff likely seems like a remote place, far from local concerns.

In the South of the country we see a much more emphatic ‘yes’ than we would expect based on Welsh language stats. The big group clustered in the bottom right of the graph is almost entirely in the very South of the country, with the exception of Flintshire and Wrexham, which are in the North East (and, with the exception of Swansea, those two deviate the least from the line). The thing all these areas hold in common is that they are traditional Labour heartlands. While all four parties were in favour of the referendum (though the Tories were a fair deal more agnostic than the other three) there is no point in denying the fact that currently there is a Labour-Plaid Cymru coalition governing in Cardiff Bay whereas there is a Conservative/Liberal Democrat coalition governing in Westminster. While I don’t think the Welsh Assembly poll taken on Monday should be taken as gold, there is an interesting breakdown of voters by party allegiance in it:

I will vote YES on Thursday: Con 25% Lab 79% Lib Dem 45% Plaid 91%

I will vote NO on Thursday: Con 75% Lab 21% Lib Dem 55% Plaid 9%

The fact that Plaid supporters were most in favour is no surprise, nor is Conservative Party supporters being most against, but the fact that many Lib Dem supporters, the 'party of localism' are against is to my mind. This suggests, to me, a partisan effect. I saw one Tory tweeter say on Wednesday that: "If you [are] Welsh and are sick of Labour control of your nation vote no tomorrow." However, the opposite is also true. Labour do not win in Wales because of a vile socialist plot, but because people vote for them. Clearly the idea of having Labour more in control of Wales than the Coalition is of great appeal to large swathes of the Welsh population. Considering the link between Labour support and the working classes it is likely that if one did a linear regression, threw in the Welsh language stat and a measure of deprivation, say, the unemployment rate in each council area, you’d get a pretty nice result. I could have done that, but I thought it might be a bit too much for a blog! (If any political scientist academic types are reading this, and I know a few do visit us, this may represent the world’s easiest journal article). Even within South Wales though, with Swansea and Blaenau Gwent as outlier’s, there is a link between Welsh language and the yes vote. With relatively strongly Welsh speaking Neath and Port Talbot voting yes at 73%, and Torfaen and Cardiff less so. Blaenau Gwent’s result may be explained by its extreme levels of deprivation – it had the highest level of unemployment in Britain in 2009. I have more trouble explaining Swansea, but it may be linked to the city historically being a centre for English migration into Wales – during the late 18th century Swansea grew by 500% as it swelled from inward migration. For a time it was Wales largest town.

The big exception to the rule on the South Wales issue, which is only slightly below the line is the Vale of Glamorgan, which is a relatively affluent area, which tends to be more of a Labour/Conservative battleground, with the West, in particularly, leaning Conservative. The Vale of Glamorgan Westminster constituency is held by the Conservative Alun Cairns as of 2010.

Moving back above the line we see Powys, which has a fair Welsh-speaking community, but a low yes vote. Powys contains a vast expanse of rural areas. Close to the English border Powys a high number of English immigrants, and high links to England. The area also tends to be fairly Liberal Democrat leaning with the Tories the second party.

Highly Welsh-speaking Carmathenshire and Ceredigion, located in the West of the country all voted pretty close to what we’d expect, as did Conwy and Denbighshire. Lastly we have Pembrokeshire, the most South-Westerly council area. Pembrokeshire demonstrates a strong North-South divide known as the ‘Landsker Line’. This is a linguistic line with Welsh speaking and identity strong in the North and the South known as ‘Little England beyond Wales’ due to the strong local identity, culture and language despite a relative remoteness. Historical research shows this dates from about the same as the Anglicisation of Devon and Cornwall, suggesting that this is a result of maritime influence.

In a sense, the Welsh referendum, therefore, demonstrates the identity lines running through Wales. Whereas it is clear that several factors were relevant in why people voted ‘yes’ or ‘no’ it seems it was a big factor whether they spoke Welsh. This will come as no surprise to any student of nationalism, being that a nation, by definition, has a shared identity, history and language, and indeed Plaid Cymru has often been described as first and foremost a Welsh language party. Language is not the sole source of identity, however, nor is identity the sole reason why someone may vote one way or another, though they often are a big factor. Nonetheless, I think the electoral geography of the Welsh referendum has a great deal to tell us about Welsh politics.


  1. Couple of thoughts.

    The opposition by Lib Dem voters is no surprise. It was ever thus. Outside of Mid Wales Lib Dem support comes largely from people with weak Welsh identification and their activists are largely drawn from people who came to Wales as students and chose to stay.

    We can't know - because the data doesn't exist - whether there is a link between parents who chose Welsh-medium education for their children and voting patterns. One suspects there is.

    Places like Rhondda Cynon Taf have around 30% of children in WM schools although the number of adult Welsh speakers is far lower. It may also be a factor in the Anglesey/Gwynedd comparison. Gwynedd doesn't offer English medium education at primary level although there are few secondaries that do. Anglesey has many primary schools where English is the main language of instruction,

  2. Whereas it is clear that several factors were relevant in why people voted ‘yes’ or ‘no’ it seems the biggest was whether they spoke Welsh.

    Nope. The biggest factor was whether they were in Plaid-Labour heartlands or in Con-Lib Dem heartlands. Language is a factor in support for Plaid in the Bro Gymraeg (Ynys Môn, Gwynedd, Ceredigion, Carmarthenshire) but it certainly isn't a factor in Labour heartlands such as Flintshire, which saw a a 20%+ swing in favour of devolution. There are only 4 majority Welsh speaking local authorities - out of a total of 22 local authorities, and of those 4 only Gwynedd had a greater Yes vote than the highest "English-speaking" Yes vote (Neath Port Talbot), which kind of deflates your argument.

  3. Sorry, that should have read 'a big factor', I've changed it now. Whereas the link is weak, and, as I covered in the post, not 100%, it is still evidently there. I do agree that party allegiance does have something to do with the vote spread (something I did cover in the post), and indeed Plaid voters and Welsh language uptake is often considered to be linked.

    That said, at the same time, there is clearly the start of a decent correlation. Neath Port Talbot is not majority Welsh speaking, as you say, but it is the most Welsh speaking area in the South of Wales. As I said, within the South Wales Labour heartland itself there appears to be a pattern of a Welsh language link to 'yes voting'.

    I actually think, in a sense, the graph shows two versions of Welsh identity. On the one hand there is the Plaid Cymru nationalist version, where Welshness is linked to Welsh language uptake. It is based in the rural villages of North and West Wales. The other is a South Wales Welsh identity based around working class industrial communities and their traditional attachment to Socialism and the Labour Party.

    Traditionally there was a third Welsh identity - based around religious nonconformism in the rural regions which was linked to a passionate support for the Liberal Party. Based on the polling figures for party support this may now have completely withered away.

  4. Turnout is more interesting to compare to Welsh speakers - by my figures that does return a statistically significant correlation...

  5. Yeah I did notice a link between turnout too. Like I said, I frankly think that any academic could probably write a pretty quick and easy journal article using some various statistics. If you went to ward level, and got employment statistics, Welsh language statistics, number of people born in Wales, GDP per capita, etc. you'd probably get a pretty damn lovely linear regression up against both the yes vote and the turnout. You might even be able to categorise them by the council areas and then run a factor analysis, possibly producing a measure of Welsh identity (isn't stats fun?).

    Course you could do all that, but I don't recommend attempting to do it in a 1,000 word blogpost.