Who has most to lose from UKIP?

When UKIP first emerged in British politics, there was a certain grim satisfaction among people on the left. Sure UKIP’s popularity was rather depressing but at least now the right-of-centre vote would be split as the left-of-centre vote had been for years.

Of course UKIP’s rise turned out to be a more complicated phenomenon than initially thought. It became clear that UKIP were performing well in traditionally Labour areas, then when the Liberal Democrat vote collapsed in the 2015 general election it seemed that UKIP had even made inroads among their voters too.

Now UKIP have arrived on the Welsh political scene, we can begin to evaluate where they seem to be picking up votes from. Do Plaid Cymru have anything to fear from them? Has Welsh Labour’s success at wrapping itself in the flag, relative to their English and Scottish sister parties, insulated them at all? What about the poor old Lib Dems?

The figure below shows the relationship between Labour share of the vote (on the X-axis) and UKIP share (on the Y-axis). As  indicated by the climbing regression line (which doesn’t include Arfon or Aberconwy, where UKIP didn’t field candidates), Labour and UKIP are competing over many of the same constituencies.


The second figure shows change in Labour’s share of the vote against UKIP’s. Here there is a tendency for seats where Labour lost a greater share of the vote since 2011 to be seats where UKIP did well. This seems consistent with Labour losing voters to UKIP, but could also be UKIP’s message playing well in the same places where Labour turnout is being depressed. Bear in mind the usual caveats about correlation not necessarily implying causation…


What about the other parties? Maybe this is a problem for everyone? It seems not. Plaid Cymru’s equivalent plots show a pretty flat (slightly negative if you squint) relationship between their vote share and UKIPs…


…and a similarly flat relationship with change in Plaid’s share of the vote. Bear in mind though that in two relatively strong Plaid constituencies (Arfon and Aberconwy), UKIP didn’t run. We would know more about the interaction of Plaid and UKIP if they had.


The Conservatives seem to be competing for very different seats to UKIP, with a clear negative relationship between their vote shares…


…and surprisingly they seem to be doing best where UKIP did well (although it’s worth pointing out that they generally lost ground – perhaps they’ll have trouble gaining steam with UKIP around, we shall see).


Finally, the Lib Dems. The four seats where they are competitive are all seats where UKIP did fairly poorly…


…but they also lost more votes where UKIP did well! This could be UKIP taking seats off the Lib Dems, but it could also be people going for another party, or indeed depressed turnout.


So to summarise. Labour appear to be vulnerable to UKIP and also competing over the same turf, Plaid seem to be relatively unaffected by them, the Conservatives seem to be competing over different seats and holding their ground better where UKIP are doing well, and the Lib Dems appear to have been vulnerable to UKIP, but their remaining areas of (relative) strength seem unfriendly to UKIP. It seems that Labour are the party with most to lose from UKIP’s foray into Wales.

Hemmed in

I’m resisting the urge for this post to be my opinions on today’s dramatic events in the Senedd –  I’m sure plenty of people who know much more than me on the topic will cover this over the next few days so I’m going to try to approach the broader issue with a quantitative flavour.

Today’s events have demonstrated how hemmed in Labour find themselves by rival parties. In many respects this might be helpful – Labour are certainly keen to promote a perception that Plaid are ganging up with UKIP and the Tories and have generally been very successful at dividing and conquering their opponents. In other respects, however, it might turn out to be more of a challenge.

MajorityMargins The plot above shows the percentage vote for each party of the Y-axis and how many percentage points that party was behind whoever won the seat on the X-axis. So if the party won the seat, it’s at the far left of the panel (and the constituency name is in the party’s colour).

The plot shows that the top target seat for every non-Labour party represented here is currently held by Labour – Gwent and Llanelli for Plaid, ‘The Vale of [insert vale here]’ seats for the Tories, Torfaen and Caerphilly for UKIP, and Cardiff Central for the Lib Dems.

Labour on the other hand seem to have an odd gap in their vote share between the seats they won and all the others, except the Rhondda and Arfon. Maybe this is evidence that Carwyn Jones is a lucky general after all, but it does suggest limited scope for Labour to go on the attack in very many seats. Even in the two that they look like they have a good chance it’s not clear how their chances would be. What happens next election in the Rhondda is frankly anyone’s guess – it will depend a great deal on how Leanne Wood performs. Labour’s share of the vote in Arfon (the constituency I live in) feels like it may be a less solid foundation than it looks – UKIP didn’t run a candidate here and Plaid are certainly well dug in here with an absolute majority.

Things can famously change quickly in politics but it certainly looks like Labour are set up to be on the defensive for the time being. They have limited target seats, and a number of tricky battles on multiple ideological and geographic fronts. It will be difficult to compose a message to pick up votes  from all of these parties at once. Labour’s days of governing alonemay be numbered. There was clearly an emotional response to the First Minister vote this afternoon but Labour would be well advised not to burn too many bridges. Coalition may well be forced upon them before long.

What if the list seats weren’t split into regions?

As you are no doubt aware, the Welsh Assembly elects members using the d’Hondt method of additional members. Essentially the list seats are preferentially allocated to parties who have fewer constituency seats than is proportional to their vote. Here’s a longer explanation. But rather than being allocated across Wales as a whole, the list AMs are allocated separately in five regions – North Wales, Mid and West Wales, South Wales West, South Wales Central, and South Wales East.

It’s not clear why we do this. The constituency AMs already represent specific areas and the list AMs are supposed to compensate for the lack of proportionality in the first-past-the-post system.If this is the case, why not elect them across Wales as a whole?

Under the existing regional list system, the AMs were allocated as follows:

  • Labour: 2
  • Plaid Cymru: 6
  • UKIP: 7
  • Conservatives 5

What if we ran the system over Wales, ignoring regions? The results look like this:

  • Labour: 0 (from 319,196 list votes!)
  • Plaid Cymru: 5 (from 211,549 list votes)
  • UKIP: 7 (from 132,138 list votes)
  • Conservatives: 3 (from 160,846 list votes)
  • Liberal Democrats: 2 (from 65,504 list votes)
  • Abolish the Welsh Assembly Party: 2 (from 44,286 list votes)
  • Greens: 1 (from 30,211 list votes)

A few things jump out:

  1. A greater number of parties get AMs, including rather surprisingly the Abolish the Welsh Assembly Party.
  2. Labour get no list AMs at all, despite getting over 32% of the list vote.
  3. Only one party with AMs would stand to profit from such a change (the Lib Dems, who would get two list AMs).
  4. UKIP do exactly the same under either system.

At the end of the day, it depends on what you want a voting system to do. Proportionality, ties to locality, support for strong governments, and simplicity (as my R code to run the d’Hondt process will attest) are somewhat mutually exclusive. The system as it stands seems relatively friendly to Labour, who of course set the assembly up. That said, it would seem perverse for them to get no list seats at all from a full 32% of the vote. What is clear though is that for better or worse, the regional character of the current list system acts as a filter on smaller parties, to the advantage of the larger parties. Which  is why I would be surprised if they adopt this system any time soon.

Carwyn Jones – The lucky general?

That Carwyn Jones is a lucky general is fast becoming a cliche of Welsh political punditry. Richard Wyn Jones of Cardiff University covers it here in the Guardian and Vaughan Roderick mentioned it a few times on S4C’s coverage and on his BBC Radio Wales programme. It certainly feels right – the numbers in close elections just always seemed to break Labour’s way. Llanelli, Cardiff Central, and the Vales of Glamorgan and Clwyd went to Labour on pretty slender margins.

Then again, Labour just won a lot more constituency seats than anyone else – 27 against just 13 for other parties. This being the case, it would be weird if Labour’s share of marginal seats wasn’t also higher than the other parties. If Labour were absolutely sweeping the board (a la the SNP in the 2015 General Election), we might expect all their opponents’ seats to be marginal but that’s not the case.

I’ve plotted the winning majority (on the y-axis) against Labour’s share of the vote (on the X-axis). The colour of the constituency’s name shows which party won it.


There’s certainly a bunch of seats which Labour won on fairly small majorities: Llanelli, the two ‘Vale of’ seats, Cardiff West. Then again, they also racked up huge majorities in Ogmore, Swansea East, and Cardiff South.

To put some numbers to this, I looked at the mean majorities by party:

  • Labour:3860
  • Plaid:5774
  • Conservative:3601
  • Liberal Democrats: 8170 – Although this value is not so much a mean as the majority of their sole seat, so not very representative.

So based on this, one could argue that it is in fact Andrew RT Davies who is the lucky general, whose Conservatives squeaked home in the most seats! Indeed the Conservatives’ majorities were consistently lower, whereas Labour’s ran the full range.

I also ran some simple statistical tests comparing the majorities of Labour seats to non-Labour seats. A bootstrapped t-test found that the difference between Labour (3860) and non-Labour (4955) seats  was no larger than you would expect by chance. The number of constituencies we have to analyse is fairly small and there are probably cleverer ways of looking at this question (perhaps looking closer at that apparent cluster of constituencies in the bottom right of the plot), but it suggests that there isn’t a major difference.

So on the surface of things, it is not clear that Labour are any luckier in this way than their rivals. There are of course other ways in which Carwyn Jones may have been lucky – perhaps the Tata crisis gave him a chance to present himself as statesmanlike, perhaps Conservative divisions on Europe hurt them, perhaps UKIP were distracted by the unpcoming referendum.. But in this narrow sense of luck, the numbers don’t seem to bear it out.









Hello world

In the small hours of election night, addled by lack of sleep, I decided I’d like to start a blog on Welsh politics. Here it is (I’ll get better at this I swear). It’s going to be one of those, oh so modern, data journalism endeavours because I’m that way inclined.

In the interest of transparency I’m centre-left and have voted Plaid Cymru, Labour, and Liberal Democrat in the three UK general elections I’ve been old enough to vote, which I suppose makes me a floating voter. I’m originally from London but moved to North Wales for university and went native (which explains the Plaid Cymru thing). I’m not planning on writing in a partisan way, but that’s easier said than done.

Anyhow, thank you for reading. Again, I’ll definitely get better at this. Comments welcome.