Why Understanding the Electoral Battleground is so Hard

It has become cliché of political commentary of late to say that politics is currently deeply uncertain and very much in flux. Actually, this is not wholly true. On the defining issue of the time, Brexit, opinions have changed little over the last three years: the main change that is evident in public attitudes is an entrenching of the differences between Remainers and Leavers, and a shrinking of the middle ground for some sort of compromise. Indeed, some very interesting evidence suggests that being a Remainer or Leaver has become a deeply-entrenched identity within significant parts of the public in the UK.

What is more uncertain is how all of this links to voting in a possible Brexit-focussed general election. To what extent might the Conservatives, under Boris Johnson, be able to claim back the Leave voters who deserted them in large numbers for Brexit party in May’s European election? And where does the Remain vote go? To Labour, as it largely did (outside Scotland) in 2017, or perhaps in increasing numbers to the Liberal Democrats, Greens, the SNP in Scotland and Plaid Cymru in Wales? And how does such an electoral context relate to the electoral map of the UK: which seats will be in play, and who will be the serious contenders in different places? Understanding this is certainly difficult.

If there is one place where elections ought to be predictable it is Wales. After all, electoral politics here has been predictable for a long time. The last general election won by anyone other than Labour was in December 1918; Labour have come first in votes and seats in Wales at every one of the last 26 general elections. It seemed for a while in 2017 that things would be different – the Conservatives actually led on first two polls. But then Labour bounded back strongly, and even ended up gaining three seats back from Conservatives; meanwhile, as in England, other parties were marginalised as two-party politics was apparently firmly re-asserted.

But even in Wales, recent changes in the electoral fortunes of parties have been so dramatic as to apparently put almost everything and everywhere up for grabs. The most recent Welsh Political Barometer poll, published last month, was quite literally the worst opinion poll for Labour in Wales ever recorded. As recently as December, Labour in Wales on 45 percent support for a general election – just four points down from their extremely strong 2017 result. July’s Barometer poll had Labour on just 22 percent. Meanwhile, five parties had their general election support recorded within a nine-point margin, from 24% to 15%.

In such a context – of large changes in party support, and close contestation between multiple parties – it becomes even more than normally hazardous to suggest how current opinion poll ratings might translate into a general election outcome. We have seen, in both 2017 and in this year’s European election, large changes in public support during the campaign period. But even if the polls were to stay roughly where they are, how national poll results would project into seat totals can be very difficult to call. When parties are seeing their support halving, or doubling, we can witness extraordinary results: anyone who doubts this should just re-examine the 2015 general election results from Scotland.

All methods that use national polling results to project seat outcomes under First Past The Post have problems in accounting for local factors: strong or weak campaigns by particular parties, strong or weak candidates, particular local issue etc. Nonetheless, in ‘normal’ political times a method like uniform national swing can generally get close to the final result; so also can the alternative method of ratio swing. And the two methods do not normally project very different results: see for instance, the discussion here.

But these are not normal times. One illustration of that is simply to show the uniform swing and ratio swing seat projections of the July Barometer poll Westminster vote intention figures. This is what I do below: the table shows the party projected to win the seat in a general election under the two methods. (A note about the table: have assumed Brexit Party is ‘UKIP Mark II’, and so used the swing from the UKIP 2017 vote as the basis for the calculations here).

Seat Uniform Swing Ratio Swing
Aberavon Labour Brexit Party
Aberconwy Conservatives Conservatives
Alyn and Deeside Conservatives Conservatives
Arfon Plaid Cymru Plaid Cymru
Blaenau Gwent Labour Plaid Cymru
Brecon and Radnor Liberal Democrats Liberal Democrats
Bridgend Conservatives Conservatives
Caerphilly Labour Brexit Party
Cardiff Central Labour Liberal Democrats
Cardiff North Conservatives Conservatives
Cardiff South and Penarth Labour Labour
Cardiff West Labour Labour
Carmarthen East and Dinefwr Plaid Cymru Plaid Cymru
Carmarthen West and S.Pembs Conservatives Conservatives
Ceredigion Liberal Democrats Liberal Democrats
Clwyd South Conservatives Conservatives
Clwyd West Conservatives Conservatives
Cynon Valley Labour Brexit Party
Delyn Conservatives Conservatives
Dwyfor Meirionydd Plaid Cymru Plaid Cymru
Gower Conservatives Conservatives
Islwyn Labour Brexit Party
Llanelli Labour Brexit Party
Merthyr Tydfil and Rhymney Labour Brexit Party
Monmouth Conservatives Conservatives
Montgomery Conservatives Liberal Democrats
Neath Labour Brexit Party
Newport East Labour Brexit Party
Newport West Conservatives Conservatives
Ogmore Labour Brexit Party
Pontypridd Labour Labour
Preseli Pembrokeshire Conservatives Conservatives
Rhondda Labour Plaid Cymru
Swansea East Labour Labour
Swansea West Labour Labour
Torfaen Labour Brexit Party
Vale of Clwyd Conservatives Conservatives
Vale of Glamorgan Conservatives Conservatives
Wrexham Conservatives Conservatives
Ynys Môn Plaid Cymru Plaid Cymru

As mentioned when the latest Barometer poll was published, a uniform swing projection thus generates the following results:

Labour: 18 seats

Conservatives: 16 seats

Plaid Cymru: 4 seats

Liberal Democrats: 2 seats

Ratio swing, by contrast, produces the following projection:

Conservatives: 15 seats

Brexit Party: 10 seats

Plaid Cymru: 6 seats

Labour: 5 seats

Liberal Democrats: 4 seats

If anyone does think they know for sure what will happen if a general election comes – well, good luck to them.


  • Kenneth Vivian

    Seeing the same number of sheep on the hillsides I am put to wonder where all the Labour voters gone. It is indeed a puzzlement.

  • jack russell

    The difference between the 2 methods is amazing.

    Crudely put the Uniform method makes Wales representation 60 / 40 left and right, but the ratio method makes it 37.5 / 62.5 left and right.

    All in the melting pot, but I suspect the actual outcome will be highly dependent on whether the next GE is pre or post 31/10.

    if pre people will have to trust BJ to stick with his promise of no WA or backstop and NO extension and in order to get brexit Tory & BXP MUST come to an agreement of some sort,

    If post 31/10 will depend what happened on 31/10 if we are out he is home and dry, if we are not out BXP will sweep the board.

  • Christian Schmidt

    I think another key issue is who will actually stand where. To give a UK example, recent opinion polls shows the Greens at around 7%. The average Green party candidates’ share of the vote in 2015 was 2.3%, an increase of 4.7 percentage points or 3 times as much. Assuming that many of these voters come from Labour, this could be very bad for Labour in many Labour-Tory marginals.

    However I would expect that the Green Party will only field about 450 candidates (as in 2017) and in particular sit out most Labour-Tory marginals (and LD-Tory-marginals too). If the polls are correct this would still see their share of the overall vote increase substantially, from 1.6% in 2017 to 5%. But for the purpose of predicting constituency winners, the effect of the increase in the Greens’ vote is surely pretty much zero…

    And I think a Plaid-LD pact in only a handful of constituencies would similarly have only a negligible impact on overall vote shares, but could be quite a big one in seat numbers?

Leave a Reply

Your email address will not be published. Required fields are marked *