Dr Rob Sutton is an incoming junior doctor in Wales and a former Parliamentary staffer.

The 2019 general election saw a major shift in voter behaviour. The Red Wall of traditionally Labour constituencies was broken, and several seats elected their first-ever Conservative MP. Working class voters switched to the Tories while Labour seemed to increasingly be the preserve of the middle classes.

The elections in May this year offer the opportunity to test whether these shifts mark a long-term transition in the UK’s political alignments. In Wales, voters will head to the polls on May 6 to elect their representative in the Senedd. Many Welsh seats, particularly in the South, have demographics and voting records which are notably similar to those of the formerly-Red Wall.

The Senedd election provides a predictive challenge for armchair psephologists. National polling is a highly imperfect barometer for estimating election outcomes. The twin problems of missing information (we only have national or regional, rather than constituency-level, data) and systematic bias (the phenomenon of the “shy Tory,” for instance) make predicting outcomes difficult.

Its mixed electoral system compounds the challenge. Of the 60 Senedd seats, 40 correspond to the Westminster constituencies and elect members through a first-past-the-post vote. The remaining 20 seats are elected through a form of proportional representation, with each of the five regions electing four members by the D’Hondt method, with voters casting separate votes for the constituency and regional candidate lists.

How can we try to predict the 2021 Senedd election without constituency level polling? YouGov provides a wealth of data through its Welsh Political Barometer. This data is disaggregated by various demographic, political and geographic features (age, gender, electoral region, social grade, vote in the 2016 EU referendum), which have the potential to be extrapolated to estimate swings in individual constituencies. Each survey contains data for over 1000 citizens, and 29 such surveys have been carried out since 2016.

We can apply these disaggregated changes to the known demographic and political characteristics of each individual constituency. To give an example: if we know how voting intention among leave voters in the 2016 EU referendum has changed, and we know approximately what percentage voted leave in each constituency, we can estimate the relative impact of that change in each constituency. In a seat where 60 per cent voted leave compared to one in which 30 per cent voted leave, we would expect the effect of the polling shift to be approximately double.

We can apply this principle to different disaggregated categories of polling data from YouGov’s Welsh Political Barometer. By collecting constituency data on age, EU referendum vote and social grade from the House of Commons Library, Dr Chris Hanretty, and Electoral Calculus respectively, we can estimate the impact of changes in support at the constituency level. The equation looks something like:

Overall change in voting intention = age-related change + EU vote-related change + social grade-related change

To give a worked example, age groups are disaggregated by YouGov into four categories. By best-fitting YouGov data for the 29 individual surveys, we find the following change in Conservative support among each age group:

  • 16-24:  -1.7 per cent
  • 25-49:  +5.8 per cent
  • 50-64:  +9.0 per cent
  • 65+:     +8.1 per cent

While Conservative support has fallen slightly among the youngest category, among the three older groups it has strengthened, driven in party by the decline of UKIP and the Brexit Party. Applying the same logic to data for the 2016 EU referendum vote and social grade data, we can build profiles of individual constituencies to estimate them, weighting each of the three components by one-third to avoid overcounting. Regional swings are calculated using trendlines for regional voter intention.

Amongst Remainers, Conservative support has fallen by 5.6 per cent, offset by support among Leavers growing by 15.5 per cent. Overall support across social grades for the Conservatives has grown, but has been stronger amongst C2DE voters, increasing by 8.2 per cent compared to 2.6 per cent amongst ABC1 voters. Regional support has increased across all regions, ranging from a low of +0.4 per cent in North Wales to a high of +9.4 per cent in South Wales West. These trends are very positive for Conservative prospects.

How does this translate into electoral outcomes in our model? In the 2016 Senedd elections, Labour won 29 seats, Plaid Cymru 12, the Conservatives 11, UKIP seven and the Lib Dems one. This model suggests a significant erosion of the size of the Labour party in the Senedd down to just 21 seats, with a corresponding growth of the Conservatives to 16 seats (+5) and Plaid Cymru to 18 seats (+6). UKIP will lose all 7 seats and both Reform and Abolish the Welsh Assembly are expected to take 2 seats. The Greens are not expected to gain any. There are a further four seats in which the Conservatives are expected to come within five per cent of the winning vote.

The model used here has several limitations. Most notably, it does not contain any constituency level data: instead, it relies on nationwide demographic and political trends to be extrapolated to a local level by using local demographic and political data. The model also assumes the effect from age, EU vote and social grade contributes equally to party polling swings at a constituency level, which is an oversimplification. There is also a chance that Labour will enjoy a significant boost late in the campaign if their post-Covid reopening goes ahead smoothly.

Yet taken together, the findings presented here suggest a very positive result for the Welsh Conservatives. There are also significant implications for what the Welsh Government could look like after May 6, and the Welsh Conservative party leadership should be preparing for these possibilities. Labour would be a long way from a majority on just 21 seats and would naturally seek to build a coalition to stay in power.

Numerically, a majority could be held by either a Labour-Plaid or Conservative-Plaid coalition, but both the Conservatives and Plaid have previously stated they would never enter into a coalition with each other, secessionism being fundamentally incompatible with the Conservative and Unionist Party’s values. There is also precedent for a Labour-Plaid coalition, so if this model proves accurate, that will be the most likely outcome.

If Labour and Plaid are unable or unwilling to form a coalition, then the resulting hung Welsh parliament may necessitate a second Senedd election soon afterwards. The Welsh Conservatives’ leadership should be quietly optimistic for the possibility of significant electoral gains, while quietly wargaming the possibility that we have a hung Welsh parliament after the May election.