The Emerging Instability in UK Elections

The 2024 general election generated the most disproportionate result of any election in UK history. It was also very close to the most disproportionate in the history of the western world. Labour gained a touch over one third of the votes but almost two-thirds of the seats. The Tories won just under a quarter of the votes but only 19% of the seats. Labour came out of the election with an almost unprecedented majority of 172 over all parties and 290 over the Tories. It achieved this despite winning only one in three of votes cast and the support of only one in five of all electors. This means that if you walked down the street and stopped adults at random only one in five would have voted Labour. In other words, four out of every five electors chose, either actively or passively not to support Labour, yet the Government we got was one of the strongest ever seen in Parliament. How did this happen and what does it mean for democracy?

The expected gap in votes between Labour and Conservatives based on pre-election polls was much greater than that which actually occurred in the general election and the over-optimistic predictions may have caused complacency among Labour voters.  Labour’s vote was overestimated by a whopping 7 percentage points and the Tory vote was underestimated by 4 percentage points. The final polls close to polling day predicted an 18-point lead for Labour over the Tories, but the actual gap was much less, at ten points. Pollster’s attempts to use more sophisticated methods than in previous elections failed to improve matters. The so-called multi-level regression and post-stratification (MRP) techniques used large samples in an attempt to allow for demographic and social differences across constituencies to predict the number of seats won. This did not work well and the average predicted number of Labour seats was over-estimated by 45.

If this had been a PR election in which the number of seats won reflected the number of votes gained, the outcome would have been hugely different. The Tories would have won 34 more seats but Labour would have won 191 fewer than they actually gained. Instead of a majority of 172 over all other parties they would have been over 100 seats short of a majority. A three-way coalition with the Greens and Liberal Democrats (whose seat share was close to their votes share of 12%) would have been needed to secure a majority. Alternatively, the 23 independents and mainly pro-Palestinian Moslem, candidates who would likely to have been otherwise pro-Labour, could have shored up a LibLab coalition with the support of Northern Ireland nationalist and Alliance MPs. Although the Independents would have been left-leaning their, support for Palestine would have exerted a difficult effect on post-election foreign policy. In summary, under PR the UK would have been left with a complex and probably unstable coalition government quite unlike what we now have.

The big losers out of the actual election were of course the Reform party. Under a PR election they would have ended up with 93 MPs instead of the five actually elected. This would not have been enough to form a winning coalition with the Tories but the 257 Tory, Reform and Ulster unionist MPs could have formed a strong oppositional block. The other big loser was the Green Party which gained 4 MPs instead of the 44 due under PR and could have been in a coalition government under PR. The SNP, Plaid Cymru and the Northern Ireland parties all gained MPs in numbers close to their proportional share of the vote because their voters were concentrated in individual regions rather than being scattered across the UK like Reform or the Greens.

First past the post elections

The geography of party support explains the pattern of proportionality and disproportionality. In a UK first-past-the-post election in which the two main parties dominate (as they did in 2019) the majority of constituency elections result in a relatively narrow win for one party over the other.  If we were to draw a bar chart for each party’s share of the vote, the chart would form a bell curve (ie. a normal distribution).  This has the unexpected consequence that the share of the seats won by each of the two main parties is close to the cube of their share of the votes. For instance, if Labour won twice as many votes as the Tories, the ratio of the vote shares would be two to one. If we cube this 2:1 ratio, we get an 8:1 ratio for the share of seats. In other world a party gaining two thirds of the votes would gain 8/9ths (or 89%) of the seats.

This bell-shaped distribution of party support depends on geography. Each constituency has close to 70,000 voters. If support for each party were randomly distributed across all the constituencies of the UK, even a slight national majority would result in the winning party gaining virtually all of the seats. A putative women’s party which was supported by virtually all females would for instance win pretty well all of the seats because females tend to slightly outnumber males in most areas.

In practice, voters are not completely randomly distributed. Instead, they are clustered. Labour voters tend to live in major cities, old industrial areas and on council estates. It is this clustering which produces the degree of disproportionality that we observe. The precise geography does not matter. Countries with FTP elections as diverse as the UK, US and New Zealand (when it had FTP elections) all tended to conform to the cube law. What matters is the size of average clusters of voters (e.g. the size of council estates or streets of affluent suburban voters) relative to the size of constituencies. If the clusters were much larger than they actually are, we would come closer to proportionality. For example, if all voters in large towns and cities voted Labour and all-rural dwellers voted Tory, most constituencies would be all-Labour or all-Tory. In this case, the voter-clusters would be larger than individual constituencies and the outcome of elections would tend to reflect national party support in votes.

These technicalities lead to predictable outcomes but are essentially arbitrary. Is this any way to run a democracy? Americans think not and use gerrymandering to rig electoral outcomes in congressional elections. US constituencies legally must contain almost precisely the same number of voters but constituency boundaries are determined by state legislatures and hence by partisan politicians. The result is that boundaries are drawn to create safe seats and to avoid marginal contests. The bell-shaped bar chart of party support across constituencies has a depression in the middle where the marginal constituencies should be. This does not greatly affect US Presidential elections where votes are generally counted across entire states, but does do so in two states where the votes are counted in congressional districts.

While the Tories were wildly unpopular in the 2024 UK election, losing two-thirds of their seats, the lacklustre Labour share of the votes, meant that the minor parties gained 43% of the votes in GB unlike 2019 when their share was 24%.  The impact of the success of the minor parties was to diminish Labour’s majority in seats.  If the 2024 election had been between only the two main parties, the cube law would have predicted a Labour majority of 318 instead of the actual majority of 172.

The latest polls

As we know, Labour have slumped in the polls in the short period since the general election. In the latest polls Labour and the Tories are neck and neck with a lowly 26% each. The main beneficiaries are Reform, up 6 points to 21% with the other minor party vote shares largely unchanged. This means that the two major parties have only 52% of the votes between them suggesting the possibility of a future realignment with one or both being replaced as frontrunner parties.

The next general election may not be held for another four years so the latest polls may not mean much but if something like these votes shares were to apply at the next election the results would be unstable. With three parties each with vote shares in the low 20s, the number of seats could vary a lot with even small changes in votes. If the current polls did apply at the next election, a uniform swing across the UK would generate the results for seats shown in the table above. Labour would be 192 seats short of a majority and even a coalition of Labour, Lib Dems, Greens and Moslem Independents would only scrape a narrow majority assuming the Sinn Fein continue to boycott Parliament. A coalition of Tories, Reform and Ulster Unionist might have 302 seats but still be 20 seats short of a majority.

Conclusion

The recent general election was perhaps a portent of a coming instability in UK politics.  The huge disproportionality between Labour’s share of seats and its share of votes temporarily disguised this instability, but a few month’s experience of Labour in power has shown the reality of tepid real support. The Tories have gained little from Labour’s falling support and the main result has been to increase Reform’s support to a level close to both Labour and the Tories. Three party elections are potentially very unstable in first past the post elections. The rise in Liberal and LibDem support in the 1970s and 1980s came close changing the face of UK politics but at actual elections these parties faded and the change was postponed. The reality at the next election may be different, with three equally matched parties all vying for power.

The question now is whether Reform continues its current rise to become the single most popular party. Labour has plenty of time to establish a record of competence, but as the current polls show they are not gaining the confidence of voters. Their plans for economic growth and public sector reform are almost all aspirational rather than detailed proposals. The fatuous aspiration for the UK to become the fastest growing economy in the G7 is emblematic of the yawning gap between aspiration and achievement.  Without any knowledge of what future G7 growth rates might prove to be, and little idea of how to accelerate UK growth rates, this aspiration reveals only the shallowness of Labour understanding of economics. Similarly, the aspiration to build 1.5 million new dwellings over the course the parliament has no chance of being realised even if the labour could be found to construct them. The private sector would need to build far more houses than it has ever achieved over the post WW2 period to meet Labour’s target and without drastic relaxation in mortgage lending it is difficult to envisage where the demand will come from.

A Labour government which falls far short of its oft-stated aspirations is unlikely to recover from its present low standing in the polls. If Reform can avoid the faction-fighting which tends to occur in emerging parties, it may achieve a sustainable parity with Labour and the Tories. However, with three evenly matched parties in a first-past-the-post election the results in future elections will be highly unpredictable.

Dr Graham Gudgin is author with Peter Taylor of Seats, Votes and the Spatial Organisation of Elections. Republished as an ECPR Classic by the European Consortium of Political Science.2012.