Damned facts and local election statistics

After the local elections earlier this month front-bench Labour MP Barry Gardiner tweeted, “BBC election headline says parties are “neck & neck”. Pity they then had to provide the figures… Labour 2,350 seats – up 77. Tory 1,332 seats – down 33. Damn facts! Getting in the way of a good story again.”

The parties were indeed neck and neck, on 35% in the BBC Projected National Share (PNS) of the vote (which I helped to calculate). In a similar exercise published in the Sunday Times, Colin Rallings and Michael Thrasher’s National Equivalent Vote (NEV) share put the Conservatives only narrowly ahead of Labour: 37% to 36%.

While not exactly a puzzle, there is clearly something that needs explaining here. So how could the parties be neck & neck, or the Tories ahead, in vote share when Labour won many more seats than the Conservatives and also made greater gains?

One part of the puzzle is easily explained. Since this month’s local elections were in mainly urban England and over 40% of the seats were in London, Labour went into these elections with 2271 seats to the Conservatives’ 1367. It was hardly surprising that Labour should have ended up with a larger seat tally. However, across Britain, Labour now have 6485 councillors to the Conservatives’ 9102. On that basis the Conservatives are way ahead. The trouble with that basis is that the number of councillors per person varies between different parts of Britain. For instance, people in (mainly Tory) shire England typically elect both district and county councillors, whilst those in (mainly Labour) metropolitan areas usually elect councillors to a single borough council. Neither the full GB councillor numbers, nor those regarding seats up for election in a particular year are a neutral guide to the parties’ relative performance.

If seat tallies are not very helpful indicators of success, what about headline figures on net gains and net losses? Surely they tell us who is on the up and who is on the way down?

Often but not always, yes, relative to when the seats were last fought. Maybe then the parties ended up neck & neck because Labour started from a lower point and made gains while the Tories started from a higher point and declined and the two met somewhere in between their positions from 2014?

No, actually the opposite. Labour were ahead 31% to 29% in the PNS in 2014, and by 31% to 30% in the NEV. There was a one-point swing to the Conservatives, from Labour, between 2014 and 2018 in both the PNS and the NEV. The real puzzle therefore is why Labour made net gains and the Tories lost seats at the same time as the Conservatives were gaining votes at a faster rate than Labour.

That both main parties were both winning more votes was overwhelmingly due to the collapse of UKIP. The 2014 local elections were held on the same day as the European Parliament elections, which UKIP won with 28% of the GB vote. Their 17% in the PNS, and 18% in the NEV were not quite so high, but still comfortably enough to put them third in the local vote. UKIP went into the elections this year notionally defending 126 seats. They won only 3. Perhaps then the solution to the puzzle is that Labour gained more seats despite the swing to the Tories because they disproportionately benefitted from the collapse of UKIP?

Again no, analysis of detailed results collected by the BBC suggests that both Labour and the Conservatives benefited from the decline of the UKIP vote, but, as the graph below illustrates, the Conservative vote went up by much more than did Labour’s where the UKIP vote fell most.

Also, the more seats UKIP lost in a council the more both Labour, but particularly the Conservatives, gained. The graph below shows that the Conservatives on average made only marginally greater gains than did Labour where UKIP lost seats. For instance, consider the two councils where UKIP lost the most seats. In both, the party lost all ten they were defending. In Great Yarmouth the former UKIP seats went 6 to the Tories and 4 to Labour. In Basildon 5 went Conservative, 3 Labour and 2 Independent.

This makes it still more puzzling that the Tories should have suffered net losses. Not only was there a national swing in the vote towards the Conservatives, but they also disproportionately benefitted from the UKIP collapse.

To unravel this puzzle it helps to look at the kinds of places where the Conservatives were losing seats. John Curtice has presented analysis of changes in vote share showing swings to Labour where Remain did best in the 2016 referendum, and swings to the Conservatives were Leave won comfortably. There is a similar pattern with respect to seats as illustrated in the graph below. The Conservative seat losses were concentrated in Remain voting areas while Tory net gains were largely in Leave areas.

Notice that there is much more variability on the Remain side than on the Leave side. Where Leave won more than 60%, the Conservative gains and losses are all between -10 and +10. Meanwhile the party made net gains in some strongly Remain voting councils but big losses in others, including Richmond (-28), Kingston (-19) and South Cambridgeshire (-16). This greater variation on the Remain side is largely due to the fact that there were more seats up for election in London and some other Remain voting areas because they had all-out elections. Meanwhile, places which voted Leave were more likely to have been District councils with just one-third of their councillors up for election.

The graph below is similar to the one above, but instead of plotting absolute numbers of gains and losses, the figures are percentages of the number of seats that were up for election this month. Looked at this way there is still a general trend for the Conservatives to do better in more Leave voting areas, but there is now much more similar levels of variation in Leave areas as there are in Remain areas. Some of the gains in Leave areas now look just as spectacular as the big losses in Remain areas. For instance, the 9 seats the Conservatives gained in Nuneaton and Bedworth represents 53% of the 17 seats that were up for election. This is similar to scale of losses in Richmond (52% of the seats up).

If all 150 of the councils that held local elections earlier this month had had the same number of seats up for election per elector, and the parties’ net gains and losses as a percentage of the total seats up in a council remained unchanged, then the Conservatives, instead of making losses, would have made a net gain of 58 seats, only 9 short of the +67 that Labour would have achieved under these assumptions.

So if we adjust for differences between councils in the number of seats up for election then the Conservative and Labour performances looks pretty similar. That resolves a large part of the part of the puzzle. But since the parties being neck and neck in the PNS implies a swing to the Conservatives, we still have to explain why the Conservatives did not secure bigger net gains than Labour.

For that we need to consider the imbalance between the places that did and did not have local elections this year. On average, in the places with local elections the Remain vote at the 2016 referendum was, at 49%, just a little above the level of the pro-EU vote in England as a whole (47%). That small difference is enough to say that, if there had been local elections across the whole of England, and if the number of councillors up for election had been the same per elector everywhere, then, based on the pattern of results, the Conservatives would have made a larger net gain of seats than Labour. Puzzle solved!

It was anticipated before the local elections that the expected headline gains for Labour would exaggerate their performance and the results for the Conservatives would correspondingly be unflattering. But it is, nonetheless, unusual for there to be a swing in votes one direction, but a swing in seats in the other. The facts do get in the way of a good story, but not in the manner that Barry Gardiner apparently supposed. The complications of local elections are such that headline statistics on seats can be misleading indicators of party performance.

Thanks to the BBC, John Curtice, Patrick English and Robert Ford for help with the data.