Bad Spreadsheets Part II: Horse and Hound edition

So: despite the extraordinary Sturm and Drang over the Tory leadership, the Boris-ing and the coming Brexit rebellions, it seems that the return of two-party politics in the UK has got us stuck in a kind of cold war. Both parties appear to be electorally maxed out at over 40% (let's not forget that on 8 June although Labour got its best vote share since 2001, the Tories got their best vote share since 1992). David Runciman, amongst others, thinks there's "no reason to suppose either party is capable of making big advances from its current position: both may be close to the limits of their appeal." Despite public Tory meltdown, neither side is moving very much. And Labour are still 60-odd seats away from a majority.

If, then, I said I thought we were a whisker away from a progressive government (assuming we can get an election) you'd be forgiven for scepticism. Needless to say, I -- like everyone else -- was wrong about the last election. So why believe me, when six months ago, right before Seven Nation Army and a 10% swing, I was arguing obstreperously that the young simply aren’t numerous enough to propel Labour anywhere close to winning?

But in fact, as the post-election surveying has shown, most pundits' first diagnosis of the Corbynite 'youthquake' was also wrong. Although turnout was up massively amongst under-25s – from around 43% in 2015 to at least 54% in 2017, and perhaps as much as 58% – it seems clear that they weren’t numerous enough to deliver Labour's big gains. Very substantial boosts in Labour support amongst 30-44 year olds were equally if not more important. My piss-poor spreadsheet was wrong not because I overestimated the electoral stranglehold of older voters, but because I hadn’t imagined there could be anywhere near as big a shift in the voting intentions of middle-aged cohorts as there was between April and June this year. The underlying demography, however, is still there. Indeed, it’s extraordinary testimony to the demographic dominance of the elderly electorate that Labour can have a majority in every age cohort under 50, and still be 56 seats down. 

Paradoxially, though, Labour’s reliance on 8 June on a middle-aged swing, rather than massive youth turnout, is good news for the potential power of the youth vote. Because it suggests that, whatever Runciman et al may say, Labour still has some powder dry. Nearly half of Labour's most supportive demographic constituencies still aren't turning out to vote. Meanwhile Tory-dominated age cohorts are already giving it everything they’ve got. Turnout amongst over-55s was pushing 70%. Amongst the Conservatives’ shock troops, the over-70s, it was around 84%, and surely can’t go much higher. 

The middle-aged swing, in other words, gives the under-30s the electoral space to be truly influential. Which in turn suggests a strategy, says my friend Mika Mino

Let's assume that age-specific voting intentions stay roughly the same (a heroic assumption, of course, and precisely what everyone was wrong about before; but I find it very hard to believe that under-25s will swing significantly towards the Tories any time soon). In other words, assume that around 62% of under-30s are going to vote Labour again, about 22% will vote Tory again, and that those who didn’t vote this time round have roughly the same voting intentions as those who did (another heroic assumption). 

In that case, maybe just going down to knife-edge marginals and helping register new under-30 voters (and who, left-wing or right-wing, could argue with youth voter registration as a democratic pursuit) would generate enough new Labour votes to tip us over the line next time round? 

Mika has been having a look at how the maths stacks up. At the risk of well-deserved ridicule, I’ve extended the thought by tabulating the numbers for Tory marginals with the slimmest majorities (many thanks to BritainElects.com for spreadsheeting the GE2017 results).

Assume that a working majority, taking into account Sinn Féin empty seats, is still 322, and that we’re after a Lab-SNP-LD-Green pact – currently on 310 seats. Labour therefore needs to win at least another 15-odd seats. And the numbers in the top 15 Tory super-marginals are pretty striking.

The six most-marginal Tory seats each need less than 1000 new under-30 voters, in theory, to fall to Labour. These are not all new Labour voters, remember, just new voters for all parties in that age-range. Across all 15 seats, we need just 21,200 new under-30 voters to come out and vote. That’s about 0.2% of the 18-to-30 population. More people bought Horse and Hound magazine last month. Seriously - we're talking about a number of people smaller than that proportion of the population for whom "Can You Prove You Own Your Horse?" is clickbait.

This is, of course, a reductio ad absurdum. It doesn't take into account age-specific variations in voting intention between constituencies (because we don't have the data, for the most part). And it's built on huge assumptions, at a time when the only rule in British politics is, of course, that all assumptions are wrong. This isn't a static fight. The Tories will seriously up their offer in the next election. There will be many, many other things to be done. Nonetheless -- to completely reverse my previous argument -- it does look like a targeted youth registration drive might be a significant component of bringing a progressive government to power.

If only we could have another election sometime soon.

Which is the thing the Fixed-term Parliaments Act seems to make impossible. That's a subject for my next post. But for now, have a look at these lovely numbers...

Constituency201520172017 ResultCon majority 2017Number of new under-30 voters needed for Lab majority
Middlesbrough South & East ClevelandLabConGain1,0202,550
Bolton WestConConHold9362,340
Northampton NorthConConHold8072,018
Stoke-on-Trent SouthLabConGain6631,658
Calder ValleyConConHold6091,523
Norwich NorthConConHold5071,268
Chipping BarnetConConHold353883
Hastings & RyeConConHold346865
Preseli PembrokeshireConConHold314785
Southampton, ItchenConConHold3178


Bad spreadsheets (election edition)

So: YouGov tells us that if only under-40s voted in next month's general election, "Jeremy Corbyn would be heading to Downing Street". In the words of YouGov pollster Chris Curtis, "age is the new dividing line in British politics", replacing social class as the most significant correlate of voter preference. But young people, famously, don't vote very much. And so, Polly Toynbee says, "To save the young from themselves, it’s time to march them down to the polling station, making voting compulsory, at least the first time."

Theresa May, 1971

Meanwhile other professional opiners as disparate as Liam Byrne and Aaron Bastani tell us that Labour has to appeal to older, largely more conservative voters -- as Bastani insists, it can't win with the millenial vote because it's concentrated in urban areas, whereas its the shires that swing it.

So: is a centrist Labour Party appealing to rightward older voters a demographic necessity if it wants to take power? Or is there demographic wiggle-room for a leftist agenda backed by a young vote, if the under-40s turn out to vote in sufficient numbers?

This seems like a central question in British electoral politics -- and perhaps the central question for the parliamentary left. I was quite surprised that I couldn't find any recent modelling on the impact of youth turnout on 2017 election predictions ( there surely has to be some, and it's likely I'm not looking hard enough).

So instead, being a fan of half-arsery to inform my political prejudices, I spent ten minutes with an Excel spreadsheet.

It was a depressing ten minutes. If you take the size of the UK voting-age population within different age brackets, and plug them into both (1) age-specific polling on 2017 voting intentions, and (2) age-specific turnout figures from the 2015 general election, then you can roughly estimate the number of votes that different parties will receive in 2017, if turnout amongst different age groups is similar to that in 2015:

Votes, GE 2017, if turnout at 2015 levels

Age group 18-19 20-24 25-29 30-39 40-49 50-59 60-69 70+ TOTAL %
Con 149826 443211 695497 1675171 2368265 2960161 3108403 3714204 15114739 44%
Lab 292843 775620 1007272 1624409 1578843 1547357 888115 754448 8468906 25%
Lib Dem 122585 295474 287792 761441 789422 672764 555072 464275 3948826 12%
UKIP 27241 92336 119913 355339 667972 807317 499565 464275 3033959 9%
Other 88534 240073 311775 659916 728697 672764 444058 348207 3494022 10%

Then you can see what happens if, say, there's a 75% turnout in all age brackets (in 2015, over-50s turned out at around 78-79%, under-25s at around 43%):

Votes, GE 2017, if turnout at 75% in all age brackets

Age group18-1920-2425-2930-3940-4950-5960-6970+TOTAL%
Lib Dem213812515362399711951802870686638700533723446419457021412%

The Tories lose just 2 percent of the vote share to Labour. Still a recipe for a landslide.

Indeed, to demonstrate the demographic impossibility of it all: let's imagine that turnout was 100% in all age brackets -- compulsory voting ,say -- and everyone aged between 18 and 30 voted Labour, with everyone else voting as they intend to do this time round:

Votes, GE 2017, if turnout at 100% in all age brackets and all 18-30 vote Labour

Age group18-1920-2425-2930-3940-4950-5960-6970+TOTAL%
Lib Dem0001269069116091485160071163159522545884399%

It would take a youth result of these Jong-Il-ian dimensions, in other words, for Labour to get a lead of just 1% in the vote share.

There's everything wrong with this simplistic model, of course: it assumes that non-voters have the same split of voting preferences as voters; it assumes that population translates straightforwardly into electorate; it ignores regional differences; it completely ignores the fact that FPTP doesn't translate vote share straight-forwardly into seats (though in only one general election, in February 1974, did one of the two main parties actually win more seats than the other with a (fractionally) lower share of the vote); and so on and so on.

So I'd be very keen to see some better modelling, and at constituency level. There must surely be some places where getting the youth vote out could swing seats.

But overall, even on this crudest of models, it looks like voter registration and voter turnout won't -- can't -- even seriously dent the Tories, let alone stop them. And that the myth that sofa-bound young people can rise up and rock the vote is just that: a patronising myth to hide the basic fact of British electoral physiology that you can't turn a pear into an apple. I suspect the same is true of BME communities, though I haven't looked into the figures.

Young people are a minority in post-demographic-transition societies: a relatively powerless, relatively property-less, minority. They differ from other minorities, of course, because they are destined over time to become the majority. But then they won't be young, and the policies they make or ratify likewise won't have the same impact on them as on others younger than them.

Perhaps, therefore, instead of either haranguing or misleading young people about their electoral potential in the face of their (rational) electoral apathy, we need to start doing what we do for other powerless minorities in majoritarian democracies: require legal safeguards to ensure that their interests are not unjustly ignored or damaged in the political process, precisely because the electoral process can't provide such safeguards. What if there was a statutory requirement that all legislation had to be tested not only against human rights and equalities law, but against its differential impact on those with longer futures? Wouldn't that automatically lead to better, long-term thinking and decision-making in politics? What would a Finance Bill look like against such a legislative test? What would such a legislative test itself look like? Could we even argue -- perhaps in court -- that the Equalities Act 2010 already provides for such a test, by including "age" amongst "protected characteristics"?

Meanwhile, suggestions greatly appreciated for more (or less oafishly used) data...