The Audit: 2024

Every year, in my last piece before Christmas, I do an audit of my posts to see what I got right and wrong, with some lessons to take into next year’s writing (here are the ones for 2022 and 2023).

I see it as having two main benefits. First, it keeps me honest during the year as I know that if I make any “bold” predictions that aren’t justified I’ll have to own up to looking daft. Secondly, it forces me to confront mistakes and try and learn something from them. If you think I’ve been too kind to myself or have any other feedback please do stick it in the comments.

I’m going to start this audit with a look back at the UK local and general elections, as that’s where I made the bulk of my quantitative predictions, before a diversion to the US elections, and then back to the UK to look at my assumptions about Labour’s first few months.

UK elections

Probably the thing I’m most pleased about since starting the substack is correctly identifying how the general election would play out from a long way back. One of my first posts in early 2022 explained why I thought a Labour majority was undervalued in the betting markets. And in summer 2023 I wrote about why I thought a Tory “wipeout”, which I defined as below 100 seats, was on the cards.

While this didn’t quite happen (a couple of points more for Labour and it would have) the factors I set out did all materialise, lead to the lowest ever Tory seat total. These were: Sunak’s weaknesses as a leader; heavy anti-Tory tactical voting; a highly efficient Labour vote, winning exactly where they needed to to maximise seat numbers; and the return of Nigel Farage.

The key point, which even the BBC was still getting wrong as the year started, was that the margin Labour needed for a big majority was going to be much lower than uniform swing calculations would imply.

The local elections in May strongly supported this view, with Labour winning most of the councils they were targeting despite a low overall vote share. My predictions for those elections were my most accurate yet. I said the Tories would lost 450-500 seats and they lost 473. As always I made council level predictions and thought the result would be: Labour 52, Tory 3, LD 13, Others/NOC 39. It was Lab 51, Tory 6, LD 12, Others/NOC 38. I got all the mayor results right as well, except I chickened out of calling the West Midlands, saying it was too close to call. Labour’s Richard Parker won by 1,500 votes.

This gave me a lot of confidence that my broad assumptions about the general election were right. In retrospect, my slight miss on vote share (I thought Labour would get 36% and they got 34%) should have been a clue that they were underperforming their polling a little.

For the general election campaign we moved to daily posts – with Dylan Difford and Josh (@beyondthetopline) helping me out with analysis. I decided, without quite thinking through what it would mean, to do seat previews for all 650 seats in the UK. Which meant I ended up, in effect, writing a 80k word book in four weeks. It was a very valuable exercise in understanding electoral geography and the composition of the new parliament, but did almost kill me. By the end I was writing for 14 hours a day, so one big lesson is to give myself more time in future.

Overall I got 89% of seat predictions right (578 out of 650) which was better than any of the models, the exit poll and all but one of the MRPs (YouGov got 580 out of 631 - they didn’t do Northern Ireland). I feel I could have done better though. I made a couple of calls that were wish fulfilment rather than likely – Gavin Williamson was never going to lose.

But my main error was to believe the polls and assume Labour’s vote share would be more like 38-39% rather than 34%, which meant they won around 30 fewer Tory seats than I expected.

Looking back through our polling analysis I think we were too slow to see the drop in Labour’s share over the final few weeks as significant, explaining it away as being mainly about tactical voting. We were also too dismissive of the idea Reform were taking Labour votes. It was true that they were mostly taking from the Tories, but the recent analysis we published from John Curtice and Lovisa Moller Vallgarda shows that there was a Labour to Reform shift in the “left behind patriots” group in the run up to the election. YouGov have also identified a genuine late Labour to Tory swing.

The only other substantive mistake was to miss how successful independent candidates would be in areas with high Muslim populations. I did highlight the issue in nearly all the seats independents won, as it was apparent from the locals that “Gaza independents” would do well, but I still thought Labour would hold on to them all when it came to a general election. And a few, like Jonathan Ashworth’s seat in Leicester were genuine surprises.

We did feature several posts during the campaign on the broader issue of vote fragmentation, and this is something I’ll be returning to a lot in 2025. It’s going to drive much of the political debate, and party behaviour, over the next four years. It’s also going to make future elections harder to call because there will be so many more variables, and polling will get even trickier.

I think my approach of going council by council, or seat by seat, is going to be increasingly the only way of making accurate assessments due to big regional, and even hyperlocal, differences in party composition. Which means a big takeaway from this year is thinking about how to do this is a somewhat more structured way, giving myself more time and bringing in more local expertise. Next year locals will be the first test with a resurgent Reform party complicating factors further.

US election

I didn’t write as much about the US election as I initially intended largely because, once Harris had replaced Biden, the nature of the race barely changed. So I only did one proper post in early September. There were lots of narrative shifts, back and forth, after that, largely based on vibes, but the data didn’t really change. In that September post I noted that the risk of systematic bias in the polls against Trump remained despite attempts by some companies to fix the issue:

“In both 2016 and 2020 there was systematic error that meant Trump did better on the day than the polls predicted. Attempts to explain this are confounded by the fact that polls were largely accurate for the 2018 and 2022 midterm elections. Some analysts think there is a Trump specific issue here. There are two theories. The first is that there are “shy” Trump voters who don’t want to admit it. But this has been thoroughly debunked by various reviews. (In any case it’s not very intuitive: shyness is not a personality trait one naturally associates with Trump supporters).

There is, though, evidence for the second theory that some types of Republican voters, particularly those who might come out for Trump but not for midterms, are less likely to respond to polling companies. Some firms are trying to adapt to this problem either by adding more variables to their weightings, like education levels and media consumption, or by shifting towards probability sampling for online polling (or both).”

But I said there was no way of knowing if these fixes would solve the problem or even create an anti-Harris bias. It turned out there was still a bias against him, and that was enough to secure a narrow win. I do wonder if I should have been bolder and said this was likely to happen again, leading to a Trump win, rather than just a possibility. Did I just not want to predict an outcome I didn’t want? On balance I don’t think there was enough data to justify making that call.

I correctly said the Republicans were likely to win the Senate, though that was not a difficult prediction, and that the House was a toss-up, which it was, as the Republicans held it by just five seats over the Democrats.

I’m planning on writing more about the US next year because Trump’s choices over personnel and policy will have such a huge impact on everything else that happens round the world. But I’m always conscious how much harder it is to write about a country you’re not living in and the risk that creates of just sticking with the conventional wisdom.

I also haven’t written as much about Europe as I planned to this year and will take the opportunity of German elections and the current challenges in France to look more at some of the broader trends affecting continental democracies.

Labour’s first six months

I don’t think anyone could accuse me of having had unduly high hopes for Labour. Back in October last year after attending their conference I wrote about what I saw as the big problem with Starmer’s proposals for governing:

“[The problem isn’t] that ‘they don’t have any policies’. They have far too many fiddly policies. No one who doesn’t spend their time reading policy documents and going to conferences will have heard of them….No, the problem is very specific. They have no substantive policies that would involve having to spend any taxpayer money. Starmer’s holding position that he wishes to run “a reforming state, not a cheque-book state” is transparent nonsense….

For a start when it comes to health, welfare and education – the three departments that account for 60% of all spending – they have no proposals for significant reform. I watched a succession of junior shadow ministers struggle through fringe events on each of these issues, acknowledging the terrible state of things right now, promising a decade of transformation, but unable to fill in any of the steps that might get us there.”

Six months into government this remains the problem. Ministers have set out substantive proposals on planning reform, local government and energy policy but they still do not have a plan for the public sector. This is exacerbated by Starmer’s discomfort with political strategy which makes it very difficult for advisers and ministers to anticipate how we wants them to deal with tricky questions. As I wrote in the Autumn:

“The only way for prime ministers to cope in our system is for those around them to be so clear about their beliefs that they can act on their behalf knowing it will be in line with expectations.”

This isn’t happening.

Moreover, the fiscal constraints remain. Anyone paying the slightest attention knew that the Conservatives pre-election tax cuts were not affordable and would have to be undone by the new government. I wrote several posts pre-election explaining why and in my August post judged the upcoming spending review would need to find “well over £60 billion” in new spending (it turned out to be £70 billion).

Labour’s refusal to acknowledge this during the campaign was a major limiting factor in what they could say or plan. But even now those cuts have been reversed, in a somewhat roundabout way, it remains their biggest problem. Most of the new money is covering old commitments around pay or capital projects that were not properly funded by the Tories. Or is keeping various collapsing systems – like special needs or criminal justice – from falling over completely.

How the government navigate these fiscal constraints will be one of the main stories of next year, especially given rising bond yields mean they are at serious risk of breaching their fiscal rules at the first point of measurement in March. This and the spending review will be a major focus for me in 2025.

Lessons for 2025

This is a generally positive audit. Readers would not have been surprised by the election result here or in the US, and would have had a good sense of the challenges facing the new government. Looking back at last year’s audit and my areas of improvement I think I’ve done well at avoiding over-reacting to immediate news stories and focusing on the big picture, that’s something I want to carry on as a feature of this substack.

I’ve done less well at avoiding parochialism and writing more about broader global trends (obviously excluding wars and international relations, which is Dad’s turf!). This is a time issue, because it requires more research to write about areas you know less well, but it’s all a confidence issue. I just feel more comfortable writing about the UK, but I really do need to push myself in 2025, as well as commissioning more guest posts from outside the UK. Otherwise there’s a risk of seeing UK problems as unique and missing the wider factors.

I’ve also not done a great job of finding things to be positive about. I did have a post on how the UK is doing much better at integration than you’d think from the media coverage. And a relatively positive post on local government reform. But it’s slim pickings. It really does feel hard to find things to be optimistic about at the moment but that’s all the more reason to push myself harder in 2025.

So my lessons to take into next year:

• Keep the focus on the big picture, and stay away from day-to-day political commentary unless there’s a very good reason to write about it.

• For election coverage develop a more systematic approach to council/seat level analysis, that allows me to build up and record local knowledge over time.

• Write more about the US, Europe and broader global trends, accepting that I will make more mistakes but it will improve analysis of UK challenges.

• Keeping trying to find more positive trends to write about – challenge my instinct towards pessimism.

Let me know what you think I should be doing differently in the comments (I’m not sensitive) and I hope all our subscribers have a great Christmas and a Happy New Year.