Sign in

Chris Prosser

@caprosser.com
1.9K followers 567 following 332 posts

Political scientist | Co-director British Election Study | Election number crunching for ITV | Trustee McDougall Trust caprosser.com

PostsRepliesMedia
Chris Prosser @caprosser.com · 29/09/2026
Critics of proportional representation: 'Yes, yes, I know that in reality politics under first-pass-the-post has been chaotic for the last decade, but in theory it provides stable two-party government, and that's what matters'.* * Not actually true in theory either.
Screenshot of newspaper article saying: 'Critics of proportional representation have said it would lead to a more chaotic form of government, with unstable coalitions and an inability to pass legislation.'
2275
Chris Prosser @caprosser.com · 17/09/2026
The second uses travel time data to classify places into an urban-rural typology, and using redistricted election and census data 1970-2024, argues that the rising urban-rural divide is driven by the activation of demographic differences between these types of places, not changing composition.
110
Chris Prosser @caprosser.com · 17/09/2026
My first paper argues that blocs have long structured voter perceptions of where the parties stand on different issues and how they feel about them (but this blociness has got stronger post-Brexit), and that changes in party viability activated the blocs in terms of voting behaviour in 2024.
100
Chris Prosser @caprosser.com · 04/09/2026
Maybe it's true of other organisations as well but I'm always amazed by how unwilling universities seem to be to do something about people who basically don't do their jobs.
1132
Chris Prosser @caprosser.com · 12/08/2026
140
Chris Prosser @caprosser.com · 10/08/2026
Working with some government data of locations *in the UK* that contains two points in the sea to the west of Denmark and three off the east of Somalia. Feeling good about the cleanliness of my own data 😂
040
Chris Prosser @caprosser.com · 05/08/2026
Obviously, 'on average' doesn't mean polls are always-and-everywhere biased to the left - there is a lot of heterogeneity between elections - but elections are much more likely to have a leftward polling bias than a rightward one.
121
Chris Prosser @caprosser.com · 05/08/2026
Breaking it down further by party family, we see a fairly consistent pattern across the left party families but more variation on the right, with polls tending to underestimate Agrarian and Conservative parties, but not Christian Democratic and Far Right parties.
120
Chris Prosser @caprosser.com · 05/08/2026
Using ~8.5k election-poll-party observations from 372 elections in 32 countries, we find a persistent partisan asymmetry: polls tend to overestimate the left relative to the right. On average across elections, polls the month before an election overstated the left by about 1.4 points
3128
Chris Prosser @caprosser.com · 29/07/2026
Just a few graphs... 😀
120
Chris Prosser @caprosser.com · 29/07/2026
It’s not out for another six months or so, but *Electoral Realignment: How Brexit Reshaped British Voting Behaviour*, my book with Ed Fieldhouse, @profjanegreen.bsky.social , Geoff Evans, @jack-bailey.co.uk and @jonmellon.bsky.social, is up on the OUP website! global.oup.com/academic/pro...
Cover of *Electoral Realignment: How Brexit Reshaped British Voting Behaviour*
By Edward Fieldhouse, Jane Green, Christopher Prosser, Geoffrey Evans, Jack Bailey, and Jonathan Mellon
13013
Chris Prosser @caprosser.com · 19/06/2026
Using a Bayesian changepoint model, we try and pin down when the class-vote association changed: the stabilisation of class voting in the 1920s; a brief drop for Labour after the Second World War; and then a more significant turn for Labour in the Blair era, and the Conservative around Brexit.
Meta-analytic changepoint models of the effect of class decile on vote share. In the top panel, points and errors bars are election-specific estimates of the effect of class decile on vote share. Lines and shaded areas are the posterior estimates of the piecewise linear estimate of the evolution of class voting since 1900. The lower panel shows the posterior distribution of the changepoints, which occur in the 1920s (Lab + Con), 1940s and 1990s (Lab), and 2010s (Con).
240
Chris Prosser @caprosser.com · 19/06/2026
Once established, the Lab-Con class relationship was remarkably stable from the 1920s into the 1990s, at which point the class-Lab association begins to decline, and then the class-Con association shifts more abruptly in 2017-19.
Marginal effect of class decile on support for Labour, the Conservatives, and Liberals at the constituency level, which shows a moderate working class-Liberal association and middle class-Conservative association between 1885-1900, and then the rapid emergence of a working class-Labour association (and declining Liberal association) from 1900-1924, and a strengthening of the middle class-Conservative association at the same time. From the 1920s-1990s this trend fluctuates but is broadly stable. Starting in the 1990s, the working class-Labour association begins to weaken, while the Conservative class association weakens quite abruptly in 2017.
141
Chris Prosser @caprosser.com · 14/05/2026
Miliband was neither especially popular nor unpopular - average feelings towards him during the 2017 election campaign were pretty similar to mid-2017-campaign Corbyn, 2024 Starmer, and 2019 Johnson.
Smoothed estimates of average feelings towards Conservative and Labour leaders during the 2015, 2017, 2019, and 2024 election campaigns.
76215
Chris Prosser @caprosser.com · 12/02/2026
Brushing up on some reading for a research ethics workshop I'm running in a couple of weeks and perhaps getting a little carried away...
160
Chris Prosser @caprosser.com · 09/02/2026
Sounds like excellent research and I will not be asking any probing questions about causal identification at this time.
A couple of teas or coffees a day could lower risk of dementia, scientists say
1101
Chris Prosser @caprosser.com · 09/02/2026
This is fun, even if the British Media, Politics, and Policy cluster is hanging out on the edge of the galaxy.
000
Chris Prosser @caprosser.com · 29/01/2026
An 'archaic' procedure such as one of the central planks of the British constitution, which establishes the supremacy of the Commons over the Lords...
Guardian headline:

Assisted dying backers could use archaic procedure to bypass ‘undemocratic’ block by peers
Exclusive: MPs backing bill to use ‘nuclear option’ of 1911 Parliament Act if it continues to be blocked by Lords
2124
Chris Prosser @caprosser.com · 14/01/2026
'Immediate reject' is a new peer review option to me... does it mean 'wow this is bad, don't even read my report just hit the reject button'?!?
392
Chris Prosser @caprosser.com · 08/01/2026
Procrastinating from writing something by analysing the number of peer reviews I do. Not sure if my new website made me more visible, or getting something into APSR made me look like a serious person, but 2025 was apparently the year journal editors remembered who I was again.
120
Chris Prosser @caprosser.com · 03/12/2025
The main thing I'm taking away from this is that more of you should listen to Olivia Chaney, so if you like indie folk, give her a listen!
130
Chris Prosser @caprosser.com · 06/11/2025
I mention the Hall, Ariss and Todorov study of basketball betting (more information -> worse performance) in the slides for one of my pol psych lectures, and now the powerpoint designer wants to illustrate the slide with some basketballs, which certainly creates an interesting juxtaposition... 😂
020
Chris Prosser @caprosser.com · 04/11/2025
I fell down a typographical rabbit hole last week and made my own R Markdown template. Quite pleased with how it looks, so take a look, and if you like it too, feel free to use it! Package here: github.com/drcaprosser/... Full explanation/example here: github.com/drcaprosser/...
4240
Chris Prosser @caprosser.com · 16/10/2025
Does it though...
030
Chris Prosser @caprosser.com · 07/10/2025
This is great - had always wondered if the telephone call thing was *actually* how they found out they'd won. In the unlikely event that I won a Nobel prize I'd probably see an unknown international number, assume it was scam, and block it.
Newspaper headline:
Nobel committee unable to reach prize winner who is ‘living his best life’ hiking off grid
Fred Ramsdell was among those honoured with a 2025 Nobel Prize in Medicine but might not know because he is somewhere in Idaho and uncontactable
1100
Chris Prosser @caprosser.com · 07/10/2025
A conclusion reached by a well justified causal research design no doubt, and not just seeing there's a correlation and asserting the causality runs in one direction... 🤔
Newspaper headline saying: 
Enjoying international cuisines makes people more tolerant, UK study finds
Frequent and varied eating found to reduce likelihood of perceiving immigrants as ‘cultural or economic threats’
3322
Chris Prosser @caprosser.com · 23/09/2025
Just spotted this excellently ironic typo in the acknowledgements of a paper I was reading 😂
A snippet of the acknowledgements section of a journal article in which the author has thanked the reviewers for their attention to detail, but misspelled 'attendtion'
1121
Chris Prosser @caprosser.com · 16/09/2025
With one line of code you can assign party colours and labels, and sort the legend. library(gbpartyscales) share_plot + scale_colour_party()
110
Chris Prosser @caprosser.com · 16/09/2025
Imagine you had a dataset of British elections results and wanted to make a plot. share_plot <- gb_shares %>% ggplot(aes(x = Election, y = `GB vote %`, colour = Party)) + geom_line() + geom_point(size = 3) + theme_bw() + scale_x_continuous(breaks = c(2019, 2024)) share_plot
100
Chris Prosser @caprosser.com · 11/09/2025
Today in 'I'm sure this is a well identified causal effect and definitely not in any way a spurious correlation, can't think of any potential confounding variables at all' 🤔
Newspaper headline saying 'Girls who play after-school sport in UK 50% more likely to later get top jobs, study finds'
1260
Chris Prosser @caprosser.com · 15/05/2025
Cool that they’ve found another copy of the Magna Carta, but I would like some more detail about this aspect of it’s provenance 😂 www.theguardian.com/uk-news/2025...
1144
Chris Prosser @caprosser.com · 13/05/2025
I think it’s simpler than that - there’s an MSOA called ‘Central Bradford’ and the 2021 census has 49.9% of its population born overseas.
111
Chris Prosser @caprosser.com · 08/05/2025
No, autocomplete, no.
1131
Chris Prosser @caprosser.com · 23/04/2025
To some extent you have to admire the ONS's steadfast refusal to allow you to easily work out the demographic characteristics of the adult (i.e. over 18) population from most of their tables, but mostly it's really annoying...
000
Chris Prosser @caprosser.com · 03/04/2025
Reading The Unaccountability Machine by @dsquareddigest.bsky.social at the moment, which I really recommend - one of those books that genuinely changes how you see the world. The last line of this section made me laugh out loud - not sure I’ve ever read a better description 😂
000
Chris Prosser @caprosser.com · 01/04/2025
Mucking about making some animated plots of constituency vote shares over time and accidently made some lava lamps.
0151
Chris Prosser @caprosser.com · 31/03/2025
Teaching my students about gganimate in their final R lab tomorrow - pretty pleased with this demo of GB vote shares from 1832 to 2024, and never fail to be impressed by this package - from line chart to this in one line of code!
3184
Chris Prosser @caprosser.com · 25/03/2025
They don’t actually mention the scaling /h in the book
100
Chris Prosser @caprosser.com · 13/03/2025
I've forgotten who I stole this example from, but thanks to whoever it was. It nicely illustrates the difficulty in accounting for confounding with observed variables and brings me joy whenever I come back to this lecture.
192
Chris Prosser @caprosser.com · 13/03/2025
Nothing like coming back to your quarto lecture slides, which worked perfectly well last year, and discovering there's been a package update... This sort of thing doesn't happen in powerpoint 😒
020
Chris Prosser @caprosser.com · 07/03/2025
My four-year-old, who is learning to read and write, has discovered that she can get to my iPad search screen without unlocking it, and if she types ‘cat’ will get lots of cat pictures to look at. For some reason they all have toothbrushes 😂
170
Chris Prosser @caprosser.com · 06/03/2025
Feels like the predatory publishing spambots aren’t even trying anymore.
020
Chris Prosser @caprosser.com · 05/03/2025
A British election factoid I hadn't noticed until I was updating my lecture slides - in 2024 only 15% of constituencies were won by a majority of votes. That is less than half the previous low of 34% recorded in 2010.
25312
Chris Prosser @caprosser.com · 05/03/2025
Putting some Taagepera and Shugart style predictive models into my lecture on the UK's electoral systems - never fail to be impressed by these things - here's predicted and actual party system size for European Parliament elections in the UK across the switch from first-past-the-post to PR.
030
Chris Prosser @caprosser.com · 01/03/2025
More academic disputes should have names like ‘the great porridge controversy’. (I confess to not having heard of the Norwegian sociologist Eilert Sundt before I read this, but he sounds great).
021
Chris Prosser @caprosser.com · 26/02/2025
It’s an old idea, but given the apparent paucity of idea in government at the moment, could be worth giving it a go…
020
Chris Prosser @caprosser.com · 24/02/2025
Because stuff we care about also varies geographically, this means that distance to random points will tend to be correlated with other variables. E.g. this is the distribution of correlations between distance to random points and Republican vote share. The mean correlation is -.2.
2130
Chris Prosser @caprosser.com · 24/02/2025
This problem exists in real maps too. E.g. here is a (lower 48 states) map of the congressional districts from the 2022 house of reps election. I do the same thing, randomly sample 10k points on the map, and then calculate the distance to those points. The edges of the map have longer distances.
270
Chris Prosser @caprosser.com · 24/02/2025
But if you calculate the *distance* from the centre of each tile to the random point, the distance is not equally spread out - it differs systematically by location. E.g. the mean distances from my above 10,000 simulations look like this
140
Chris Prosser @caprosser.com · 24/02/2025
The points will be equally spread out across the different tiles, e.g. this is the results from 10,000 simulated draws. So far, so simple...
120