Chris Hanretty

@chanret.bsky.social

I teach politics at a university in the UK. I'm interested in electoral systems, public opinion, and the politics of non-majoritarian institutions like courts and regulators. ORCID: 0000-0002-8932-9405

Google search box with text "odyssey parent and child screening"

Cover of Patrick Radden Keefe's London Falling

Paper plate turned upside down, with a triangle filled with interior polygons drawn on the paper plate in felt tip pen

Electoral democracy scores from V-Dem. Value for the US in most recent year: 0.73; value for Paraguay 0.58. Credible intervals don't overlap.

Electoral democracy measures from V-Dem; Canada's figure is consistently higher than the figure for Bosnia.

Reverse "Kombucha girl" meme: left panel: "Data exists in XML", happy face; right panel, "XML is malformed", unhappy face.

Chart showing levels of electoral democracy in Czechia and South Korea from the V-Dem project. Current levels: South Korea: 0.82 [CI: 0.78 - 0.88]; Czechia 0.85 [CI: 0.83 - 0.91]

Chart showing levels of electoral democracy in Mexico and South Africa from the V-Dem project. Current levels: South Africa: 0.74 [CI: 0.69 - 0.81]; Mexico 0.47 [CI: 0.42 - 0.53]

Map of the world, with UN regions colour-coded

Table showing excess death rates for different UN regions from the linked study

Regression table showing no significant associations on representation for dummy variable measuring proportionality, and Gallagher index; significant associations for district magnitude

Table of results taken from https://democracy.wandsworth.gov.uk/mgElectionResults.aspx?XXR=0&ID=12&AC=EDIT_ELECTION&RPID=212461733

Quote from report: "the fact that political attention has shifted towards these voters, who do not feel that universities have benefited them, drives much of the conversation about universities. It is perhaps the biggest political challenge universities face. The increased political salience of widening participation can be seen through this lens, as can concern about immigration and international students."

Screenshot of Claude output. Original paper: https://doi.org/10.1111/sjos.12575

R code which shows a data frame being created through a call to a function named `collapse_netherlands`

$Formula: Bayesian R^2 = \frac{V_{n=1}^N y_n^{pred s}}{V_{n=1}^N y_n^{pred s} + var_{res}^{s} where var_{res}^{s} = (\sigma^2)^s for a linear regression model with equal variances$

Two posts, one by Steve Fisher predicting Reform will win 2260 council seats, and one about Reform asking a rival party councillor to stand as a paper candidate

Over the last thirty years the West has preoccupied itself thanklessly in an expeditionary capacity in the invertebrate civil wars of others. It ought to have learned that it is impossible to maintain an integrated multi-valent society once neighbours start kidnapping each other’s children and murdering them with hand drills, blowing up each other’s cultural events, slaying each other’s teachers and religious leaders, and tearing down their icons. It is soberingly worth noting, moreover, that plenty of instances of all those things have occurred already in the West and all of them have occurred in France alone in the last five years.

XKCD web-comic, "Dependency". A large array of shapes at top is labelled "All modern digital infrastructure"; a slender column supporting these shapes is labelled "A project some random person in Nebraska has been thanklessly maintaining since 2003"

"Here as elsewhere alliances will play a vital role in between the two rounds of voting.

Under the complex voting system, it is quite possible that as many as five candidates in the capital will make it through to the run-off on 22 March."

Quote reads: "You get a lot of people that like to write you off as soon as you reach 40, it's all downhill from there, is what you hear," she said. "I think Elana and I are both proof that that's not true.""

Lamppost with two stickers attached: one sticker of Diana, Princess of Wales; below, a heart-shaped sticker with text, "English Common Law"

$Mathematical formula. LaTeX code: \underbrace{\frac{1}{N'_S}}_{\substack{\text{Inverse of}\\ \text{district figure}}} = \underbrace{\color{OliveGreen}\frac{1}{N_S}}_{\substack{\text{Inverse of}\\ \text{national figure}}} + \underbrace{\frac{S - {\color{BurntOrange}m}}{S - 1}}_{\substack{\text{Finite} \\ \text{pop.}\\ \text{correction}}} \cdot \frac{1-{\color{OliveGreen}\frac{1}{N_S}}}{{\color{BurntOrange}m}}$

Extended quote from Shugart and Taagepera, Votes from Seats:

Mathematical equations omitted in this alt text.

"We have already seen in Chapter 7 that for the number of parties winning at least one seat, the model without adjustment for embeddedness, is a good fit. But this lack of necessary adjustment applies only to [this quantity]. We show in this chapter that adjustment is absolutely needed to make sense of other outcomes of interest, including the effective numbers of both seats and votes (as Figures 10.1 and 10.2 at the start of this chapter showed). The details of the derivation of this embeddedness parameter, k, are complex. The resulting
formula is [omitted]
The formula is baffling, and those readers wanting to see our logic and many steps of calculation should refer to the appendix, where we spell it out in detail.

Title: "Electoral districts as samples from the national population"

Abstract: The relationship between national and district-level electoral is a foundational topic in the study of elections. I argue that selected district-level electoral outcomes can be modelled as though they were the outcome of a process involving random sampling without replacement. For seat-level outcomes, district magnitude is analogous to sample size. For vote-level outcomes, I argue that the role of sample size is played by a pseudo-magnitude, or district magnitude plus a constant. My theory gives point predictions and variances for the effective and 'raw' numbers of seat-winning parties, and the effective number of vote-winning parties. I estimate linear and nonlinear Bayesian regressions to test this theory using CLEA data. The principal coefficients in these regressions are within one to three percent of their predicted values; predictions from this theory are between 2% and 45% more accurate than predictions from the leading account of district-level electoral outcomes.

$R code: ```{r} ### What if we just model the postcode effect as a random intercept? gam_postcode_re <- gam(index_nimbyism ~ educ + sex + party + interior_quality + treated_rd_relative + s(years_rel_2013, by = treated_rd_relative, k = 2) + s(post_code, bs = 're'), data = df) ### Effect of -0.223 (-0.465 to 0.0186) tidy(gam_postcode_re, parametric = TRUE, conf.int = .95) |> filter(term == "treated_rd_relative") ```$

$R code: ```{r} ### (2) with bezirke rd_bezirke <- rdrobust(y = df$index_nimbyism, x = df$years_rel_2013, c = 0, h = 5, b = 5, covs = model.matrix(~ educ + sex + interior_quality + bezname - 1, data = df)[,-1]) ### Effect of -0.19 SDs (95% CI: -0.952, 0.250) summary(rd_bezirke) ```$

$R code: ```{r} ### Load libraries library(tidyverse) library(here) library(rdrobust) library(mgcv) library(broom) here::i_am("source_results/cjh_edit.R") # Read main dataset df <- read_rds(here::here("data", "data_main.rds")) |> filter(!is.na(treated_rd_relative)) # Standardize continuous outcomes df <- df |> mutate(index_nimbyism = as.vector(scale(index_nimbyism))) ### Estimate RDDs ### (1) with postcode rd_post_code <- rdrobust(y = df$index_nimbyism, x = df$years_rel_2013, c = 0, h = 5, b = 5, covs = model.matrix(~ educ + sex + interior_quality + post_code - 1, data = df)) ### Effect of -0.6 SDs (95% CI: -1.264, -0.273) summary(rd_post_code) ```$