Political Analysis

Logo for Political Analysis with the text 'Political Analysis' in white font on a red background, above the hashtag 'OpenAccess' on a yellow background.

#OpenAccess from @polanalysis.bsky.social -

Detecting Formatted Text: Data Collection Using Computer Vision - cup.org/3IMLbjB

"This letter describes a workflow process for structured text extraction using free models and software"

- Jonathan Colner

#FirstView

Banner reading "POLITICAL ANALYSIS" in white capital letters on a red background, with a yellow strip bearing the hashtag "OpenAccess" below.

#OpenAccess from @polanalysis.bsky.social -

Bin-Conditional Conformal Prediction of Fatalities from Armed Conflict - cup.org/4mjbJHl

- David Randahl, Jonathan P. Williams & Håvard Hegre

#FirstView

They use three experimental designs to derive bounds for interactions between the treatment and the moderator. These bounds allow researchers to assess the sensitivity of findings to both priming and post-treatment bias. You can read the full paper here: www.cambridge.org/core/journal...

Currently in FirstView: In “Priming Bias Versus Post-Treatment Bias in Experimental Designs,” @mattblackwell.bsky.social, Jacob Brown, @sophieehill.bsky.social, Kosuke Imai, and Teppei Yamamoto analyze the trade-off between post-treatment and priming biases in survey experiments.

They illustrate this method by studying the relative importance of partisan, racial, gender, and religious identities. 30% of respondents offered random responses and not accounting for this can affect substantive conclusions. You can read the full paper here: www.cambridge.org/core/journal...

Currently in FirstView: In “Addressing Measurement Errors in Ranking Questions for the Social Sciences,” Yuki Atsusaka and @sysilviakim.bsky.social examine the statistical consequences of measurement error and introduce a framework for improving ranking data analysis.

Using simulations, the authors demonstrate MAUP-related inconsistency in regression results. This paper identifies both MAUP concerns and best practices. You can read the full article here: www.cambridge.org/core/journal...

Currently in FirstView: In “The Modifiable Areal Unit Problem in Political Science,” @dwlee.bsky.social, @melissazrogers.bsky.social‬, and Hillel David Soifer discuss the MAUP and how the size of spatial units and the location of their borders affect empirical results.

Promotional graphic for "Political Analysis" featuring text 'New Issue Online' set against a yellow and red background.

NEW ISSUE from @polanalysis.bsky.social -

Political Analysis - Volume 33 - Issue 3 - July 2025 - cup.org/4nduNIG

All articles in this issue are #OpenAccess.

The authors use eye-tracking to study the effects of odd attribute combinations of survey respondents. The effect of these combinations is minimal and does not impact respondent attention, search, or choice behavior substantially. Read the full paper here: www.cambridge.org/core/journal...

Currently in FirstView: In “Odd Profiles in Conjoint Experimental Designs: Effects on Survey-Taking Attention and Behavior,” Kirk Bansak and Libby Jenke consider how survey-takers respond to odd combinations of conjoint attributes.

Logo graphic for "Political Analysis" with the hashtag "OpenAccess" on a yellow and red background.

#OpenAccess from @polanalysis.bsky.social -

Fixed Effects, Lagged Dependent Variables, and Bracketing: Cautionary Remarks - cup.org/4dS2taa

- Matei Demetrescu, Manuel Frondel, Lukas Tomberg & Colin Vance

#FirstView

The CSAS method is then applied to several examples, including the identification of local political concerns. CSAS identified both locally salient issues and national issues not traditionally associated with local government. Read the full paper here: www.cambridge.org/core/journal...

Currently in FirstView: In “Crowdsourced Adaptive Surveys,” @yamilrvelez.bsky.social introduces a methodology (CSAS) that converts open-ended text from participants into survey items and applies a multi-armed bandit algorithm to determine which questions should be prioritized in the survey.

These results demonstrate that “asking” instruction-tuned LLMs for the position of short texts produces valid position estimates. The position estimates of the best LLMs are as accurate as crowdsourced position estimates. You can read the full paper here: www.cambridge.org/core/journal...

Currently in FirstView: “Positioning Political Texts with Large Language Models by Asking and Averaging.” @glemens.bsky.social‬ and @ainagallego.bsky.social use a variety of LLMs to take in political texts and position political actors in policy and ideological space.

The paper uses simulations to estimate pivotal probabilities in both IRV and single-member plurality electorates. These pivotal probabilities are a crucial ingredient in formal treatments of voting as a strategic activity. Read the full paper here: www.cambridge.org/core/journal...

Currently in FirstView: “The Probability of Casting a Pivotal Vote in an Instant Runoff Voting Election.” @samuelbaltz.bsky.social derives all the ways that a single ballot can change the result of an instant runoff voting election, for any number of candidates and voters.

Words used in open-ended prompts tend to be correlated over survey waves. Their new method, which leverages variation in contextually common words, provides efficient quantitative representations of open-ended responses. You can check out the paper here: www.cambridge.org/core/journal...

Currently in FirstView: “Categorizing Topics Versus Inferring Attitudes: A Theory and Method for Analyzing Open-ended Survey Responses.” @willrhobbs.bsky.social and @jongreen.bsky.social outline and validate a theory to separate noise from signal in open-ended responses.

They find that confidence intervals from single-topic studies capture the typical effect from a population of topics at well below the nominal rate. You can check out the paper here: www.cambridge.org/core/journal...

Currently in FirstView: In “The Limits (and Strengths) of Single-Topic Experiments,” @scottclifford.bsky.social & @carlislerainey.bsky.social examine the generalizability of single-topic studies, focusing on how often confidence intervals capture treatment effects from a larger population of studies

He provides a conceptual overview of text generation, offers guidance for researchers, and offers a technique for improving text quality. This method is validated using three unique examples. You can check the paper out here: www.cambridge.org/core/journal...

Currently in FirstView: “Synthetically generated text for supervised text analysis.” @ahalterman.bsky.social proposes using LLMs to generate synthetic training data for training smaller, traditional supervised text models.

When comparing representations in high dimensional space, distance measures can exhibit statistical bias. The authors introduce and validate a bias correction for the squared Euclidean distance. You can check out the paper here: www.cambridge.org/core/journal...

Currently in FirstView: “Measuring Distances in High Dimensional Spaces Why Average Group Vector Comparisons Exhibit Bias, And What to Do about it” by Breanna Green, @willrhobbs.bsky.social, Sofia Avila, Pedro Rodríguez, Arthur Spirling, and @bstewart.bsky.social

✨Sharing our new publication in @polanalysis.bsky.social @cambridgeup.bsky.social "Constructing Vec-tionaries to Extract Message Features from Texts: A Case Study of Moral Content" #OpenAccess

www.cambridge.org/core/journal...

Using elections in Colombia and Finland, the authors highlight the sensitivity of RDD estimates to implementation choices. Ultimately, the type of implementation makes little difference when the CEF around the cutoff is approximately linear. You can read it here: www.cambridge.org/core/journal...

Currently in FirstView: "When Can We Trust Regression Discontinuity Design Estimates from Close Elections? Evidence from Experimental Benchmarks" by Leandro De Magalhães, Dominik Hangartner, Salomo Hirvonen, @jmerilainen.bsky.social, Nelson Ruiz, and @jannetukiainen.bsky.social.

Using the overlaps in 95% confidence intervals to perform significance tests does not work. The authors provide packages in R and Stata to overcome these problems. You can check out the paper (which is still open access!) here: www.cambridge.org/core/journal...