Mapping Hostility and Its Reception

Attacks Against Politicians in Germany, 2021–2025

Cornelius Erfort

Witten/Herdecke University

Elias Koch

Hertie School
(Data Science Lab/DYNAMICS)

June 19, 2026

Motivation

• Hostility toward politicians — widely salient; GLES: 78% online abuse, 30% threats (GLES 2023, 2025)
• Prior work on consequences, mostly from self-reports (Ponce 2019; Håkansson and Lajevardi 2024; Alizade et al. 2024)
• Gap: little systematic evidence on police-recorded incidents themselves, or on how politicians take them up in communication

The data

10,730 recorded incidents

• Federal government responses to Kleine Anfragen: lists of police-recorded offenses against politicians and party structures
• Per incident: date, location, targeted party, criminal offense; for a subset a short narrative of what happened

Linked sources

• Candidates & MPs (n = 2,929 federal; 1,266 state)
• GLES candidate studies 2021 + 2025 (N = 1,099)
• Facebook posts (n ≈ 430,000 from 337 politicians)

What does recorded hostility look like?

Mostly low-level: posters, graffiti, messages

Appendix: Examples from the records

Within each target type

Appendix: Static maps

Where attacks happen. Dots = incidents over population density. Attacks broadly follow people, but eastern clusters (Berlin hinterland, Saxony, Saxony-Anhalt, Thuringia) exceed what population density would predict. Static federal/state maps are in the appendix.

Patterns in the records

Who attacks whom

Appendix: Electoral calendar How they attack Federal by party State by party

Linking candidate surveys to police records

GLES 2021 + 2025 × attacks (N = 1,047)

Appendix: Perceived security Full GLES specifications

Attacks as a communication resource

Who talks about attacks?

Does talking about attacks pay off?

Within-politician estimates: percent change in engagement on attack-related posts vs. the same politician’s other posts (author + week FE; pooled reference in diamonds)

Appendix: Pooled estimates Robustness Heterogeneity

Takeaways

1. 10,730 attacks linked to politicians, surveys, and posts
2. Low-level, thinly spread — campaign-driven, ideological
3. Weak link between survey and police data
4. Uneven politicization — payoff in attention

Thank you

Cornelius Erfort · cornelius.erfort@uni-wh.de

Elias Koch · e.koch@hertie-school.org

References

Alizade, Jeyhun, Fabio Ellger, Marius Grünewald, and Thomas Tichelbaecker. 2024. “Does Political Violence Undermine Descriptive Representation? The Case of Women in Politics.” European Journal of Political Research.

GLES. 2023. GLES Kandidierendenstudie 2021. (ZA7704; Version 2.0.0) [Data set]. GESIS, Köln. https://doi.org/10.4232/1.14100. https://doi.org/10.4232/1.14100.

GLES. 2025. GLES Kandidierendenstudie 2025. (ZA10102; Version 1.0.0) [Data set]. GESIS, Köln. https://doi.org/10.4232/5.za10102.1.0.0. https://doi.org/10.4232/5.za10102.1.0.0.

Håkansson, Sandra, and Nazita Lajevardi. 2024. “Representation at Risk: Evaluating Levels and Consequences of Violence Against Immigrant-Background Politicians.” American Political Science Review, 1–16.

Ponce, Aldo F. 2019. “Violence and Electoral Competition: Criminal Organizations and Municipal Candidates in Mexico.” Trends in Organized Crime 22 (2): 231–54.

Appendix

Examples from the records

29.03.2022

Gera

Property damage · Party premises · The Left · Offline

Unknown perpetrators threw a stone at the shop window of the Die Linke constituency office and damaged the outer pane.

27.02.2023

Leipzig

Threat · Party premises · FDP · Mail

The FDP district office received a postcard by mail. Text: “If your war comes here, we’ll kill you. Period.”

12.06.2024

Lüchow

Insult · Party premises · SPD · In person

The accused entered an SPD MP’s office and insulted a staff member with the words “idiot”, “pig” and a homophobic slur; the insults were also directed at the openly gay MP.

Source: parliamentary inquiries (Kleine Anfragen).

Where attacks happen (static maps)

Federal targets (n = 1,968)

All targets (n = 7,874)

Attacks track the electoral calendar

Federal

targets; dashed lines = federal elections

State

sample; grey bars = covered legislative periods

How they attack

Federal

State

Verbal/online hate speech dominates; threats & physical violence are rare – left- and right-wing perpetrators use similar repertoires

Repertoires by party (federal)

Repertoires by party (state)

Perceived security and recorded attacks

(a) Federal targets, 6 months pre-election

(b) All targets through GLES field end

GLES: full specifications (federal, 6 months pre-election)

Engagement payoff: pooled within-politician estimates

Engagement payoff: robustness across specifications

Engagement payoff: heterogeneity