When Do Autocratizing Incumbents Lose Elections?

Interactive companion to the research paper.
Explore data from 105 re-election contests across 63 countries (1995–2024).

Overview

This website provides interactive access to the data underlying the paper. We identify 105 elections in which an autocratizing incumbent (or their party) sought re-election during an episode of democratic erosion. The dataset spans 63 countries from 1995 to 2024. A subsample of 51 post-2016 elections is hand-coded from Freedom House country reports to measure the scope and visibility of autocratizing acts.

105
Elections
63 countries, 1995–2024
29
Incumbents Defeated
28% overall defeat rate
39%
Defeated in Democracies
21 of 54 elections
16%
Defeated in Autocracies
8 of 51 elections

Geographic Distribution of Cases

Countries with at least one re-election contest during an episode of autocratization. Color indicates the number of elections in the dataset. Hover for details.

Election Explorer

Browse all 105 elections in the dataset. Each row is one re-election contest during an episode of autocratization. Polyarchy shows the V-Dem Electoral Democracy Index (0–1). Visibility, scope, and constraint indicators are within-country standardized. Click column headers to sort. Click any row with a ⓘ icon to see election details (incumbent, challenger, vote shares, and electoral system).

Country Year Regime Outcome Polyarchy Visibility Scope Horiz. Constraints

Key Concepts

Our analysis focuses on two dimensions of autocratization that shape whether incumbents can get away with democratic erosion at the ballot box.

Visibility of Autocratization

Not all forms of autocratization are equally visible to voters. We distinguish between high-visibility repression—overt, physically coercive, or incontrovertibly undemocratic actions such as jailing opposition leaders, shutting down media outlets, or using violence against protesters—and low-visibility repression—subtle, legalistic, or technical maneuvers such as court packing, regulatory changes to media ownership, or procedural restrictions on civil society organizations.

The key insight is that high-visibility repression is more electorally legible: ordinary citizens can recognize these actions as threats to democracy without requiring elite framing or specialized knowledge. Low-visibility actions, by contrast, may erode democratic institutions just as effectively but are less likely to trigger voter backlash because they are harder to detect and interpret.

Scope of Autocratization

Scope measures the breadth of democratic erosion—how many distinct institutional domains are simultaneously under attack. Rather than measuring the depth of any single dimension, scope captures whether an incumbent is targeting the judiciary and the media and elections and civil liberties at the same time.

We operationalize scope by tracking six V-Dem institutional pillars (freedom of expression, freedom of association, clean elections, civil liberties, judicial constraints, and legislative constraints). When an indicator declines by at least 0.05 on V-Dem's 0–1 scale relative to its recent average, that pillar is classified as “under attack.” The scope count is the number of pillars simultaneously eroding. Broader scope may signal to voters that democratic backsliding is pervasive, or it may indicate that the incumbent has consolidated enough power to prevent electoral accountability.

What Do We Find?

High-visibility repression is associated with a higher probability of incumbent defeat in the pooled sample. This effect is concentrated in democracies, where information environments are more open and voters can more readily translate their awareness of repression into electoral punishment. In autocracies, the relationship is weaker: even visible repression may not lead to defeat when elections themselves are compromised. Meanwhile, the erosion of horizontal constraints (checks and balances) also predicts defeat, suggesting that institutional degradation can signal regime vulnerability.

Bivariate Bootstrap Estimates

Each coefficient is from a separate bivariate OLS regression of incumbent defeat (1/0) on a standardized predictor. Estimates use pre-election means (t−3 to t−1). Thick bars show 90% bootstrap CIs; thin bars show 95% CIs (1,000 replications). Filled points denote significance at the 90% level. Positive values indicate a greater likelihood of incumbent defeat.

Scope and Visibility of Autocratization (Hand-Coded)

Based on 51 post-2016 elections hand-coded from Freedom House country reports. Each point is one election. Scope counts the number of distinct autocratizing tactics used by the incumbent in the 24 months prior to the election. Visible tactics are overt, public, or violent actions clearly legible as undemocratic; subtle tactics are procedural or institutional changes less likely to be noticed by ordinary citizens. Hover over a point for details.

Freedom House Coding Datasheets

Browse the detailed hand-coding for each of the 51 post-2016 elections. Each entry records an autocratizing action identified from Freedom House country reports, including the year, a description, the institutional target, and whether the action was coded as subtle or visible.

Year Action Target Code Type

Case Study Timelines

Explore how key indicators evolved relative to the election year (t=0) for each election. The x-axis shows years before and after the election. Select an election and indicator to plot. Note: some indicators may have missing values for certain years depending on data availability.