Living meta-analysis dashboard on how voters punish politicians who break democratic rules

Across many democracies, researchers have run the same kind of experiment: show people a pair of candidates — alike in most ways, except that one has done something that breaks a democratic rule (ignoring courts, harassing journalists, trying to stay in power) — and see who they pick. This page pools 13 such studies, covering 32,935 people across 14 democracies, to ask two simple questions: how often do voters reject the rule-breaker, and does it matter whether that candidate is from their own party?

Each study asked people to choose between candidate profiles whose features (age, party, policies, and so on) were shuffled at random. Because only chance separates the profiles, the share who turn away from the rule-breaker is a clean measure of how much people are willing to defend democratic rules at the ballot box.

In every country, most voters reject the rule-breaker

Each dot is one country in one study: the share of voters who chose the candidate who respected democratic rules over the one who broke them. In all 14 democracies the share sits above half — from a low of about 53% to a high of about 73%. The vertical line marks the pooled average of 61%; the dashed line at 50% is a coin flip.

Bars show 95% confidence intervals. “A coin flip” (50%) would mean breaking democratic rules costs a candidate no votes at all; every country lands well above it.

Which kinds of rule-breaking cost the most?

“Breaking democratic rules” covers many different acts. Some studies put several of them in front of voters at once, which lets us ask which ones cost a candidate the most support. Each dot is that cost — how far a candidate’s support falls when they take the action instead of respecting the rule.

• Measured in several studies — a steadier estimate.
• From a single study — one country only, so read it with more caution.

Costs are drawn from the studies that report each action separately (Frederiksen 2022, Graham & Svolik, Mares & Visconti, Saikkonen & Christensen, and Scoggins). Closely related acts are grouped under one plain-language label; bars are 95% confidence intervals.

… but party loyalty has a price

Rejecting a rule-breaker is easy when the alternative is from your own side. The hard test is when the candidate who breaks the rules is from your own party, so punishing them means voting for the rival. The chart below splits the studies that let us see both situations.

• Loyalty not at stake — both candidates share the voter’s party, so rejecting the rule-breaker costs nothing.
• Crossing party lines — the rule-breaker is the voter’s own party, so punishing them means voting for the other side.

Pooled across these studies, rejection falls from 59% when loyalty isn’t at stake to 38% when it is — a drop of about 21 points. People still care about democratic rules, but partisanship pulls hard in the other direction.

Each row is one country in one study. The gap between the two dots is the “price of party loyalty” — how many voters stop defending democratic rules once doing so means crossing party lines.

The price of loyalty is much bigger for some parties than others

We can measure that price for individual parties. It ranges widely: supporters of some parties barely waver, while others drop their defense of democratic rules by 30–40 points once their own side is the one breaking them.

Each row is a party’s supporters. The dot is the price of party loyalty — how far their rejection of a rule-breaker falls when the rule-breaker is their own party. Bars are 95% confidence intervals; rows crossing zero are not statistically distinguishable from no penalty.

The bottom line has held steady as the evidence piled up

A fair worry about any single study is that the next one might overturn it. So here we replay the evidence in the order it arrived: each point is the pooled rejection rate using every study published through that year. Early on, with only a country or two, the estimate is shaky — but as studies accumulate it settles close to 61% and stays there.

Bars are 95% confidence intervals; they narrow as more countries and samples enter the pool. The dashed line marks today’s pooled average. The wide bar in 2018 shows how little could be pinned down when only two countries had been studied.

The studies

Every row is one published candidate-choice experiment in the analysis. Click a row to see the ways a candidate could break democratic rules in that study, what else the candidates differed on, and the share of voters who rejected the rule-breaker.

This is a living dashboard: we add new candidate-choice experiments as they appear, using the same design. If you have run a study that fits and it isn’t here yet, email marc.jacob@nd.edu and we’ll include it.

“Year” is when the survey was run. Studies marked with ⓘ in the party-loyalty analysis let us compare own-party and cross-party choices.