MIGRATION AND LABOR LAWS IN COMPARISON
Hate speech, voting and the legal protection of non-citizens
Hostile speech about migrants circulates in constitutional democracies of every region, and parties that campaign against immigration win large shares of the vote in some of them and very little in others. This comparison asks two linked questions across 34 democracies, 18 Western and 16 non-Western. First, how closely is hate speech in party politics associated with votes for parties that oppose immigration? Second, do countries whose law protects non-citizens explicitly, in the constitution or in a statute written for migrants, see less of either than countries that rely on general guarantees written with citizens in mind?
- The question
- Legal protection at a glance
- Speech and votes
- What the data show
- Impunity
- Explore the data
- Country profiles
- Method and codebook
- From research to practice
- Sources and downloads
I · THE QUESTION
Conventional and explicit protection
Most constitutions and anti-discrimination laws were drafted for citizens. Some speak of "all citizens" and leave foreigners to judicial interpretation; many protect "everyone" or "every person" without saying whether that includes people without the nationality of the state; and many list race, ethnicity or national origin as protected grounds, which courts have often read to exclude nationality and immigration status. A smaller group of countries protects non-citizens in so many words: the constitution states that foreigners enjoy the same rights, or a migration statute guarantees migrants equal treatment and outlaws xenophobic discrimination. International law points in the same direction: the UN Committee on the Elimination of Racial Discrimination asks states to protect non-citizens expressly (General Recommendation No. 30) and to address racist hate speech against them (General Recommendation No. 35).
Hypotheses
H1. Party hate speech is less frequent in countries whose law explicitly protects non-citizens.
H2. Parties that oppose immigration win a smaller share of the vote in those countries.
H3. The association between party hate speech and the electorate's support for restricting immigration is weaker in those countries.
The null hypothesis in each case is that there is no difference. The hypotheses were fixed before the first run of the data, and the five tests for H1 and H2 are corrected together for multiple comparisons.
What this comparison can and cannot show
It shows associations across countries at one point in time. Explicit protection may reduce hostile rhetoric, but countries that are already more open may also be more likely to adopt explicit protection, and electoral rules, history and the size and origin of migrant populations shape both. The findings are a starting point for case studies and for a longer panel, not evidence of effect.
Hate speech often reaches migrants through religion, as scholarship on the religious incorporation of migrants describes (Martikainen 2020), and a national-origin or religion ground may protect them in practice even where nationality is not named. The coding records the words of the law; the notes record how courts read them.
II · LEGAL PROTECTION AT A GLANCE
How explicitly each country's law protects non-citizens
Four features of domestic law, each coded 2 when the text names non-citizens, nationality or citizenship, 1 when it uses general wording, and 0 when there is no protection of that kind. A country is in the explicit group when its constitution or a migrant-specific statute scores 2; all others are in the conventional group. The order is regional, not a ranking.
III · SPEECH AND VOTES
Party hate speech and the anti-immigration vote
Party hate speech is the V-Dem expert estimate of how often major parties use hate speech, with the sign reversed so that higher means more often; the scale is centered near zero across all countries in the world. Vote measures come from the most recent national legislative election in 2015–2019 that V-Party covers. "Restrictive" parties are those V-Party experts place closer to opposing immigration than to a neutral position; "nativist" parties are restrictive parties that also appeal to the people against elites at least half the time. The electorate's stance is the vote-weighted average immigration position, from 0 (support) to 4 (strong opposition). Click a column heading to sort.
V-Dem measures hate speech against any group, not only migrants. Elections from 2020 onward, including the gains of anti-immigration parties in several European countries and in Japan, are not yet in the vote data; see the method section.
IV · WHAT THE DATA SHOW
Explicit and conventional protection compared
Group means, Welch's t-test, a permutation test with 5,000 reshuffles of the group labels, and Holm's correction across the five tests. The figures are computed in your browser from the published data, so they change when the data are corrected.
How closely hate speech and votes move together
Pearson's r with a 95% confidence interval, and Spearman's ρ, which is less sensitive to outliers such as Sri Lanka, where V-Party places both main parties as restrictive.
Is the link weaker where protection is explicit? (H3)
Ordinary least squares: electorate stance on party hate speech in the election year, explicit protection, their interaction, the liberal democracy index and a Western indicator.
V · IMPUNITY
Are crimes against migrants punished?
Impunity has a legal side and a practical side. On the legal side, the table records whether criminal law treats hostility toward a victim's nationality or foreignness as an aggravating circumstance, coded like the features above but kept out of the index. On the practical side, it shows V-Dem's indices of access to justice and of equal protection across social groups (0 to 1, higher is better) and the rule of law. These indices are not specific to migrants.
The most direct measure of impunity is attrition: how many hate crimes recorded by the police are prosecuted and how many end in a sentence. The OSCE Office for Democratic Institutions and Human Rights publishes these figures for its participating states, which include the 16 European and North American countries here. The table shows the median yearly share of police-recorded hate crimes that were prosecuted and that ended in a sentence, 2019–2024. ODIHR publishes these stages only for all hate crimes together, not by bias, and police, prosecutors and courts count cases differently, so the shares are approximate; years above 100% are left out. Six countries report no sentencing figures at all. No comparable figures exist for the Asian, African and Latin American countries compared, where hate crimes against migrants are rarely recorded as such.
VI · EXPLORE THE DATA
Test your own question
Redefine what counts as explicit protection, choose which countries to include and weight the legal features yourself. Then compare groups, check whether two measures move together, or fit a regression with controls. Results show confidence intervals, p-values and a plain reading, and you can download the dataset you built.
The three hypotheses under your settings
The same five tests as in section IV, rerun with your definition and your choice of countries. Section IV keeps the definition fixed in advance; compare the two to see how much the result depends on coding choices.
Popularity and hate crime
Does hate crime run higher where anti-immigrant parties win more votes? Choose a measure of their popularity and a measure of hate crime or group violence, then compare. The tool reports Pearson and Spearman correlations, a permutation test and, if you wish, a regression slope with controls, and it tests every pairing at once with Holm's adjustment. Cross-national hate crime data are scarce: the Pew and ACLED measures cover all 34 countries but are not specific to migrants, and the OSCE ODIHR figures cover only the 16 European and North American countries.
The data is loading.
Your own analysis
The data is loading.
VII · COUNTRY PROFILES
Each country in detail
VIII · METHOD AND CODEBOOK
How the data was built
The unit of analysis is the country. The legal features record the law on paper (de jure) as in force on the coding date, with the year each provision took effect so that the law can be matched to the year of the election.
Case selection
Constitutional democracies with competitive national elections, chosen to balance Western democracies (Western Europe, North America, Australia and New Zealand) and non-Western democracies in Asia, Africa and Latin America, and to include the main destination countries for migrants in each region. India, Indonesia, Mexico and the Philippines are classified by V-Dem as electoral autocracies in 2025; they are kept because they hold competitive elections under written constitutions, and the democracy variable lets you set them aside. The Western and non-Western labels are conventional and contestable.
Coding rules
A score of 2 requires the words of the provision to name non-citizens, foreigners, migrants, nationality or citizenship. "Nation" and "national group" count as naming nationality; "national origin" alone does not, because courts have often read it as ancestry (for example Espinoza v. Farah in the United States and Ealing v. Race Relations Board in the United Kingdom). Interpretations by courts are recorded in the note, not in the score. Every cell is marked S (secondary) until it has been checked against the official text.
Outcome measures
Party hate speech: V-Dem v16, item v2smpolhate, 2025 and the election year, sign reversed. Online targets: v2smhargr, share of experts naming religious or ethnic groups. Votes: V-Party, items v2pavote, v2paimmig and v2papeople, original-scale means; where V-Party reports an alliance without a position, the alliance takes the position of its leading party, as shown in the profile. Kenya has no usable V-Party election in 2015–2019.
Limitations
Thirty-four countries give little statistical power and the countries are not a random sample. V-Dem's hate speech item is not specific to migrants. The vote data stop in 2019, while the legal coding and the 2025 speech measure are recent. Expert surveys may themselves be shaped by a country's legal culture. The impunity indicators are partly general rather than migrant-specific, and the OSCE ODIHR figures cover the European and North American countries only and depend on how much each police force records.
Evidence from hate crime and local studies
Hate crime research points the same way. A scoping review of 76 European studies from the past decade finds that terrorist attacks, the 2015 refugee arrivals and the COVID-19 pandemic were followed by rises in hate crime, and that cross-national and longitudinal studies remain scarce (Suárez-Vergne 2026). In the United States, an analysis of the National Crime Victimization Survey finds that non-citizens have higher odds of hate crime victimization than people born in the country, while official hate crime statistics do not record legal status (McCann and Boateng 2022). National averages also hide local variation. In Santiago, in Chile, which guarantees migrants equal treatment by statute, reports on a citizen reporting platform that mention migrants are more likely to contain hate speech, and the hate speech concentrates in neighborhoods with many recent arrivals rather than in established migrant communities (Graells-Garrido et al. 2026, preprint). Where the law does not name non-citizens, the data often do not count them either.
How to cite
Institute for Migrant Rights (2026). Hate Speech, Voting and the Legal Protection of Non-Citizens: A Comparative Dataset of 34 Constitutional Democracies, version 1.0. https://theimr.org/hate-speech-voting/
IX · FROM RESEARCH TO PRACTICE
Using the comparison
The coding shows, for each country, which legal provision a migrant targeted by hate speech or discrimination can rely on, and whether that provision names them. Advocates can use the gaps it identifies in submissions to national human rights institutions and to the UN treaty bodies, which regularly ask states to protect non-citizens expressly.
X · SOURCES AND DOWNLOADS
Sources and downloads
Static CSV files: Country data · Legal codings with notes and sources · Codebook · Parties · Treaty status · BibTeX. Licensed CC BY 4.0. V-Dem and V-Party data are used under their own terms; cite them as listed below.
This page summarizes laws and published data for research and advocacy. It is not legal advice. Found a mistake? Write to secretariat@theimr.org.