Europeans' support for refugees of varying background is stable over time

The study investigates how European public attitudes toward refugees respond to repeated humanitarian crises, specifically comparing the Syrian refugee crisis (2015–2016) and the Ukrainian crisis (2022). Competing hypotheses are evaluated: that the Ukraine war could revive European solidarity and increase support for refugees, potentially more for Ukrainians, versus that repeated large inflows, economic pressures, and rising populism could reduce support or polarize it, and possibly crowd out support for non-Ukrainian groups (e.g., Muslims). Given Europe’s central role in the global asylum regime, understanding the stability and evolution of public preferences is critical. The authors aim to assess whether preferences over refugee attributes have changed, whether overall support has shifted, and whether any increase for Ukrainians comes at the expense of others.

The paper builds on research showing voter attitudes influence policy in democracies and on a broad literature identifying drivers of attitudes toward immigrants and, to a lesser extent, refugees. Prior work during the Syrian crisis highlighted the roles of economic, humanitarian, and religious considerations in shaping preferences. The study also engages with debates on potential preferential treatment of Ukrainians due to perceived cultural fit, race, religion, and education, as well as concerns about asylum fatigue in the context of rising populism. Methodologically, it leverages conjoint analysis literature for causal inference on multidimensional preferences and acknowledges evidence on the external validity of conjoint designs, including validation against real-world behavior and eye-tracking studies. Supplementary analyses reference European Social Survey trends indicating stability in immigration attitudes over two decades.

Design: Two large-scale public opinion surveys using paired-profile conjoint experiments were fielded in 15 European countries during two waves: 2016 (amid the Syrian crisis) and May–June 2022 (amid the Ukraine war). The 2016 wave included ~18,000 vote-eligible respondents; the 2022 wave included ~15,000. Respondents evaluated randomized asylum seeker profiles varying across multiple attributes. Attributes: Country of origin (e.g., Syria, Afghanistan, Eritrea, Iraq, Pakistan, Kosovo, Ukraine), religion (Muslim, Christian), reason for migrating (political persecution, religious persecution, economic opportunities; ‘war’ added in 2022), asylum testimony consistency (no inconsistencies, minor inconsistencies, major inconsistencies), gender, age (e.g., 21 vs 62 years), previous occupation (unemployed, skilled worker, teacher, farmer, doctor), special vulnerabilities (no surviving family, victim of torture, handicapped), and host-country language skills (fluent, none). Outcomes: Primary outcome was forced choice between two profiles (binary indicator of chosen profile). Secondary outcomes included a 1–7 rating of each profile and a dichotomized acceptance indicator derived from ratings. A separate feeling thermometer (0–100) captured warmth toward groups (Ukrainians, other origins, compatriots) in 2022. Sampling and weighting: Samples were re-weighted via entropy balancing to match each country’s age, gender, and education margins; robustness checks used unweighted data and alternative weights also balancing political ideology. The design and analyses were preregistered (https://osf.io/jd8n3/). Analysis: Linear least-squares regressions with cluster-robust standard errors estimated average marginal component effects (AMCEs) on forced choice. Statistical tests included two-sided t-tests and two one-sided tests (TOST) for equivalence. Comparative analyses excluded 2022 profiles randomized to ‘war’ for fair comparison with 2016. Subgroup analyses examined heterogeneity by country and respondent ideology; additional tests considered European solidarity as a moderator and assessed potential substitution effects across origins and religion.

• Warmth toward Ukrainians (2022): Mean feeling thermometer score for Ukrainian asylum seekers was 62.5 versus 42.7–46.9 for other non-Ukrainian groups; compatriots scored 79.5 (t = 50.08–62.92; max P < 0.00001; n = 14,856).
• Stability of attribute preferences (2016 vs 2022): Effects of key attributes on forced choice were remarkably similar across waves. Representative 2022/2016 effects (percentage points, pp):
  • Major inconsistencies in testimony: −9.6 to −10.7 pp vs none (t = 23.42–28.38; max P < 0.00001; n = 178,740 in 2016; n = 148,460 in 2022).
  • Male vs female: −5.8 to −6.0 pp (t = 18.72–19.64; max P < 0.00001).
  • Age 62 vs 21: −5.9 to −6.1 pp (t = 14.39–15.54; max P < 0.00001).
  • Previously employed vs unemployed: +4.8 to +14.2 pp; high-skilled (accountant/teacher/doctor) +7.6 to +14.2 pp (t = 9.06–25.05; max P < 0.00001).
  • Victim of torture vs none: +9.8 to +11.2 pp (t = 18.95–23.65; max P < 0.00001).
  • Muslim vs Christian: −8.8 to −10.7 pp (|t| = 21.39–28.04; P < 0.00001).
  • No language skills vs fluent: −11.1 to −11.7 pp (t = 27.19–30.08; P < 0.00001).
  • Economic reasons vs persecution/war: −13.3 to −18.9 pp (t = 25.50–36.26; max P < 0.00001).
• Minimal differences across time: Only 4/21 attribute-effect differences (excluding origin) were statistically significant at P < 0.05; largest absolute difference was 1.9 pp. TOST equivalence tests with ±3 pp bounds rejected non-equivalence for all differences (P < 0.05), indicating practical equivalence.
• Country-of-origin (Ukraine) effect: In 2022, being Ukrainian increased choice probability by 5.5 pp compared with non-Ukrainians (t = 11.71; P < 0.00001; n = 148,460), versus 0.9 pp in 2016 (t = 2.13; P < 0.05; n = 178,740). The 4.6 pp difference across waves was significant (t = 7.18; P < 0.00001). However, the Ukraine effect was smaller than effects for religion, skills, and gender in 2022.
• European solidarity moderation: The 2022 Ukraine effect was 7.5 pp among respondents with higher European solidarity vs 4.7 pp among those with lower solidarity; difference 2.7 pp (|t| = 2.61; P < 0.01; n = 148,460). Supplementary analyses suggest causal moderation.
• No general preference for European origin: In 2022, asylum seekers from Kosovo were not treated differently than those from Syria, Afghanistan, Eritrea, Iraq, or Pakistan (effects not significant; TOST within ±3 pp rejected non-equivalence).
• Overall acceptance increased: Pooling countries, the share of profiles accepted rose by 4.9 pp from 2016 to 2022 (t = 13.41; P < 0.00001; n = 178,740 in 2016; n = 118,807 in 2022; ‘war’ profiles omitted). Twelve countries showed significant increases; none showed decreases.
• No substitution toward Ukrainians: Acceptance increased for both Ukrainian and non-Ukrainian profiles in most countries; no country had a negative point estimate for non-Ukrainians. Acceptance of Muslim profiles increased significantly in 9 countries; only Sweden showed a small, non-significant decrease (−0.62 pp).
• Across ideology: In 2022, left-wing respondents accepted 61.3% of profiles vs 42.9% for right-wing (difference 18.4 pp; t = 28.10; P < 0.00001). Acceptance increased from 2016 to 2022 by 6.5 pp (left) and 4.4 pp (right) (both P < 0.00001).
• Additional indicators: Categorical rejecters decreased by 3.0 pp; support for granting asylum increased by 1.3 pp at home and 4.5 pp in Europe (t = 2.83–10.43; P ≤ 0.005).

The findings address the key questions by showing that European preferences over refugee attributes are highly stable across two major crises and that overall support has slightly increased, contradicting expectations of asylum fatigue or zero-sum substitution. The modest Ukraine-specific boost in 2022 appears secondary to long-standing preferences for traits that Ukrainian refugees disproportionately possess (female, younger, Christian, higher education/skills), implying that observed support reflects enduring evaluative criteria rather than a transient surge. Increased support extends to non-Ukrainians and Muslims and is evident among both left- and right-leaning respondents, indicating broad-based resilience rather than polarization-driven shifts. These results suggest that public attitudes toward asylum seekers are durable and that crisis-induced inflows of a neighboring population do not reduce support for other groups. The robustness of the conclusions is supported by preregistered analyses, weighting schemes, multiple outcome measures, and validation literature on conjoint external validity.

The study demonstrates that European support for refugees is resilient: attribute-based preferences are stable between 2016 and 2022, and overall acceptance has modestly increased. Support for Ukrainians in 2022 is driven chiefly by enduring preferences for specific traits, with a smaller, additional Ukraine-specific solidarity effect. Crucially, increased support does not come at the expense of non-Ukrainian or Muslim refugees, and gains occur across the ideological spectrum. These insights inform theory by highlighting the stability and generosity of asylum attitudes even amid repeated crises and rising populism, and inform policy by indicating durable public consensus on preferred refugee profiles and limited risk of crowd-out when prioritizing Ukrainians. Future research could track dynamics between crises with higher-frequency measurements, investigate causal mechanisms behind solidarity effects, and assess policy communication strategies that align with stable public preferences.

Two main limitations are noted. First, as with all surveys, there are external validity concerns related to the measurement instrument and sample representativeness. Nonetheless, prior validation shows conjoint choices correlate with real-world behavior and eye-tracking indicates alignment between stated choice and attention; results are robust to alternative weighting (including ideology) and unweighted analyses, and the same sampling mechanism was used in both waves. Second, data were collected at two points (2015–2016 and 2022) but not in between, limiting visibility into interim fluctuations. Supplementary analyses using European Social Survey data (2002–2020) suggest long-run stability in related immigration attitudes, consistent with the present findings, but continuous measurement would be needed to fully characterize temporal dynamics.