PEPSY Core Paper 1 - Final Hauptfeld Analysis

Author

POLITSOLID Team

1 Setup

The final analysis uses data/POLITSOLID_Hauptfeld_weighted_18052026.dta. Main analyses are estimated without population-size weights because the paper studies cross-national patterns across seven countries rather than aiming to estimate an EU-population average. The cleaned file contains 8066 interviews across 7 countries. The primary outcome is the preregistered territorial solidarity index, computed as the mean of willingness to pay for redistribution at the municipal, country, and EU levels.

The region match linked NUTS context to 8066 respondents. Regional public-service quality from EQI qualityp is available for 7319 respondents, using the 2021 full-NUTS2 file so German respondents can be matched at NUTS2. Relative regional income is available for 8066 respondents.

2 Preregistration Coverage

The local preregistration reference is preregs/core1_prereg.txt. The table checks whether the final file contains the variables needed for each preregistered hypothesis.

hypothesis status available_items missing_or_note
H1 Ideology Fully testable poleconred, poleconsmr, leftright
H2a Objective resources Fully testable country income item(s), slfstatus
H2b Perceived capacity Fully testable fresil, ecohhld
H3a Objective need/risk Fully testable public payment items, EmploymentStatusKP
H3b Perceived risk Fully testable prskcope, prskunemp, prskunwork, prskcare
H4 Norms and contribution Partially testable taxnorm, fraudnorm rednrmdsc is absent in the Hauptfeld .dta; H4 uses the two available norm items
H5a Generalized social trust Fully testable soctrst, percfair, perchelp
H5b Institutional trust Partially testable trstparl, trstjudi, trstmuni, trstgov, trsteucom, satdem trstpltcns and trstparts are absent in the Hauptfeld .dta
H5c Public-service evaluations Fully testable pssatgen and public-service quality/fairness items
H6a Regional public-service quality Fully testable EQI qualityp 2021 full-NUTS2 file used for service-quality context
H6b Relative regional prosperity Fully testable Eurostat regional GDP per capita

3 Descriptive Europe Maps

These maps summarize regional patterns in the matched survey/context data. They are descriptive checks and should not be read as model-predicted effects.

4 Country Outcome Means

The table reports unweighted country means for the main solidarity outcomes, matching the estimand used in the models and maps.

country n_interviews analysis_interviews Deserving solidarity Migrant solidarity Territorial solidarity Deserving-minus-migrant gap
Croatia 1124 1124 3.905 2.782 3.049 1.122
France 1115 1115 3.708 2.555 3.055 1.153
Germany 1600 1600 3.899 2.778 3.202 1.121
Italy 1117 1117 3.993 3.187 3.394 0.806
Poland 1005 1005 3.777 2.764 3.231 1.014
Spain 1054 1054 4.094 3.201 3.543 0.893
Sweden 1051 1051 4.181 3.230 3.687 0.951

H4 and H5b are partial tests. H4 uses taxnorm and fraudnorm; rednrmdsc is not in the final .dta. H5b uses the available institutional trust items; politician and party trust are not in the final .dta.

5 Index Composition

Most hypotheses are tested with indices. Each item is cleaned for nonresponse, put on a common 0-1 direction where needed, averaged within respondent, and then standardized for the regression models.

hypothesis index components raw_variables direction note
H1 left_right_ideology Left-right self-placement leftright Lower = left; higher = right Ipsos coding confirmed as 1 Left to 11 Right
H2a objective_resources_index Household income decile and subjective social status country-specific hincome_*; slfstatus Higher = more resources
H2b perceived_capacity_index Financial resilience and household economic evaluation fresil; ecohhld Higher = more perceived capacity
H3a objective_need_index Receipt of public payments and unpaid/inactive labor-market status pubpaypens; pubpaychild; pubpayunemp; pubpaydisab; pubpayrent; pubpaymininc; EmploymentStatusKP Higher = more objective need/risk
H3b perceived_risk_index Risk of not coping, unemployment, inability to work, and family-care need prskcope; prskunemp; prskunwork; prskcare Higher = more perceived risk
H4 norms_legitimacy_index Citizen tax morale and welfare-fraud norms taxnorm reversed; fraudnorm reversed Higher = stronger perceived fair contribution norms Partial test: rednrmdsc is absent from the Hauptfeld file
H5a social_trust_index General social trust, perceived fairness, and perceived helpfulness soctrst reversed; percfair; perchelp Higher = more generalized social trust
H5b institutional_trust_index Trust in parliament, judiciary, local authorities, government, European Commission, and satisfaction with democracy trstparl; trstjudi; trstmuni; trstgov; trsteucom; satdem Higher = more institutional trust Partial test: trstpltcns and trstparts are absent from the Hauptfeld file
H5c service_quality_index Overall service satisfaction, service quality, local services, fairness, efficiency, helpfulness, and integrity pssatgen; psqhlth; psqschl; psqtxath; psqtrnsp; psqbnft; psqloc; psefair; pseeffi; psehelp; pseintgrt Higher = better service evaluations
H6a regional_service_quality_within_country_z Regional public-service quality within country EQI qualityp Higher = better regional service quality than the country average Context check using the 2021 full-NUTS2 EQI qualityp file
H6a regional_service_quality_country_mean_z Country mean regional public-service quality EQI qualityp Higher = higher national/regional average service quality Between-country component from the 2021 full-NUTS2 EQI qualityp file
H6b relative_regional_income_z Regional GDP per capita centered within country Eurostat regional GDP per capita Higher = richer than the national regional average Context check; preregistered expected sign is negative

6 Imputation Summary

This section reports how missing values were handled before constructing the main indices. Income is the only variable with substantial missingness. Because income is later used to construct equivalized income and the objective-resources index, missing income is imputed before calculating these measures.

Income is imputed separately within each country. Where possible, missing income is predicted using financial resilience, household economic assessment, age, gender, education, employment status, household size, and household composition. If a prediction cannot be produced, the country mean is used as a fallback. Outcome variables are never imputed.

Overall, 902 income values are imputed, corresponding to 11.2% of the sample. Most other variables have very low missingness, below 5%, and are imputed using simple country-level mean or mode imputation. Respondents missing the dependent variable are excluded from the relevant models.

country Observed income Regression imputed Country-mean fallback Total imputed
Croatia 88.9 0.0 11.1 11.1
France 90.9 4.0 5.1 9.1
Germany 88.6 4.7 6.8 11.4
Italy 85.9 5.0 9.0 14.1
Poland 85.8 5.6 8.7 14.2
Spain 88.8 4.6 6.6 11.2
Sweden 93.0 2.9 4.2 7.0
Variable Rule Imputed %
Income Country regression + fallback 11.2
Household size Country mean 4.4
Household economy Country mean 2.9
Gender Country mean 0.5
Migrant background Country mean 0.5
Employment status Country mode 0.3
Education Country mode 0.2

The main imputation concern is income. Other variables have only minor missingness, so their imputation is unlikely to drive the substantive results. As a robustness check, the main models can also be shown using complete cases only.

7 Preregistered Hypotheses

  • H1 Ideology. More left-wing individuals show higher political solidarity than more right-wing individuals.
  • H2a Objective resources. Individuals with higher objective socioeconomic resources show higher political solidarity.
  • H2b Perceived capacity. Net of objective resources, individuals with higher perceived capacity to contribute show higher political solidarity.
  • H3a Objective need/risk. Individuals with higher exposure to social risks show higher political solidarity.
  • H3b Perceived risk. Net of objective indicators, individuals with higher perceived personal risk show higher political solidarity.
  • H4 Norms and perceived contribution. Individuals who believe that others contribute fairly show higher political solidarity.
  • H5a Generalized social trust. Individuals with higher generalized social trust show higher political solidarity.
  • H5b Institutional trust. Individuals with higher institutional trust show higher political solidarity.
  • H5c Public-service evaluations. Individuals with better public-service evaluations show higher political solidarity.
  • H6a Regional public-service quality. Residents of regions with higher public-service quality show higher political solidarity.
  • H6b Relative regional prosperity. Residents of relatively better-off regions show lower territorial solidarity.

8 Testing Strategy

The primary tests are unweighted OLS models with country fixed effects and the preregistered respondent controls. Block A covers left-right ideology, resources, capacity, need, perceived risk, and the risk-by-capacity interaction. Block B covers norms, social trust, institutional trust, and public-service evaluations. The full model combines both blocks.

I treat a hypothesis as confirmed if the unweighted full-model estimate points in the preregistered direction and its 95% confidence interval excludes zero. Directional but imprecise estimates are mixed. Wrong-direction or near-zero estimates are not confirmed. H6a and H6b are tested separately as preregistered regional-context hypotheses.

9 Descriptive Checks

The plot sorts respondents into five roughly equal-sized bins for each focal individual-level predictor, then plots the mean predictor value against the mean outcome in each bin.

10 Main Models

Characteristic
Block A
Block B
Full
Beta 95% CI p-value Beta 95% CI p-value Beta 95% CI p-value
H1 Left-right ideology -0.329 -0.350, -0.308 <0.001


-0.266 -0.286, -0.246 <0.001
H2a Objective resources 0.063 0.031, 0.094 <0.001


0.005 -0.025, 0.034 0.764
H2b Perceived capacity 0.108 0.083, 0.133 <0.001


0.036 0.012, 0.059 0.003
H3a Objective need/risk 0.054 0.030, 0.077 <0.001


0.029 0.008, 0.051 0.008
H3b Perceived risk -0.029 -0.052, -0.006 0.012


0.008 -0.013, 0.029 0.455
H3 Risk x Capacity interaction 0.001 -0.020, 0.022 0.930


-0.004 -0.023, 0.015 0.681
H4 Norms and legitimacy


0.083 0.063, 0.104 <0.001 0.073 0.052, 0.094 <0.001
H5a Social trust


0.192 0.169, 0.215 <0.001 0.158 0.135, 0.181 <0.001
H5b Institutional trust


0.183 0.155, 0.211 <0.001 0.158 0.129, 0.187 <0.001
H5c Service evaluations


0.150 0.122, 0.178 <0.001 0.135 0.107, 0.163 <0.001
Abbreviation: CI = Confidence Interval
model n r_squared adj_r_squared
Block A 7159 0.222 0.220
Block B 7868 0.268 0.266
Full 7058 0.340 0.338

11 H6 Regional Public-Service Quality Models

These models examine whether regional public-service quality is associated with political solidarity. The H6 predictor is EQI qualityp from the 2021 full-NUTS2 file.

To separate regional from country-level variation, I split EQI into two components. The first is a within-country component: each region’s EQI score minus the average EQI score of its country. This coefficient asks whether respondents living in higher-quality regions are more solidaristic than respondents living in lower-quality regions within the same country. The second is the country mean: the average EQI score across regions in each country. This coefficient asks whether countries with generally higher regional service quality also show higher average solidarity.

This specification is descriptive. It should not be interpreted as a causal mediation test. The models use unweighted OLS with NUTS2-clustered standard errors. Country fixed effects are not included, because they would absorb the country-mean component. The within-country coefficient is therefore closest in spirit to a country fixed-effects comparison, while the country-mean coefficient captures between-country differences.

eqi_source_year matched_respondents matched_nuts2_regions countries_covered note
2021 7319 124 Croatia, France, Germany, Italy, Poland, Spain, Sweden 2021 full-NUTS2 EQI qualityp is used to avoid Germany being aggregated to NUTS1.
Model N NUTS2 Within-country EQI estimate [95% CI] Within-country EQI p Country-mean EQI estimate [95% CI] Country-mean EQI p
Baseline 7319 124 -0.026 [-0.065, 0.012] 0.183 0.084 [0.03, 0.139] 0.003
+ demographics 7319 124 -0.022 [-0.06, 0.017] 0.272 0.075 [0.025, 0.126] 0.003
+ left-right ideology 6816 123 0.005 [-0.029, 0.039] 0.771 0.08 [0.028, 0.131] 0.003
+ Block A non-ideology 6510 123 0.005 [-0.029, 0.039] 0.779 0.062 [0.014, 0.109] 0.011
+ trust/service 6504 123 -0.01 [-0.045, 0.026] 0.601 -0.032 [-0.073, 0.009] 0.131
Full model 6420 123 -0.01 [-0.046, 0.025] 0.571 -0.031 [-0.072, 0.009] 0.128

Across specifications, the within-country EQI coefficient is small and statistically insignificant. The country-mean coefficient is positive in the simpler models, but becomes negative and statistically insignificant once trust and service-evaluation variables are included. This suggests that H6 is not supported by robust within-country regional variation; the initial country-level association appears to be absorbed by attitudinal evaluations of institutions and services.

12 Hypothesis Assessment

hypothesis hypothesis_text how_tested verdict evidence
H1 More left-leaning individuals show higher political solidarity than more right-leaning individuals. Unweighted full model using standardized left-right self-placement, where higher values mean more right-wing. Confirmed beta = -0.266, 95% CI [-0.286, -0.246], p = <0.001
H2a Individuals with higher objective socioeconomic resources show higher political solidarity. Unweighted full model using income and subjective social status. Mixed beta = 0.005, 95% CI [-0.025, 0.034], p = 0.764
H2b Individuals who perceive higher personal capacity to contribute show higher political solidarity. Unweighted full model using financial resilience and household economic evaluations. Confirmed beta = 0.036, 95% CI [0.012, 0.059], p = 0.003
H3a Individuals with higher exposure to social risks show higher political solidarity. Unweighted full model using public-payment receipt and labor-market inactivity. Confirmed beta = 0.029, 95% CI [0.008, 0.051], p = 0.008
H3b Individuals who perceive higher personal risk and future need show higher political solidarity. Unweighted full model using perceived risk plus the preregistered risk x capacity interaction. Mixed beta = 0.008, 95% CI [-0.013, 0.029], p = 0.455; interaction beta = -0.004, p = 0.681
H4 Individuals who believe that others contribute fairly show higher political solidarity. Partial unweighted test using taxnorm and fraudnorm; rednrmdsc is unavailable in the final data. Confirmed beta = 0.073, 95% CI [0.052, 0.094], p = <0.001
H5a Individuals with higher generalized social trust show higher political solidarity. Unweighted full model using social trust, perceived fairness, and perceived helpfulness. Confirmed beta = 0.158, 95% CI [0.135, 0.181], p = <0.001
H5b Individuals with higher institutional trust show higher political solidarity. Partial unweighted test using available institutional trust items; politicians and parties are unavailable. Confirmed beta = 0.158, 95% CI [0.129, 0.187], p = <0.001
H5c Individuals with more positive public-service evaluations show higher political solidarity. Unweighted full model using public-service satisfaction, quality, and procedural evaluations. Confirmed beta = 0.135, 95% CI [0.107, 0.163], p = <0.001
H6a Residents of regions with higher public-service quality show higher political solidarity. Unweighted H6 within/between model using 2021 full-NUTS2 EQI qualityp; verdict is based on the within-country coefficient. Not confirmed beta = -0.01, 95% CI [-0.046, 0.025], p = 0.571
H6b Residents of relatively better-off regions show lower territorial solidarity. Unweighted full + context model using within-country centered regional GDP. Not confirmed beta = 0.014, 95% CI [-0.01, 0.038], p = 0.257

The table above reports the unweighted full-model result for each hypothesis and flags the two partial tests directly in the how_tested column.

13 Secondary Outcomes

In the prereg we also promise secondary analyses of group-specific solidarity. Thus I repeat the unweighted full model using three alternative outcomes:

  • solidarity with children, older people and poor people
  • solidarity with EU and non-EU migrants
  • the gap between those two indices
Characteristic
Deserving solidarity
Migrant solidarity
Welfare chauvinism
Beta 95% CI p-value Beta 95% CI p-value Beta 95% CI p-value
H1 Left-right ideology -0.169 -0.188, -0.150 <0.001 -0.446 -0.470, -0.422 <0.001 0.276 0.254, 0.299 <0.001
H2a Objective resources 0.031 0.004, 0.059 0.026 -0.017 -0.052, 0.018 0.345 0.048 0.014, 0.082 0.005
H2b Perceived capacity 0.003 -0.019, 0.025 0.816 0.024 -0.003, 0.052 0.086 -0.022 -0.049, 0.006 0.118
H3a Objective need/risk 0.007 -0.014, 0.027 0.535 0.061 0.035, 0.087 <0.001 -0.055 -0.079, -0.030 <0.001
H3b Perceived risk 0.006 -0.013, 0.026 0.520 0.045 0.019, 0.070 <0.001 -0.038 -0.062, -0.015 0.002
H4 Norms and legitimacy 0.090 0.069, 0.111 <0.001 0.040 0.017, 0.064 <0.001 0.050 0.026, 0.073 <0.001
H5a Social trust 0.110 0.088, 0.131 <0.001 0.204 0.177, 0.231 <0.001 -0.094 -0.120, -0.068 <0.001
H5b Institutional trust 0.022 -0.005, 0.049 0.108 0.222 0.187, 0.256 <0.001 -0.200 -0.232, -0.168 <0.001
H5c Service evaluations 0.115 0.087, 0.142 <0.001 0.105 0.071, 0.139 <0.001 0.009 -0.023, 0.042 0.571
H3 Risk x Capacity interaction -0.012 -0.030, 0.006 0.203 0.002 -0.021, 0.025 0.850 -0.014 -0.036, 0.008 0.204
Abbreviation: CI = Confidence Interval
term_label direction evidence
H1 Left-right ideology Wider deserving-minus-migrant gap beta = 0.276, 95% CI [0.254, 0.299], p = <0.001
H5b Institutional trust Narrower deserving-minus-migrant gap beta = -0.2, 95% CI [-0.232, -0.168], p = <0.001
H5a Social trust Narrower deserving-minus-migrant gap beta = -0.094, 95% CI [-0.12, -0.068], p = <0.001
H3a Objective need/risk Narrower deserving-minus-migrant gap beta = -0.055, 95% CI [-0.079, -0.03], p = <0.001
H4 Norms and legitimacy Wider deserving-minus-migrant gap beta = 0.05, 95% CI [0.026, 0.073], p = <0.001
H2a Objective resources Wider deserving-minus-migrant gap beta = 0.048, 95% CI [0.014, 0.082], p = 0.005
H3b Perceived risk Narrower deserving-minus-migrant gap beta = -0.038, 95% CI [-0.062, -0.015], p = 0.002

For welfare_chauvinism, positive coefficients mean that the model predicts a larger gap between solidarity with children, older people and poor people on the one hand, and solidarity with EU and non-EU migrants on the other. Negative coefficients mean that the gap is smaller.

Significant positive predictors of the gap are H1 Left-right ideology, H4 Norms and legitimacy, H2a Objective resources. These predictors are associated with relatively more solidarity toward the deserving-target groups than toward migrants.

Significant negative predictors of the gap are H5b Institutional trust, H5a Social trust, H3a Objective need/risk, H3b Perceived risk. These predictors are associated with a smaller deserving-minus-migrant gap.

14 Ideology Sign Checks

The Ipsos left-right self-placement item is coded from 1 = Left to 11 = Right. The main models therefore use left_right_ideology_z, where positive coefficients mean more right-wing respondents score higher on the outcome.

outcome term estimate conf.low conf.high p.value interpretation
Territorial solidarity Left-right ideology (higher = more right-wing) -0.266 -0.286 -0.246 0 More right-wing ideology predicts lower political solidarity.
Welfare chauvinism Left-right ideology (higher = more right-wing) 0.276 0.254 0.299 0 More right-wing ideology predicts higher welfare chauvinism.

Because higher left_right_ideology_z means more right-wing, a negative coefficient in the territorial-solidarity model implies that more left-wing ideology predicts higher political solidarity.

15 Country Heterogeneity

The full model is re-estimated separately by country without country fixed effects. The plot focuses on ideology, perceived capacity, and service evaluations to check whether the main effects are driven by one country.

16 Appendix

16.1 Outcome Distribution

16.2 Missingness

variable meaning n_total n_missing pct_missing
left_right_ideology Left-right self-placement, where lower values are left and higher values are right 8066 576 7.1
perceived_risk_index Perceived risk of future need or deterioration index 8066 346 4.3
norms_legitimacy_index Norms and perceived contribution index; partial Hauptfeld version uses tax morale and welfare fraud norms 8066 181 2.2
perceived_capacity_index Perceived personal capacity to contribute index 8066 94 1.2
institutional_trust_index Institutional trust index; partial Hauptfeld version excludes politicians and parties 8066 36 0.4
objective_need_index Objective exposure to social need/risk index 8066 8 0.1
solidarity_territorial Territorial solidarity index: willingness to pay for redistribution at municipal, country, and EU levels 8066 0 0.0
objective_resources_index Objective socioeconomic resources index 8066 0 0.0
social_trust_index Generalized social trust index 8066 0 0.0
service_quality_index Public-service satisfaction, quality, fairness, efficiency, helpfulness, and integrity index 8066 0 0.0
income_equiv Household income equivalized by square-root household size 8066 0 0.0
age Respondent age 8066 0 0.0
gender_binary Gender indicator: 0 male, 1 female 8066 0 0.0
education_level Highest education level 8066 0 0.0
migrant_background Respondent born outside country of interview 8066 0 0.0
hincome_imputed Indicator for household income deciles imputed by country-specific regression or fallback mean 8066 0 0.0

The unweighted full model retains 7058 of 8066 respondents after listwise deletion (87.5%).