| hypothesis | status | proxy |
|---|---|---|
| H1 Regional governance quality | Tested with EQI 2024 | EQI 2024 regional governance quality |
| H2 Perceived governance quality | Tested now | perceived_governance_index, service_quality_index, institutional_trust_index |
| H3 Attenuation after adding perceptions | Tested with EQI 2024 + perceptions | Compare objective-context models with and without perceptions |
| H4 Ideology moderation | Tested now | perceived_governance_index × ideology_right |
| H5 Regional governance change | Not testable from available files | Requires at least two EQI waves |
| Additional prereg step: welfare-state moderation | Not testable from available files | Requires welfare-state-size context |
PEPS-Y Core 2: Governance and Redistribution
1 Setup
2 Preregistered Hypotheses
H1 Regional governance quality.“The regional governance quality is positively associated with higher individual willingness to contribute to public redistribution.”H2 Perceived governance quality.“The individual perception of governance quality is positively associated with higher individual willingness to contribute to public redistribution.”H3 Attenuation after adding perceptions.“The association between regional governance quality and willingness to contribute to public redistribution is attenuated when perceived government quality is included (consistent with mediation).”H4 Ideology moderation.“The association between perceived governance quality and willingness to contribute to redistribution is moderated by ideology: it is negative among right-wing citizens and positive among left-wing citizens.”H5 Regional governance change.“Improvements in regional governance quality (ΔEQI = EQIt2 - EQIt1 at NUTS2) are positively associated with willingness to contribute to redistribution.”
The preregistration also includes an additional context-level step on welfare-state moderation. That check is retained in this report as an optional context model, but it cannot be evaluated from the pilot file alone.
The automatic GISCO-based region match linked contextual governance information to 355 interviews.
3 Missingness Diagnostic
The preregistration specifies listwise deletion as the primary approach when missingness is limited, with multiple imputation as a fallback. The table below reports per-variable missing rates for all key analysis variables.
| variable | n_total | n_missing | pct_missing |
|---|---|---|---|
| education_level | 355 | 52 | 14.6 |
| income_equiv | 355 | 35 | 9.9 |
| ideology_right | 355 | 7 | 2.0 |
| gender_binary | 355 | 1 | 0.3 |
| migrant_background | 355 | 1 | 0.3 |
| solidarity_territorial | 355 | 0 | 0.0 |
| perceived_governance_index | 355 | 0 | 0.0 |
| service_quality_index | 355 | 0 | 0.0 |
| institutional_trust_index | 355 | 0 | 0.0 |
| age | 355 | 0 | 0.0 |
After listwise deletion the H2 model retains 273 and the H4 model retains 270 of 355 observations. If overall loss exceeds ~20%, multiple imputation should be considered per the preregistration.
4 Visual Checks
5 OLS Models for H2 and H4
This section estimates three separate OLS models around the governance hypotheses. The H2 composite model asks whether the overall perceived-governance index is associated with solidarity. The H2 components model then separates that composite into service-quality evaluations and institutional trust, which lets us see which part of the broader governance perception measure is actually carrying the association. The H4 moderation model adds ideology and the governance-by-ideology interaction to test whether the governance association differs across the left-right spectrum.
Because the substantive predictors are standardized, the coefficients can be read as the expected change in the territorial solidarity index associated with a one-standard-deviation increase in each predictor, holding the controls constant. In the compact comparison table below, coefficient estimates include significance stars, and rows with p < 0.05 are bolded to make the clearest pilot signals easier to identify.
| model | term_label | estimate | conf_int | p_value |
|---|---|---|---|---|
| H2 composite | Perceived governance composite | 0.243*** | [0.117, 0.368] | <0.001 |
| H2 components | Institutional trust component | 0.245** | [0.097, 0.393] | 0.001 |
| H2 components | Service quality component | 0.029 | [-0.107, 0.165] | 0.676 |
| H4 moderation | Perceived governance composite | 0.190*** | [0.084, 0.296] | <0.001 |
| H4 moderation | Governance × ideology | 0.068 | [-0.019, 0.154] | 0.127 |
| H4 moderation | Right ideology | -0.250*** | [-0.342, -0.158] | <0.001 |
5.1 Traditional Regression Tables
The compact table above is useful for quick cross-model comparison. The tables below present the same results in a more conventional regression-reporting format using gtsummary::tbl_regression(), which is closer to what you might include in an appendix or a methods supplement.
Characteristic |
H2 composite |
H2 components |
H4 moderation |
|||
|---|---|---|---|---|---|---|
Beta 1 |
SE 2 |
Beta 1 |
SE 2 |
Beta 1 |
SE 2 |
|
| Perceived governance composite | 0.243*** | 0.054 | 0.190*** | 0.051 | ||
| Service quality component | 0.029 | 0.062 | ||||
| Institutional trust component | 0.245*** | 0.069 | ||||
| Right ideology | -0.250*** | 0.045 | ||||
| Perceived governance composite * Right ideology | 0.068 | 0.040 | ||||
| 1
p<0.05; p<0.01; p<0.001 |
||||||
| 2
SE = Standard Error |
||||||
5.2 Model Fit Comparison
The preregistration calls for reporting model fit and explained variance. The table below compares adjusted R² across the governance models.
| model | n | r_squared | adj_r_squared |
|---|---|---|---|
| H2 composite | 273 | 0.172 | 0.138 |
| H2 components | 273 | 0.180 | 0.142 |
| H4 moderation | 270 | 0.288 | 0.252 |
6 Hypothesis-by-Hypothesis Assessment
The table below makes the pilot verdict explicit and places it close to the main OLS results, where readers are most likely to ask the practical question: which hypotheses are already supported, which look mixed, and which still cannot be tested without the contextual governance variables.
| hypothesis | hypothesis_text | how_tested | pilot_verdict | evidence |
|---|---|---|---|---|
| H1 | The regional governance quality is positively associated with higher individual willingness to contribute to public redistribution. | OLS model using EQI 2024 as the objective governance measure, with the same respondent controls and country fixed effects as the main micro-level models. | Not confirmed in the pilot | beta = -0.265, 95% CI [-0.506, -0.024], p = 0.032 |
| H2 | The individual perception of governance quality is positively associated with higher individual willingness to contribute to public redistribution. | Tested with the perceived-governance composite and then decomposed into service-quality and institutional-trust components. | Confirmed in the pilot | Composite: beta = 0.243, 95% CI [0.117, 0.368], p = <0.001; service-quality component: beta = 0.029, 95% CI [-0.107, 0.165], p = 0.676; institutional-trust component: beta = 0.245, 95% CI [0.097, 0.393], p = 0.001 |
| H3 | The association between regional governance quality and willingness to contribute to public redistribution is attenuated when perceived government quality is included (consistent with mediation). | Two-step comparison of the objective-governance coefficient before and after adding perceived governance to the model. | Mixed pilot evidence | Objective-only: beta = -0.265, 95% CI [-0.506, -0.024], p = 0.032; with perceptions: beta = -0.225, 95% CI [-0.445, -0.004], p = 0.047 |
| H4 | The association between perceived governance quality and willingness to contribute to redistribution is moderated by ideology: it is negative among right-wing citizens and positive among left-wing citizens. | Tested with the perceived-governance × ideology interaction in the main model and checked again across the secondary solidarity outcomes. | Mixed pilot evidence | Governance main effect: beta = 0.190, 95% CI [0.084, 0.296], p = <0.001; ideology main effect: beta = -0.250, 95% CI [-0.342, -0.158], p = <0.001; interaction: beta = 0.068, 95% CI [-0.019, 0.154], p = 0.127 |
Substantively, the pilot gives the clearest support to H2 rather than H1, H3, or H4. Perceived governance is positively associated with territorial solidarity in the composite model, and that pattern appears to be driven much more by institutional trust than by service-quality evaluations. The ideology-moderation hypothesis is not confirmed in the pilot: right-wing orientation is strongly associated with lower solidarity overall, but the interaction term itself is imprecise and does not show the preregistered pattern that governance should matter differently for left- and right-leaning respondents. The new objective-governance models also run against the preregistered expectation: the EQI coefficient is negative rather than positive, and it remains negative after perceived governance is added, so there is only weak attenuation and no support for the intended mediation story. On the secondary outcomes, the perceived-governance measure remains positive for migrant-targeted and broader group solidarity, while also predicting lower welfare chauvinism, which is directionally coherent with the general argument even though the moderation mechanism is not clearly supported.
7 Interaction Plot for H4
8 Ordered Logit Robustness
This section asks whether the governance results are robust to both item-level disaggregation and a more appropriate model for ordinal outcomes.
The main OLS specifications above treat the territorial solidarity index as the primary outcome. But it also smooths over the fact that the original survey items are ordinal response scales, not continuous measures.
The ordered-logit models below do two things: First, they check whether the main findings still hold when the three territorial items are analyzed one by one rather than averaged together. Second, they use an estimator that is designed for ordered categories, which is a cleaner match to the measurement scale of the original survey responses.
| outcome | term_label | estimate | conf.low | conf.high | p.value |
|---|---|---|---|---|---|
| plsolcntry | Perceived governance composite | 0.402 | 0.112 | 0.697 | 0.007 |
| plsolcntry | Governance × ideology | 0.037 | -0.189 | 0.261 | 0.744 |
| plsolcntry | Right ideology | -0.519 | -0.784 | -0.259 | 0.000 |
| plsoleu | Perceived governance composite | 0.450 | 0.176 | 0.731 | 0.001 |
| plsoleu | Governance × ideology | 0.280 | 0.052 | 0.517 | 0.018 |
| plsoleu | Right ideology | -0.724 | -0.976 | -0.479 | 0.000 |
| plsolmuni | Perceived governance composite | 0.438 | 0.150 | 0.732 | 0.003 |
| plsolmuni | Governance × ideology | 0.026 | -0.199 | 0.249 | 0.820 |
| plsolmuni | Right ideology | -0.642 | -0.912 | -0.379 | 0.000 |
9 Notes
- The preregistration text uses inconsistent numbering between the hypothesis list and the later analysis steps. This report follows the main hypothesis numbering H1-H5 and then treats welfare-state moderation as an additional preregistered analysis step when context data are available.
- External governance indicators from EQI 2024 are now attached automatically, but EQI change and welfare-state-size inputs are still unavailable. The report therefore activates H1 and H3 while keeping H5 and the welfare interaction out of scope for now.
- The preregistration specifies random-effects and/or fixed-effects models to account for respondent clustering within regions and countries. Because the pilot sample is small and does not include the contextual variables that would benefit from a multilevel structure, this report uses OLS with country fixed effects and HC3 robust standard errors. The full-sample analysis should implement mixed models with random intercepts for region and/or country, and should include the between/within decomposition required for the H3 mediation test.
- The H2 “components” model (separating service quality from institutional trust) implements the preregistered secondary analysis of alternative perception operationalisations. If the components show divergent patterns, this suggests the composite may mask distinct channels.
- The results here are best used as pilot checks of directional coherence and measurement behavior.
10 Secondary Outcome Models
The preregistration specifies running the governance models on group-specific solidarity outcomes (children, older people, the poor, EU migrants, non-EU migrants) and related indices. The models below apply the H4 specification (governance × ideology) to these secondary outcomes.
| outcome | term_label | estimate | conf.low | conf.high | p.value |
|---|---|---|---|---|---|
| solidarity_deserving | Perceived governance composite | 0.049 | -0.064 | 0.163 | 0.393 |
| solidarity_deserving | Governance × ideology | 0.035 | -0.087 | 0.158 | 0.574 |
| solidarity_deserving | Right ideology | -0.166 | -0.260 | -0.072 | 0.001 |
| solidarity_group | Perceived governance composite | 0.110 | 0.006 | 0.213 | 0.039 |
| solidarity_group | Governance × ideology | 0.056 | -0.045 | 0.156 | 0.278 |
| solidarity_group | Right ideology | -0.301 | -0.387 | -0.215 | 0.000 |
| solidarity_migrant | Perceived governance composite | 0.200 | 0.063 | 0.338 | 0.005 |
| solidarity_migrant | Governance × ideology | 0.086 | -0.026 | 0.199 | 0.133 |
| solidarity_migrant | Right ideology | -0.503 | -0.612 | -0.394 | 0.000 |
| welfare_chauvinism | Right ideology | 0.337 | 0.231 | 0.443 | 0.000 |
| welfare_chauvinism | Governance × ideology | -0.051 | -0.180 | 0.078 | 0.439 |
| welfare_chauvinism | Perceived governance composite | -0.151 | -0.288 | -0.013 | 0.033 |
