Social Policy and Political Trust: Long-Term Effects of Welfare Generosity in Europe

Political Trust Mood Changes

Citizens’ trust in political institutions, or political trust, reflects the support individuals harbor for their political system and decision-making processes (Easton, 1975). Fundamentally, trust is rooted in the belief that political institutions and actors will adhere to expectations of positive behavior, even in the absence of monitoring by citizens (OECD, 2017). Traditionally, political support measures are categorized as “diffuse” or “specific,” where specific support pertains to day-to-day political management (covering concrete decisions, actors, or events), while diffuse support takes an expansive view (covering political regimes and communities) (Easton, 1975). Political trust, occupying a middle ground between these two ends, remains independent of daily politics but is contingent on ongoing evaluations. Within this framework, political trust encapsulates a fundamental belief that political entities will uphold principles of “good” governance.

National levels of political trust are shaped by citizens’ evaluations of how political institutions have performed, their perceptions of procedural effectiveness, and the ethical conduct of political entities. Therefore, a country may experience higher levels of trust when citizens believe in the competence, reliability, responsiveness, and/or openness of political actors (Fiske, 2002; OECD, 2017). At the aggregate level, despite citizens’ varied perceptions, this trust-as-evaluation approach thus argues that the quality of policies should to some extent matter for a country’s trust “mood,” that is, the country’s average level of trust (Bussi et al., 2022; Kumlin and Haugsgjerd, 2017).

Conceptually, a “mood” reflects a broad and relatively stable inclination of people, in this case, of citizens of a given country, to trust political institutions. Originating from research on public opinion (Stimson, 1991), a policy mood refers to the general disposition or shared feelings of the public toward policies over time and across different circumstances. Policy mood studies emphasize that aggregated attitudes represent broad and shifting sentiments, influenced by short- and long-term political and policy developments, which in turn shape how publics perceive policies and political institutions. For instance, the policy mood can shift to a greater demand for government intervention in response to periods of economic downturn, only to swing back toward preferences for market-driven solutions during times of economic stability. Yet, such shifts occur within existing ideological and institutional boundaries, as long-standing political and economic developments shape the range of viable policy sentiments of the public. Accordingly, a mood refers to the overarching tendency, or unobserved underlying climate, rather than the mere sum of isolated individual opinions at a given moment (Claassen, 2019, 2022, 2024; Stimson, 1991).

Past studies have highlighted how a country’s national level of political trust is influenced by the short- and long-term output of economic policies, such as the unemployment rate and growth of the economy, as well as short- and long-term changes in the quality of political decision-making processes, such as (control over) corruption (Friedrichsen and Zahn, 2014; Goubin and Hooghe, 2020; Ruelens et al., 2018; Van Erkel and Van der Meer, 2016; Van der Meer and Hakhverdian, 2017). Citizens use these factors to assess the performance of their political institutions. Hence, a country’s trust mood should evolve in response to long-term political, economic, and policy developments, while fluctuating due to momentary shocks, events, and shifts in opinion. Though these relationships are well established in the literature, there are limited studies simultaneously estimating the impact of both short- and long-term economic and political performance over an extended period of time across multiple countries. Hence:

In evaluating the performance of political institutions, the output of social welfare policies, aiming to provide crucial social rights and services to citizens, can be a key criterion to evaluate the performance of political institutions (Kumlin, 2004; Kumlin and Haugsgjerd, 2017). Welfare generosity refers to the extent to which the social policies of a given country provide comprehensive and equitable social protection, aiming to safeguard citizens against economic risks, meet their basic needs, and ensure access to opportunities. Such policies include ensuring replacement incomes in times of unemployment or sickness, poverty relief measures, pensions in old age, and promoting childcare to facilitate work (Esping-Andersen, 1990; Nicaise and Lehweß-Litzmann, 2018). While the trust-as-evaluation literature emphasizes the impact of performance evaluations for policy processes and outputs (Van der Meer and Hakhverdian, 2017), previous research has primarily explored the output of economic policies and the procedural fairness of political decision-making processes. Consequently, it remains unclear whether the generosity of social welfare policies in providing social protection, which we conceive of as a “policy output” in the Eastonian sense (Easton, 1965), is used as a performance evaluation that shapes the public’s tendency to trust political institutions.

Studies have shown how governments that implemented austerity measures in social spending usually lost votes in the election following implementation (Armingeon et al., 2016; Horn, 2021). Research indicates that citizens are more likely to support incumbent governments when satisfied with social policies at various levels of government (Giger, 2010), which also positively influences satisfaction with democracy and political trust (Lühiste, 2014). Although these relationships have been documented both between countries and over time within countries (Haugsgjerd, 2018; Kumlin and Haugsgjerd, 2017), the relationship between welfare efforts and political trust has remained relatively unexplored and undertheorized. To address this, we theorize three mechanisms that link a country’s welfare effort to its political trust mood.

Combined Welfare Efforts Foster Political Trust in the Long Run

The first mechanism relates to the long-term enduring impact of sustained welfare efforts. This association between trust and social policies is related to performance of the government to ensure its citizens’ welfare (Bussi et al., 2022; Pierson, 1993). Social security policies aim to guarantee the quality of life for all citizens, as they serve to protect and maintain the quality of life for citizens in times of need. If governments have invested in programs that provide healthcare, education, and social security over an extended period of time, they show that they care about the well-being of citizens and seek to improve their lives. Over time, such efforts can have the ability to foster the feeling among all citizens that their welfare is protected by their government from any risks, be that illness, old age, or unemployment (Kumlin and Haugsgjerd, 2017). In consequence, when governments have invested in a generous welfare state, they create a sense of solidarity between citizens and the state, hence improving the political trust mood.

Shifting perspectives from the benefits of resources for individuals to the nation as a collective provides an explanation as to why economically well-off citizens too would become more trusting because of greater welfare efforts (Esping-Andersen, 1990; Kumlin and Haugsgjerd, 2017; Van Kersbergen, 2006; Wilkinson and Pickett, 2011). Importantly for contemporary times, social policies can further serve to mitigate the effects of economic inequality (Shore, 2019). When governments provide support for low-income households via social welfare policies, particularly in times of need, it can reduce the impact of disparities between the rich and poor (Lühiste, 2014; Wilkinson and Pickett, 2011). Such efforts can be interpreted as welfare efforts to create a level playing field, which can boost trust (Goubin and Hooghe, 2020; Shore, 2019). This level playing field, in turn, facilitates social cohesion and greater horizontal solidarity between citizens within a country (Goubin, 2018). Hence, greater welfare efforts also relate to higher political trust among citizens who do not end up needing these efforts.

Yet importantly, such an impact is not likely to manifest as a result of short-term changes, or per say as a result of a single enduringly generous social policy domain. Rather, fostering the belief that the state ensures citizens’ welfare—that is, fostering the belief in “good” governance—requires sustained effort across policy domains to show that the state will protect citizens whenever the need arises. Hence, it is expected that the sustained and enduring combined welfare efforts of the state is what develops higher political trust moods. This leads to:

Polarized Public Opinion and (Un)Deserving Short-Term Welfare Efforts

Although this long-term perspective of state welfare efforts can foster a positive political trust mood, the short-term challenges of achieving enduringly generous social policies can be more contested (Kumlin and Haugsgjerd, 2017; Kumlin and Stadelmann-Steffen, 2014). Welfare policies have heterogeneous impacts on different population segments, which makes it difficult to discern average effects on political trust across a national population with varying and fluctuating conditions of need. For instance, though some welfare policies are essential for the well-being of socio-economically vulnerable citizens, such as the unemployed, those who are economically well-off might see these same policies as burdensome on the state’s budgetary deficit and causing unnecessarily high taxes (Lim and Moon, 2022; Rudolph and Evans, 2005). Additionally, policymakers often incorporate blame avoidance strategies when enacting austerity policies, and can implement changes incrementally or stealthily (Jensen et al., 2014; Soss and Schramm, 2007). Consequently, within-country changes in welfare efforts are not likely to be noticed by all citizens and likely to be interpreted differently by various citizens from differing socio-economic positions.

Accountability and saliency diffusion are further exacerbated by the normative frames through which citizens view welfare policies. Most citizens agree on the necessity of actions aimed at reducing inequality (Dallinger, 2022). However, within this consensus, there are significant variations due to the contention over who qualifies as “deserving” of welfare benefits. These normative disagreements influence the reception of policies designed to aid contentious groups, such as the unemployed, and less contentious groups, such as the elderly (Laenen and Gugushvili, 2021; Van Oorschot et al., 2017).

Beyond diffused accountability and normative desirability, the fiscal pressures on maintaining a mature welfare state have become a growing cause of concern in the face of aging populations, spikes in unemployment, continued migration, and economic globalization (Kumlin and Nemčok, 2024). Although the discussion of decades past assumed that welfare states would be resilient, the sustainability, affordability, and deservingness of generous social welfare policies is today a central point of discussion. Confronted with reform pressures, this shapes perceptions of deservingness as to which risks the state is and is not responsible for (Goerres et al., 2020; Kumlin and Goerres, 2022; Pierson, 1996, 2001). Concretely, in most European countries, the state has become largely uncontested in its responsibility to protect against life-course risks, such as old age, that are effectively guaranteed for all, but is contested in its responsibility to protect against labor-market risks, such as unemployment insurance. Indeed, research indicates that the welfare reform pressures can be exaggerated depending on its association to (un)deservingness (Jensen, 2012; Laenen and Gugushvili, 2021; Van Oorschot et al., 2017).

Based on these arguments, one might expect that greater welfare efforts to ensure life-course risk insurance are less contentious and more meaningful to all citizens, as all citizens are effectively guaranteed to experience them. This would imply that nationally aggregate levels of political trust should be more sensitive to changes in the generosity of such policies. In contrast, expanding labor market insurance policies is likely to be more contested, and unpopular across most citizens. This clash in desirability is likely to itself cause a decrease in a country’s political trust mood due to increasing the salience of polarization around social policy issues. Hence:

Dependent Variable: Capturing Political Trust Moods

This study builds on a novel statistical method developed by Claassen (2019, 2022) to construct a dataset with a broader spatial-temporal scope compared to most previous studies on nationally aggregate political trust levels. Building upon the tradition of “mood” measures that address the need for high-quality aggregate public opinion data over large time spans (Jackman, 2005; Stimson, 1991), we pooled data on trust and confidence in various political institutions from six different survey projects ¹ spanning from 1981 to 2022 and covering 31 European countries. ² With the proliferation of international surveys, multiple studies have sought to pool data (e.g. Van Erkel and Van der Meer, 2016), but were confronted with fragmentation of national time-series data and discrepancies in political trust item wording and scale both across and within surveys. This method is innovative as it overcomes such data restrictions by leveraging Bayesian latent variable modeling techniques that smooth over gaps in time series while accounting for item bias and time-invariant cross-national non-equivalence as well as random error (Claassen, 2019, 2022), with recent findings indicating that it can be applied to political trust measures too (Claassen, 2024; Kołczynska et al., 2024; Valgarðsson et al, 2024).

To illustrate the exact operationalization of the trust mood measure, it is helpful to discuss concrete survey items. Sporadically since 1981, the European Values Study included the following item: “Please look at this card and tell me, for each item listed [National Parliament and Civil Services], how much confidence you have in them, is it a great deal (4), quite a lot (3), not very much (2) or none at all (1)?” In contrast, items included in the European Social Survey on a bi-annual basis since 2002 ask, “Using this card, please tell me on a score of 0-10 how much you personally trust each of the institutions I read out [politicians, political parties, and the national parliament]. 0 means you do not trust an institution at all, and 10 means you have complete trust.” Despite both survey questions asking explicitly about trust and confidence in political institutions, the two items differ in wording, response scales, and institutional focus, while also having differing country samples and time frames, each of which introduces some type of bias in measuring political trust. For the purposes of this study, we excluded items that attempt to capture political trust implicitly to limit bias introduced by question wording, meaning item batteries needed to explicitly ask about trust or confidence. Moreover, to sufficiently account for bias introduced by survey projects, items from surveys with less than three waves were excluded. Last, we only included survey items fielded in multiple European countries across multiple waves, in turn reducing bias introduced by country-specific question interpretations. Table A.1 in Appendix A shows the list of items that were used for the computations of political trust mood scores.

In the simplest terms, to account for these item variations, responses are first recoded into affirmative (1) and dissenting (0) responses to ensure comparability (more information about the statistical principles and equations applied can be found in Appendix A). Biases introduced by survey-item differences are accounted for by using a survey-item loading score in the measurement model of political trust, ensuring variation due to wording or format does not distort the estimates. As items can function differently across countries, a country-item parameter is used to ensure that variation is not driven by country variations in the interpretation of questions. To smooth over time gaps, estimates for each year are informed by their prior level, producing consistent estimates of the political trust mood across Europe over time. This approach yields scores for country-years that reflect fluctuations in the overarching tendency of publics to trust their political institutions. As is common in latent measurement models, these scores are standardized around a grand mean of zero with a standard deviation of one, allowing for direct comparisons across time and countries.

Scores are estimated using Bayesian Markov-Chain Monte Carlo methods via the R package CmdStanModel (Carpenter et al., 2017; Češnovar et al., 2021). Four parallel chains were run with 1000 samples, where the first 500 samples are used as warm-up and the remaining samples analyzed for the posterior distribution. Models converged with R ^ s between 0.95 and 1.05. This resulted in 1066 country-years as based on 3335 nationally aggregated public opinion measures on political trust. This allows for one of the most comprehensive analyses of trends in trust in Europe to date (Claassen, 2024; Kołczynska et al., 2024; Norris, 2011, 2022; Valgarðsson et al, 2024).

The remainder of this article will focus on the latent country-year estimates of political trust, also known as political trust mood scores. To give insight into what the scores imply, Figure 1 visualizes all the country-years for which the method was applied. The political trust “mood” scores reflect the extent to which a given country-year is trusting of their political institutions relative to the grand mean fixed at 0. Accordingly, scores reflect specifically when and which European countries had more or less political trust during the last four decades relative to each other as opposed to being based on an item scale or percentage of affirmative responses. From these average trends, and as can be expected from past research (Haugsgjerd, 2018), we can see that European countries experienced a particularly turbulent period following the 2008 recession. While the average level of trust in 2020 was slightly above the grand mean (x̄ = 0.12), the average across Europe from 2009 to 2017 hovered around −0.20. Beyond the average trend across countries, Figure 1 orders countries with higher scores in 2020 toward the top and lower scores to the bottom of the Figure.

OPEN IN VIEWER

Figure 1.

Political Trust Mood Scores Across Europe over the Last Four Decades.

As this ordering clearly shows, and as can be expected from past findings (Norris, 2011, 2022; Zmerli and Van der Meer, 2017), there is group of high trust countries with positive scores, mainly in northern Europe (e.g. Switzerland, Sweden, and Germany), and low-trust countries with negative scores, mainly in southern and eastern Europe (e.g. Poland, Slovakia, and Greece). This highlights how there are both enduring differences between countries and yearly fluctuations that overlap with established causes of political trust, hence providing quality data on how much a country was trusting of their political institutions.

Established Causes of Trust: Economic and Political Performance

To examine whether the descriptive trends are indeed significantly related to established causes of differences in national-level political trust, we include key economic and political performance factors (Friedrichsen and Zahn, 2014; Van Erkel and Van der Meer, 2016). To test hypothesis 1a for economic performance, we include the unemployment rate and GDP per capita. Data for this is obtained from the World Development Indicators (Armingeon et al., 2018). To test hypothesis 1b for political performance, we include control of corruption as measured in the World Governance Indicators (WGI) by the World Bank (2018). This index reflects aggregate perceptions of the extent to which there is control over the use of public power for private gains and quality of policy formulation.

Welfare State Generosity

To examine how welfare efforts are associated with a country’s political trust mood, we use data from the Comparative Welfare Entitlement Project (Scruggs, 2023). The CWEP indices focus on citizens’ rights to social insurance, considering aspects beyond expenditure, such as how the money is spent, and policies are structured. The indices account for qualification and waiting periods, expected duration of benefits, standard replacement rates, contribution proportions, and coverage rates. Scores are determined by summing these characteristics and classifying countries as less generous, averagely generous, or more generous, based on their deviation from the mean score in the CWEP dataset. These classifications are combined so that higher values indicate the risk insurance policies are more generous. At the time of producing this study, the CWEP is yet to incorporate data on Eastern European countries, leading to a reduction of the total country-year sample (Scruggs and Ramalho Tafoya, 2022).

The CWEP contains data on three public insurance policy programs that take up large portions of in-cash benefit transfers of social policies across countries: Pension, Unemployment, and Sickpay. Additionally, the CWEP provides a combined generosity score that is based on the sum of scores for the three respective programs. For hypothesis 2, we use the combined generosity index to test whether enduringly high combined welfare efforts foster positive political trust moods. For hypotheses 3a and 3b, we use the pension ³ and unemployment ⁴ generosity indices to test whether short-term changes with differing degrees of contentiousness have diverging associations to political trust moods.

Given that the level of social expenditure can influence the interpretation of changes in welfare efforts, we control for the total social expenditure as a percentage of a country’s GDP (Scruggs and Ramalho Tafoya, 2022). To capture differences in social policy expenditure, this study uses data from the OECD (2019). Social expenditure refers to the public and mandatory, as well as voluntary private, expenditure within government social policies. Importantly, expenditure itself can be confounded by other factors that do not reflect the extent to which citizens are ensured welfare and protected against risks. For example, expenditure can be higher due to the increased “objective” demand for specific policies (e.g. a growing population of elderly and higher unemployment rates) or not be reflective of generosity given the manner in which policies are structured (e.g. high benefit duration for low replacement rates). Nonetheless, as citizens can relate changes to cash benefit welfare policies with state budgetary deficits, we control for the total level of social policy expenditure to isolate the impact of welfare efforts, and with expenditure for the elderly and unemployment specifically.

Modeling and Sampling Strategy: Enduring and Transient Efforts Across Europe

To disentangle the impact of short- and long-term welfare efforts, this study leverages hierarchical random-effect within-between (REWB) models, wherein observations at year t are nested within countries i. Specifically, for all variables, an across time mean variable (x̄_i) is created to gauge long-term differences between countries. Furthermore, this across time mean is subtracted from the original value (x_ti − x̄_i), capturing the year’s difference to the country’s mean over time, and hence reflecting short-term changes occurring within countries. To account for spurious effects, a time trend control β₀(t) is added to the model. Moreover, to isolate the short- and long-term estimates for welfare efforts, controls (v_i) are added with similar within-between transformations (for a detailed discussion on the advantages of REWB models, see Bell et al., 2018; Fairbrother, 2014). The advantage of this method is that we can disentangle the impact of persistent disparities between countries on political trust from unfolding internal dynamics within countries. Moreover, past studies showcase that utilizing REWB models is fruitful in understanding how macro-level dynamics over time and across countries impact political trust (e.g. de Blok et al., 2022; Goubin and Hooghe, 2020).

The equation is as follows:

y it = β 0 ( t ) + β 1 ( x ti − x ¯ i ) + β 2 ( x ¯ i ) + v i + ε it

Because of data availability across the various data sources, there is a notable reduction in the total country sample when including the CWEP indices, and a further reduction in the total year sample when including the WGI indices. This results in a total balanced sample of 15 countries (Austria, Belgium, Finland, France, Germany, Greece, Ireland, Italy, the Netherlands, Norway, Portugal, Spain, Sweden, Switzerland, and the United Kingdom) across two decades (1998, 2000, and yearly since 2002 until 2018 before the onset of Covid-19). Though smaller than desired, this 285 country-year sample does provide a variety of changing economic conditions, political performance, and welfare regimes, while remaining largely within a comparable geographic-historical scope of (Western) European countries in the early twenty-first century. Nonetheless, given lower power, we keep our models parsimonious. An overview of the descriptive statistics for the original factor scales, as well as their within-between transformation for this sample of country-years is provided in Table 1. In section “Robustness Analysis,” as part of the robustness analyses, we first run the main analysis models using robust standard errors. Additionally, we run models that exclude political performance, allowing tests with a wider time span (annual observations from 1991 to 2018). ⁵ All robustness analyses are found in Appendix B.

OPEN IN VIEWER

Table 1.

Descriptive Statistics of Within-Between Transformations.

Variables (and within-between transformation)	Min	Max	Mean	SD
Economic & Political Performance
Unemployment Rate (% Working Population)	2.49	27.69	8.21	4.64
. . . between	3.65	16.40	8.21	3.59
. . . within	-8.17	12.01	0.00	2.95
GDP per Capita (divided by 1000)	11.53	103.55	42.13	18.02
. . . between	19.71	75.07	42.13	14.79
. . . within	-40.27	28.48	0.00	10.29
Control Corruption (multiply by 10)	-1.90	24.54	15.20	6.49
. . . between	1.46	22.69	15.20	6.31
. . . within	-5.39	6.09	0.00	1.54
Welfare Efforts
Combined Generosity	21.78	45.46	34.71	5.22
. . . between	24.69	44.31	34.71	5.07
. . . within	-3.23	5.44	0.00	1.23
Pension Generosity	8.79	15.80	12.74	1.58
. . . between	8.96	14.54	12.74	1.48
. . . within	-1.65	2.62	0.00	0.56
Unemployment Generosity	3.75	15.21	10.50	2.84
. . . between	5.10	14.57	10.50	2.76
. . . within	-2.11	2.49	0.00	0.65
Total Social Policy Expenditure (% GDP)	13.14	32.65	24.63	3.79
. . . between	17.70	30.23	24.63	3.11
. . . within	-5.56	6.43	0.00	2.17
Old Age Expenditure (% GDP)	2.39	14.79	8.84	2.72
. . . between	3.58	13.22	8.84	2.46
. . . within	-3.61	3.24	0.00	1.17
Unemployment Expenditure (% GDP)	0.13	3.58	1.24	0.78
. . . between	0.26	3.00	1.24	0.70
. . . within	-1.29	1.35	0.00	0.36

Main Results: Long- and Short-Term Correlates of Political Trust Moods

Table 2 shows the results of the REWB models. Starting with the null model (0), we find an intraclass correlation (ICC) of 70%, indicating that much variation in political trust mood scores in (western) Europe during the last two decades is explained by enduring between country differences as opposed to transient fluctuations. This aligns with what we see in Figure 1, where the fluctuations in how much citizens in a country trust their political institutions largely hover slightly above, slight below, or on the average level of trust, while the ultimate country rankings do not change substantially.

OPEN IN VIEWER

Table 2.

REWB Models on the Short- and Long-term Impact of Welfare Performance on Political Trust Moods.

	(0)	(1)	(2)	(3)	(4)	(5)	(6)	(7)
	std. Beta	std. Beta	std. Beta	std. Beta	std. Beta	std. Beta	std. Beta	std. Beta
Economic Performance
Unemployment Rate (btw)		0.060(0.116)			0.058(0.097)	0.039(0.109)	0.064(0.098)	0.137(0.113)
Unemployment Rate (wi)		–0.349*** (0.021)			–0.372*** (0.021)	–0.363*** (0.022)	–0.325*** (0.026)	–0.296*** (0.032)
GDP per Capita (btw)		0.472*** (0.112)			0.418*** (0.097)	0.427*** (0.101)	0.427*** (0.100)	0.426*** (0.092)
GDP per Capita (wi)		0.091** (0.028)			0.089** (0.027)	0.097*** (0.028)	0.074** (0.027)	0.089** (0.028)
Political Performance
Control of Corruption (btw)		0.448*** (0.110)			0.398*** (0.094)	0.445*** (0.100)	0.399*** (0.094)	0.540*** (0.097)
Control of Corruption (wi)		0.160*** (0.023)			0.159*** (0.022)	0.163*** (0.022)	0.154*** (0.022)	0.159*** (0.022)
Overall Welfare Performance
Combined Generosity (btw)			0.555*** (0.165)		0.175* (0.070)		0.168* (0.073)
Combined Generosity (wi)			0.046		–0.073***		–0.044*
			(0.032)		(0.019)		(0.022)
Specific Welfare Performance
Pension Generosity (btw)				–0.154(0.179)		0.088(0.075)		0.071(0.065)
Pension Generosity (wi)				–0.103** (0.038)		–0.056* (0.023)		–0.028(0.025)
Unemployment Generosity (btw)				0.456* (0.179)		0.129(0.072)		0.186* (0.081)
Unemployment Generosity (wi)				0.088** (0.033)		–0.057** (0.021)		–0.037(0.021)
Welfare Spending Controls
Total Social Expenditure (btw)							0.020(0.063)
Total Social Expenditure (wi)							–0.093** (0.032)
Old Age Expenditure (btw)								0.128* (0.063)
Old Age Expenditure (wi)								–0.047(0.040)
Unemployment Expenditure (btw)								–0.072(0.081)
Unemployment Expenditure (wi)								–0.071** (0.025)
Random Effect
σ2	0.24	0.08	0.24	0.23	0.08	0.08	0.07	0.07
τ00	0.57	0.06	0.32	0.35	0.04	0.04	0.04	0.03
ICC	0.70	0.41	0.57	0.61	0.33	0.36	0.34	0.29
N	15 Countries
Observations	285
AIC	476.845	212.779	481.821	481.289	212.712	218.792	221.382	232.228

Within-country effects were abbreviated to wi and between-country effects to btw. Estimates shown are Standardized Beta Coefficients. *p < 0.05, **p < 0.01, ***p < 0.001.

To see whether the political trust mood scores are related to established causes in the literature on national-level political trust, we start by discussing the short- and long-term impact of economic and political performance factors. After including the within-between effects for economic and political performance (model 1), the ICC already drops to 41%, suggesting around 30% of between country differences in political trust moods relate to economic and political performance. Here, we hypothesized that positive economic conditions, measured by the unemployment rate and GDP per capita, and political conditions, measured by the control over corruption, are related to higher political trust moods. As Table 2 shows, there is indeed a positive impact of greater short- and long-term control over corruption. However, we find that only within-country increases in the unemployment rate are related to lower political trust moods, while there is no significant estimate for between-country differences. ⁶ We do find that both higher within- and between-country GDP per Capita levels are positively related to political trust moods. Hence, these results provide support for both hypothesis 1a and 1b on the impact of economic and political performance.

Next, we turn to hypothesis 2 on the enduring impact of combined welfare efforts. For this, we hypothesized that the between-country differences in welfare generosity relate to higher political trust moods, while we did not expect this for short-term within-country changes. Indeed, as we see in Table 2, across model iterations, there is a significant positive between-country estimate for combined welfare generosity, while the within-country estimate is insignificant or negative. Although not to the same extent as economic and political performance, combined welfare efforts do explain around 23% of the ICC. The contrast in estimate direction with the amount of variance explained underscores how enduringly strong welfare efforts can facilitate more positive perceptions of the government, but that achieving this incrementally can be a contested journey. Accordingly, this finding provides support for the theoretical mechanism underlying hypothesis 2 on the long-term combined welfare efforts.

Interestingly, when examining specific welfare policies—the generosity of old age and unemployment insurance—there is no clear direct impact of short- and long-term differences on political trust. We anticipated that increased generosity in old age spending would positively correlate with political trust, while more generous unemployment benefits would be associated with a decline in trust. However, as shown in models 3, 5, and 7 of Table 2, the results for pension and unemployment benefit generosity are largely inconsistent. Pension generosity yields an insignificant between-country estimate, while the within-country estimate is negative across all three models testing hypothesis 3a. Moreover, much less variance in political trust moods between countries is explained by these specific welfare efforts, with an ICC drop of less than 10%. We speculate that this may be due to heightened discussions on reform pressures stemming from aging populations and increasing polarization around pension spending. Consequently, we reject hypothesis 3a (Short-Term Uncontested Welfare Performance). Similarly, the coefficients for unemployment benefit generosity vary in both sign and significance, offering no clear evidence of an effect from long- or short-term changes. Therefore, we also reject hypothesis 3b (Short-Term Contested Welfare Performance).

Robustness Analysis

The results of the robustness analyses are presented in Appendix B. To ensure the robustness of these analyses, models were rerun using robust standard errors (Table B.2). First, there are no substantial changes in the economic and political performance indicators. ⁷ This result is in line with the literature, and they explain a substantial amount of variation in political trust mood scores, as indicated by the drop in the ICC. Second, for combined welfare generosity, although there is a decrease in the significance level, the positive between-country effect remains significant at a p < 0.10 level. Given the statistical power and the number of between-country effects in these models that substantially explain national-level political trust, this result indicates tentative robustness. Third, when using robust standard errors for the models containing pension and unemployment benefit generosity, the positive effect of between-country unemployment generosity in models 3 and 7 stays significant.

To increase the statistical power of the models, we ran additional analyses that excluded the control of corruption indicator (Table B.3), as data for this indicator is largely limited to the twenty-first century, allowing an additional decade of observations to be included in the models. ⁸ Starting with economic performance, the within and between effects for the GDP per capita, and the within effect of the unemployment rate, remain significant, while the between effect of the unemployment rate remains insignificant. Turning to the estimates for the combined generosity index, countries with an enduringly more generous package of welfare policies do have significantly higher political trust moods, while within-country changes do not have a significant relation. This also holds true after applying robust standard errors (Table B.4). For specific welfare efforts, all effects become insignificant. This result aligns with our discussion on the importance of combined welfare efforts, as opposed to certain welfare domains. Nonetheless, the between-country effect for unemployment generosity does remain significant at a p < 0.10 level when accounting for economic performance, in contrast to our discussion on deservingness, though the estimates become insignificant once controlling for expenditure. Accordingly, hypothesis 1 (impact of economic and political performance) and hypothesis 2 (long-term impact of combined welfare efforts) are accepted, while hypotheses 3a and 3b are rejected.