1 Introduction

European societies have seemed to be in constant crisis mode since the turn of the millennium. The financial and the debt crises have been followed by the refugee crisis and the rise of populist movements, now being eclipsed by a pandemic and the consequences of the Russian invasion of Ukraine. During all of these crises, politicians and societal players have never tired of emphasizing the importance of social integration (Pitas und Ehmer 2020). This matches scientific findings showing the positive consequences of social integration as social capital on the contextual and the individual levels (Paxton 2002; Portela et al. 2013).Footnote 1 Unsurprisingly, a longstanding question in this field of research is how social integration can be created (Grunow et al., this issue). Next to individual factors that foster social integration, established findings emphasize the importance of two contextual factors in crafting social integration: civil society and political institutions (Stolle and Rothstein 2007). A crucial institutional aspect is welfare state policies that have been shown to matter for social integration (Kumlin and Rothstein 2005; Rothstein 2001). We revisit the role of welfare state policies for volunteering as a key indicator of social integration creating social capital. In particular, we ask whether the link between welfare state policies and volunteering varies across social groups and policy areas.

Generally, two arguments have been put forward in the literature regarding the role of welfare states for social integration (Gundelach et al. 2010; van Oorschot and Arts 2005). While the “crowding-out” thesis assumes that extensive welfare spending hampers social integration by suppressing private obligations of mutual support, the “crowding-in” thesis argues that a generous welfare state can foster social integration by providing resources and establishing a culture of caring (Kumlin and Rothstein 2005; van Oorschot and Arts 2005; Rothstein 2001). Overall, the crowding-in thesis has received more empirical support, but it has also become clear that reality is more complex than these two conflicting arguments suggest. Research shows that the relationship might be conditional on the concrete welfare state policies and the potential recipients of support (Ferragina 2017; Stadelmann-Steffen 2011; Visser et al. 2018). Again, other studies indicate that welfare state expenditures are differentially related to varying aspects of social integration (Kääriäinen and Lehtonen 2006; van Oorschot and Arts 2005).

We build on these nuanced perspectives on the link between welfare state policies and social integration and qualify existing research in two respects. First, we focus on volunteering as a key aspect of social integration. As Putnam (2000, pp. 116–117) puts it, “[V]olunteering …—our readiness to help others—is by some interpretations a central measure of social capital …. Thus, any assessment of trends in social capital must include an examination of trends in volunteering.” While social networks and regular contacts might be restricted to groups of friends and family, volunteering has the potential to bridge societal groups and build strong and sustainable social ties among members of a society (Dekker and Van den Broek 2005). Thus, we examine how welfare state policies relate to volunteering and, thereby, add to the literature that has mostly focused on social contacts, networks, and group membership. Only a few exceptions have explicitly studied volunteering, but these studies are restricted to either single countries (Gundelach et al. 2010; Suzuki 2017) or specific forms of volunteering (Stadelmann-Steffen 2011). Building on the crowding-in thesis, we assume that generous welfare states foster a culture of caring and mutual support and are, therefore, on average positively correlated with volunteering. Second, we take a group- and area-specific approach to examine whether particular welfare state policies, i.e., unemployment and pension benefits, are particularly positively related to volunteering among the targeted groups. Thereby, we paint a more fine-grained picture of the link between welfare states and social integration and delve into the mechanisms behind this link. We argue that it should matter whether or not an individual benefits from a state’s welfare policies. Regarding groups targeted by welfare state policies, we define those who currently or very likely benefit from a policy. Currently unemployed people or people with low education facing a high risk of becoming unemployed would be the target group for unemployment policies, for instance, while current pensioners or people close to retirement age would be the target group for retirement policies. For these targeted groups, crowding-in effects should be more likely because the welfare state addresses their needs and frees up resources for civic engagement. If, for instance, pension schemes are generous, pensioners and those who will retire soon do not have to worry about how to make a living. Instead of taking a marginal job or worrying about their future, they can put their time and energy into volunteering. In light of this line of reasoning, we study whether policies in specific areas of the welfare state unfold crowding-in effects for the targeted groups. In sum, we expect welfare state policies to be positively related to volunteering via two distinct mechanisms: by fostering a culture of helping and caring in the entire population (cultural mechanism) and by freeing up resources such as time and money for actual and potential target groups of welfare state transfers (resource mechanism).

Empirically, we approach this topic using survey data from the European Values Study 2017/2018 (EVS 2020), which is combined with welfare state data from the Comparative Welfare Entitlements Dataset (Scruggs et al. 2017) and Eurostat (2021). In total, we are able to cover 23 European countries in our comparative analysis. To estimate the link between welfare state spending and volunteering, we use hierarchical linear probability models. Our results lend support to the crowding-in thesis and demonstrate that the welfare state can enable social integration in the form of volunteering. Area- and group-specific analyses demonstrate that specific welfare state programs particularly enable their beneficiaries to engage in voluntary work: Particularly, pension welfare strengthens volunteering among the retired and the elderly. At the same time, unemployment welfare also reinforces volunteering among the less educated, but the results are slightly less robust. Finally, against expectations, the unemployed do not seem to profit from unemployment welfare.

Our findings can inform policy makers about side effects of welfare state reforms. Against the backdrop of demographic change, increased public expenditures during the current pandemic, and the economic consequences of the Russian invasion of Ukraine, discussions about spending cuts are likely to regain momentum in European societies. Thus, it is important to consider how welfare state spending is related to social integration and social capital across different societal groups. Spending cuts might backfire by harming social integration, particularly for vulnerable groups, and by fostering inequality in the acquisition of social capital.

2 Theoretical Framework

Research on social capital considers volunteering to be a key aspect of social integration. Bekkers (2008, p. 641) defines volunteering as “activities that benefit another person, group, or cause and that are carried out by individuals by their own choice and without pay”. It is important to distinguish volunteering from group membership. Volunteers do unpaid work that could be done by another person for payment, such as train an amateur soccer club (third-person criterion) (Bühlmann and Schmid 1999; Freitag et al. 2016). Thus, volunteering goes beyond membership in terms of commitment and is likely to produce closer and more sustainable ties. Doing voluntary work is a costly form of social integration that necessitates commitment over a longer time span. Accordingly, research has shown that individuals need to invest time and resources in these kind of activities (Wilson 2012).

While the link between welfare states and associational membership, social networks, or social trust has been studied before (Ferragina 2017; Gelissen et al. 2012; Kääriäinen and Lehtonen 2006; Kumlin and Rothstein 2005; van Oorschot and Arts 2005; van Oorschot and Finsveen 2009; Rothstein 2001; Sage 2015), volunteering is fairly understudied in this line of research (for exceptions, see Gundelach et al. 2010; Stadelmann-Steffen 2011; Suzuki 2017). What is more, a number of studies point to differential effects across different components of social integration. Kääriäinen and Lehtonen (2006), for instance, found a positive link between extensive welfare states and participation in associations, while there was no relationship with social networks and social support. Meanwhile, Oorschot and Arts (2005) reported a negative link between Nordic welfare states and trustworthiness but mixed findings regarding other forms of social capital. In addition, they found varying associations for spending and regime indicators of the welfare state. Overall, it is difficult to draw general conclusions on the link between welfare states and volunteering as a central aspect of social integration based on existing research. Yet volunteering as a resource-intense form of social integration should be particularly vulnerable to the lack of resources and, hence, sensitive to a compensation of resources by means of welfare state policies.

We address this gap in the literature and focus on the link between welfare state policies and volunteering as one of the cornerstones of civil society. Thereby, we build on a rich tradition of studies that discuss the role of welfare state policies for social integration. Two conflicting arguments have emerged in this literature. The crowding-out thesis assumes that an extensive welfare state takes responsibility away from civil society and individuals and, thereby, hampers the creation of social ties and social capital (Gundelach et al. 2010; van Oorschot and Arts 2005). In contrast, the crowding-in thesis argues that an extensive welfare state might enable civic engagement. Considering various empirical tests of these conflicting arguments, support for the crowding-in thesis is indeed stronger. A number of studies indicate that extensive welfare states relate to higher levels of various facets of social integration (Kääriäinen and Lehtonen 2006; Kumlin and Rothstein 2005; van Oorschot and Arts 2005; Rothstein 2001). If anything, crowding-out seems to take place mainly in forms of social integration that the welfare state directly encroaches, i.e., in the areas of elderly care, poverty, or health (Stadelmann-Steffen 2011). Thus, we build on this existing research and expect that crowding-in should occur for volunteering.

Two different mechanisms are assumed to account for the positive link between welfare state spending and voluntary activities: a cultural mechanism and a resource mechanism. The cultural mechanism refers to the argument that a generous welfare state is likely to establish a culture of helping and caring in a society. Societal norms of social responsibility and practices of care are supposed to be established between individuals and should “crowd in” volunteering (Visser et al. 2018). These societal norms are supposed to generally increase the levels of volunteering in a society as a whole, not only among recipients of welfare state transfers. Thus, we can expect an overall positive relationship between more extensive welfare state policies and volunteering:

H 1

A more extensive welfare state is positively correlated with a higher propensity to volunteer.

The second mechanism that can account for a positive link between welfare state policies and volunteering is the resource mechanism. It assumes that welfare state benefits free up resources such as time and money, as well as cognitive resources that individuals might spend for creating a public good (Gundelach et al. 2010; Kumlin and Rothstein 2005; van Oorschot and Arts 2005; Rothstein 2001; Visser et al. 2018). Voluntary engagement requires that these resources be available in sufficient quantities. Thus, the resource mechanism assumes that those who lack the necessary resources to engage in voluntary work and are targeted by the welfare state to support them should primarily benefit from welfare state policies. Taken seriously, this argument indicates that the relationship might be conditional on the type of welfare state spending and on the societal group that is targeted by this policy. The resource mechanism is not supposed to uniformly increase levels of volunteering in a society but only among actual and potential recipients of welfare state transfers. For these groups, volunteering should be more likely because of established social norms in generous welfare states (cultural mechanism) and because of additional resources they receive as beneficiaries of welfare state policies (resource mechanism). Resources come on top of societal norms for them and should further strengthen the positive relationship between welfare state generosity and volunteering among them. Nonbeneficiaries, in contrast, are unaffected by the resource mechanism—for them, crowding-in only occurs due to the cultural mechanism. Put differently, crowding-in effects are most likely to occur for those who actually or potentially benefit from a welfare state program. For instance, high pensions should particularly enable the elderly to engage in volunteering because they do not need to invest their time and resources in marginal employments to supplement their pension. Studying welfare state spending and social forms of volunteering directly touched by welfare states, Stadelmann-Steffen (2011) shows that such group-specific effects matter: Crowding-out occurs only for the well off, while crowding-in happens only for individuals with low income. She concludes that “crowding out and crowding in go hand in hand” and that it is therefore particularly relevant to pay attention to group-specific relationships (Stadelmann-Steffen 2011, p. 150). Visser and colleagues (2018, p. 276) come to a similar conclusion studying informal contacts: “Governments spending more on social protection may provide disadvantaged individuals with resources that relieve their financial burden, which allows them to (continue to) have intimate contact.” Thus, it is key to consider group-specific effects when analyzing the link between welfare state policies and social integration.

Building on this, we argue that the link between welfare state policies and social integration should vary across social groups and welfare state areas. Welfare state policies can craft social integration by making up for a lack of financial, time, or cognitive resources, which are crucial for volunteering. They will, however, not only provide targeted groups with missing resources that are necessary for volunteering but will more generally satisfy basic needs among targeted groups and, thereby, allow them to take care of their social needs (Visser et al. 2018, p. 264). Thus, crowding-in effects of welfare state policies should vary across groups and be most likely for targeted groups (see Visser et al. 2018 for a similar argument on informal social capital). These targeted groups include actual and potential beneficiaries of welfare state policies. In this respect, it is important to notice that not everyone has the same chance of becoming a beneficiary. Due to certain characteristics, some individuals have a higher risk or chance to be dependent on transfers in the near future. This makes welfare state policies in a specific area more relevant to these individuals and affects their resources for volunteering to a greater extent. They can keep on staying active members of society instead of taking up additional jobs if they can rely on generous welfare state schemes. In conclusion, welfare state policies likely enable volunteering for those individuals who currently benefit or will potentially benefit from it in the future. To disentangle these group-specific effects, we consider area-specific generosity of the welfare state by studying unemployment and pension policies. We focus on these two areas of the welfare state, in which direct financial transfers are the most important instruments. These financial transfers should play a relevant role in freeing up time and cognitive resources for volunteering among targeted groups.

With regard to unemployment policies, we expect two groups to benefit: unemployed individuals and lowly educated individuals. Unemployed people as the clear target group of these welfare state measures should immediately benefit. Financial aids should lower the burden of unemployment because they at least partly compensate for the previous income of the unemployed (Visser et al. 2018). This also reduces the psychological burden and is supposed to free up cognitive resources for volunteering among the unemployed. Moreover, considerations of reciprocity might offer an alternative explanation for increased levels of volunteering in this group of beneficiaries. Unemployed individuals who experience support by the state and society might feel obliged to give something back to society. This is supposed to motivate them to become volunteers. Yet it is also possible that this motivation only materializes in voluntary engagement once they have overcome the stigma of unemployment. Next to the unemployed, we also expect individuals with low levels of education to benefit as a potentially targeted group for unemployment policies. Particularly in European knowledge societies, lowly educated people carry a high risk of becoming unemployed. Generous unemployment programs will reduce their financial and psychological risks and enable them to be actively involved in society. They can rely on state payments that will bridge wage losses if their unemployment risk materializes. This might prevent them from taking up additional jobs, for instance, and keep them from withdrawing from social life and societal activities.

H 2

More extensive unemployment policies are more positively correlated with a higher propensity to volunteer among the unemployed.

H 3

More extensive unemployment policies are more positively correlated with a higher propensity to volunteer among the lowly educated.

The second area of welfare state policies that we consider concerns pensions. These financial transfers are even more suited to craft volunteering because they will not be subject to social stigma. Again, we can identify two social groups that should particularly benefit from pension policies: the retired and the elderly. If, for instance, replacement rates for pensions are high, pensioners do not have to work in minority employments to supplement their pension (Visser et al. 2018). Thus, they are supposed to have more time for volunteering. Similarly, generous pension policies will already unfold positive effects for those who are approaching retirement. The closer people get to retirement, the more they should value a generous pension policy. High financial transfers are supposed to reduce their potential financial risks and burdens and, related to that, the psychological burden as well. They do not have to look for a temporary job to supplement their pension or to worry about how to make ends meet as pensioners. Consequently, we argue, time and cognitive resources will be freed up for voluntary activities among the elderly, if pension policies are generous.

H 4

More extensive pension policies are more positively correlated with a higher propensity to volunteer among the retired.

H 5

More extensive pension policies are more positively correlated with a higher propensity to volunteer among the elderly.

3 Research Design

To test the hypothesized relationships, we made use of cross-sectional survey data from the European Values Study 2017/2018 (EVS 2020), which included over 29,000 respondents in 23 Western and Eastern European democraciesFootnote 2: Austria, Bulgaria, Croatia, the Czech Republic, Denmark, Estonia, Finland, France, Germany, Hungary, Iceland, Italy, Lithuania, the Netherlands, Norway, Poland, Romania, the Slovak Republic, Slovenia, Spain, Sweden, Switzerland, and the United Kingdom.

For our dependent variable, volunteering, we used a survey question asking respondents whether they did voluntary work in the last 6 months, with answers being either (0) no or (1) yes.Footnote 3 As to our key independent variable, we measured the size of the welfare state and two key longstanding welfare state programs—unemployment and pensions—using two different measures and data sources: (1) We used replacement rates from the Comparative Welfare Entitlements Dataset (Scruggs et al. 2017). This measures how much of the previous income of a typical worker is compensated in the case of pensions, sickness, and unemployment, taking the average of a single worker as well as that of a single earner with a spouse and two children. For our measure of the overall size of the welfare state, we took the average of the replacement rates of these three programs. Because of our theoretical focus, we excluded sickness replacement rates when looking at individual welfare programs and focused only on unemployment and pension replacement rates. (2) We used welfare state spending as a percentage of the gross domestic product (GDP) from Eurostat (2021), including both a measure of total welfare state spending aggregated over all programs (pensions, sickness, unemployment, disability, family/children, housing, social exclusion, and survivors) as well as specific welfare state spending on the two aforementioned programs, unemployment and pensions.Footnote 4 These measures have seen widespread use in welfare state research and have sparked a debate over which measure is more appropriate (Castles 1994; Green-Pedersen 2004; Ferrarini et al. 2013). The advantage of social expenditure data is their widespread availability in the OECD world. However, this measure is also contingent on developments in the economy and demographic changes that do not amount to actual differences in the size of the welfare net. It is standardized by GDP, and spending is affected by how many people in a country have a right to welfare state benefits, i.e., the share of people unemployed or older than retirement age. Replacement rates, in contrast, do not suffer from these issues. Instead, they focus on only one part of the welfare state, considering only income replacement but no other social benefits. In addition, data availability is much more limited. In fact, the latest data available, which we use in our analysis, are from 2010 and are absent for two of our countries, Croatia and Iceland.Footnote 5 Since both of these measures have their distinct advantages, we employ both separately in our analysis (Kunißen 2019). In order to keep the welfare state measures comparable, we rescaled them to range from 0 to 1.

For hypotheses H 2 to H 5, we looked at the relationship between welfare state programs and volunteering for specific focus groups. In doing so, we distinguished between actual and potential beneficiaries of the welfare state programs. For actual beneficiaries of unemployment benefits, we made use of a dummy variable for (1) unemployed vs. (0) not unemployed respondents. As to potential beneficiaries of unemployment benefits, we looked at the highest level of education achieved (nine-level International Standard Classification of Education [ISCED] classification), following the logic that respondents with a lower education carry a substantially higher risk of being unemployed. For actual beneficiaries of pension benefits, we used a variable differentiating (1) retired from (0) nonretired respondents. Finally, we used the age of respondents, since the salience of pensions should generally increase with age as people come closer to their retirement age.

We controlled for a set of sociodemographic variables that are commonly employed in research on this topic (e.g., van Oorschot and Arts 2005; Visser et al. 2018). Our models control for the age and gender of the respondents, their highest level of education achieved (nine-level ISCED classification), and their income decile. Additionally, we considered their marital status (single, married, widowed, divorced/separated), whether they had children, whether they had a first- or second-generation migration background, and their political ideology on a left–right scale (squared). Table A1 in the Online Appendix shows a detailed overview of summary statistics for all variables used in the analysis.

We accounted for the multilevel data structure by running hierarchical linear probability models with individual respondents at the first level and countries at the second level.Footnote 6 All of our models employ robust standard errors (SEs). In the first part of our analysis, we tested the overall as well as the area-specific relationships between welfare programs and volunteering. For this, we regressed our dependent variable volunteering Voli on our respective welfare measures Welfj as well as a set of control variables CVi.

$$\mathrm{Vol}_{i}=\beta _{1}\mathrm{Welf}_{j}+\beta _{2}CV_{i}+u_{j}+\varepsilon _{ij}$$
(1)

In the second part, we address whether the link between welfare state programs and volunteering is group-specific and depends on whether an individual is targeted by a certain welfare program. For this, we added interaction terms between our individual welfare programs and our variables for actual and potential beneficiaries Beni. To be precise, we tested interaction effects of our unemployment and pension benefit variables with unemployment status and education as well as retirement status and age, respectively.

$$\mathrm{Vol}_{i}=\beta _{1}\mathrm{Welf}_{j}+\beta _{2}\mathrm{Welf}_{j}*\mathrm{Ben}_{i}+\beta _{3}CV_{i}+u_{j}+\varepsilon _{ij}$$
(2)

4 Empirical Findings

Figure 1 presents a first descriptive overview of the hypothesized correlation between the welfare state and volunteering. In a comparison of the 23 countries in our sample, the share of respondents doing voluntary work varies substantially from a minimum of 10.3% in Spain to a maximum of 47.6% in Norway. Despite the conceptual and empirical differences between the two welfare state measures, they are both clearly positively correlated with the share of respondents doing voluntary work in each country—mean replacement rates at r = 0.551 and total social expenditure as a share of the GDP at r = 0.594. Table 1 presents the results of our hierarchical linear probability regression models, which more formally tested this relationship by evaluating voluntary work at the individual level and by including a range of control variables. Both models confirm the correlational evidence: Respondents living in a country with a more extensive welfare state are significantly more likely to do voluntary work. In fact, increasing the mean replacement rate by a full standard deviation (i.e., by 0.221) raises the likelihood to engage in voluntary work by 6.0%. Increasing the replacement rate from its minimum to its maximum value increases the likelihood to do voluntary work from 11.5% to 38.0%. Similarly, a standard deviation increase in the total social expenditure (i.e., by 0.322) results in a 5.6% higher likelihood to do voluntary work. Over the full range of total social expenditure, the likelihood of volunteering increases from 16.5% to 35.1%. Overall, these findings lend support to hypothesis H 1 and indicate a crowding-in effect: Generous welfare state policies are related to higher levels of volunteering. The proposed cultural mechanism seems to be at work, meaning that an extensive welfare state increases the general levels of voluntary work in a society.

Fig. 1
figure 1

Country-level correlation between share of respondents doing voluntary work and welfare state measures. (Data sources: Eurostat, 2021; EVS, 2020; Scruggs et al., 2017)

Full size image
Table 1 Regression models of the relationship between welfare state and voluntary work. (Data sources: Eurostat, 2021; EVS, 2020; Scruggs et al., 2017)
Full size table

Looking at the control variables, most of them point in the expected direction. While voluntary work does not seem to differ for age or gender, we find substantial effects for both education and household income. In addition, marital status does not make any difference, whereas having children increases the likelihood to do voluntary work. Volunteering is less common for respondents with a second-generation migration background, but not for those with a first-generation migration background. Finally, left-wing respondents tend to be more active as volunteers.

While this confirms our expectation of a positive relationship between the welfare state and voluntary work, this overarching measure may hide differences between specific welfare programs. Therefore, Fig. 2 displays the relationship between two types of welfare programs and volunteering. It compares the coefficients of the mean replacement rate or total social expenditure as discussed in Table 1 with the respective coefficients for pension and unemployment benefits. Table A2 in the Online Appendix shows the full models. In general, estimates for the area-specific welfare state programs confirm the overall pattern of a positive link between a generous welfare state and volunteering. For both replacement rates as well as for social expenditure, unemployment and pension benefits also significantly foster voluntary work by themselves.

Fig. 2
figure 2

Regression coefficients of the relationship between welfare state programs and voluntary work. Displayed are hierarchical linear probability regression coefficients with 90% confidence intervals. Each coefficient displays a single model that includes one of the welfare state measures. (Data sources: Eurostat 2021; EVS 2020; Scruggs et al. 2017)

Full size image

In general, we thus find a positive relationship between the welfare state and voluntary work irrespective of the specific welfare state program. There may, however, be differences between these specific welfare programs for certain groups of individuals. Coming to the core of our analysis, we argued that for those who are most likely to benefit or those who do actually benefit from a specific welfare state program, this program should display a particularly strong relationship with the likelihood to engage in voluntary work. In particular, we argued for four interactions: First, we expected the relationship between unemployment benefits and volunteering to be stronger for those who receive them, i.e., unemployed respondents. With regard to potential recipients, lower educated respondents are particularly likely to work in insecure jobs and face a higher risk of unemployment. The relationship between unemployment benefits and volunteering should thus be stronger for respondents with a lower education. Second, we also expected the relationship between pensions and volunteering to be stronger for those who benefit from them, i.e., retired respondents. In addition, the salience of pensions should generally increase with age as people come closer to their retirement age. Also, more elderly people become less capable of working and thus more dependent on their pension income. The relationship between pensions and volunteering should thus be stronger for the elderly as well.

Figure 3 displays the results of our interaction models for unemployment benefits. It shows the marginal effects of unemployment replacement rates and social expenditure on unemployment respectively for unemployed vs. not unemployed respondents (panel a) and at different levels of education (panel b). The results are mixed for actual and potential recipients. On the one hand, we find no evidence that actual recipients of unemployment benefits profit more from them in terms of being more likely to volunteer. The interaction term of unemployment replacement rates with unemployed status is positive at 0.069 (SE = 0.063) but nonsignificant. For unemployment expenditure, the interaction term is −0.101 (SE = 0.078) and is also nonsignificant. On the other hand, our models for potential benefits confirm the expectations of our second hypothesis for both welfare state measures. The interaction term of unemployment replacement rates with education is negative at −0.009 (SE = 0.005) and significant at the 10% level. At the lowest level of education, a standard deviation increase in the unemployment replacement rate (i.e., by 0.271) raises the likelihood to engage in voluntary work by 5.0%, while at the highest level of education, the same increase amounts to only a 3.0% increase in the likelihood to do voluntary work. The result for unemployment expenditure is substantively similar. The interaction term is negative at −0.014 (SE = 0.008) and also significant at the 10% level. At the lowest level of education, a standard deviation increase in the unemployment expenditure (i.e., by 0.281) makes respondents 3.8% more likely to do voluntary work, whereas at the highest level of education, the same increase only leads to a 0.1% higher chance of volunteerism. The full models can be found in Table A3 in the Online Appendix.

Fig. 3
figure 3

Marginal effects of unemployment benefits on voluntary work at different levels of unemployment status and education. Displayed are marginal effects of unemployment welfare measures on voluntary work a for unemployed vs. not unemployed respondents as well as b at different levels of education with 90% confidence intervals, calculated from hierarchical linear probability regressions. Below the graphs are the coefficients, standard errors, and significance levels of the interaction coefficients. Data sources: Eurostat 2021; EVS 2020; Scruggs et al. 2017

Full size image

Figure 4 displays our results for pension benefits. It shows how the marginal effects of pension replacement rates and pension expenditure change for retired vs. nonretired respondents (panel a) as well as according to the age of respondents (panel b). The hypotheses for both actual and potential beneficiaries of pensions receive empirical support. Looking at actual beneficiaries, the positive relationship between pension programs and volunteering is stronger for retired respondents. The interaction term for pension replacement rates with retired status is positive at 0.103 (SE 0.044) and significant at the 5% level. For nonretired respondents, a standard deviation increase in the pension replacement rate (i.e., by 0.268) increases the likelihood to volunteer by 5.1%, while for retired respondents, the same increase raises volunteering by 7.7%. At the same time, the interaction term for pension expenditures is also positive at 0.148 (SE = 0.025) and significant at the 0.1% level. A standard deviation increase in the pension expenditure (i.e., by 0.315) increases the likelihood to engage in voluntary work by only 2.9% for nonretired respondents, but it increases by 7.3% for retired respondents. Looking at potential beneficiaries, the results are substantively similar when using age as the moderator. The interaction term for pension replacement rates is positive at 0.003 (SE = 0.001) and significant at the 5% level. Respondents aged 18 (the minimum in our data set) are 3.6% more likely to engage in voluntary work and respondents aged 82 (the maximum) are 5.9% more likely to do so when pension replacement rates increase by a standard deviation. Similarly, the interaction term for pension expenditures amounts to 0.004 (SE = 0.001) and is significant at the 0.1% level. Respondents aged 18 are 0.5% more likely to do voluntary work, and respondents aged 82 are 4.1% more likely. As Fig. 4 illustrates, pension expenditures and volunteering are not only significantly positively linked for pensioners but are also significantly positively linked in younger age groups. An explanation for this finding might be that prospectively high pension benefits free up resources in earlier life stages as well. Private savings become less relevant, and this reduces financial burdens across all age groups. Still, the significant interaction effect indicates that the elderly, and particularly the retired, benefit more strongly from these welfare state expenditures. The full models can be found in Table A4 in the Online Appendix.

Fig. 4
figure 4

Marginal effects of pension benefits on voluntary work at different levels of retirement status and age. Displayed are marginal effects of pension welfare measures on voluntary work a for retired vs. nonretired respondents as well as b at different ages of respondents, with 90% confidence intervals, calculated from hierarchical linear probability regressions. (Data sources: Eurostat 2021; EVS 2020; Scruggs et al. 2017)

Full size image

Next, we tested whether actual or potential beneficiaries profited more from the welfare state program that targeted them or also from the respective other welfare state program. If the resource mechanism holds, we should find weaker or no interaction effects for welfare state programs that do not target the recipients. In other words, unemployment status or education should not affect the positive relationship between pension programs and volunteering, while retirement status or age should not affect the positive relationship between unemployment programs and volunteering. This was tested by including measures for both welfare programs simultaneously as well as interaction terms with the variables designating beneficiaries into the models. The results of these models are displayed in Table A5 and Table A6 in the Online Appendix. For retired and elderly respondents, the resource mechanism argument finds full support. The interaction with pension welfare remains significant and similar in size even when controlling for unemployment welfare, and at the same time, there is no significant interaction with unemployment welfare. For unemployed and less educated respondents, the results are more mixed. As in the main models, the interaction between unemployment status and unemployment welfare remains nonsignificant. In addition, the already substantively fairly small interaction between education and unemployment replacement rates becomes nonsignificant, whereas the interaction with unemployment expenditure remains significant and even becomes slightly greater. At the same time, there is no significant interaction with pension replacement rates. For pension expenditure, however, we find an interaction with unemployment status and education. Pension expenditure actually seems to be less beneficial for unemployed or lower educated respondents. This means, at the same time, that the cultural mechanism does not strike through for these groups. There may be two reasons for this finding: First, the unemployed or lower educated respondents not only are not the direct beneficiaries of pension expenditure, but they may actually profit less from them in the future, seeing as most pension systems are insurance based. Second, pension programs are generally popular and less likely to be cut (Jensen et al. 2018; Wolf et al. 2014). Higher pension expenditures may in times of permanent austerity increase the pressure on other, less popular welfare programs—in particular, labor market and unemployment schemes (Jensen 2012; Pierson 1994). (Potential) beneficiaries of unemployment schemes may thus be less certain that they will keep receiving (or will receive) these benefits in the future. Given the substantial correlations between the two welfare programs, however, these results should be taken with a grain of salt.

Finally, we conducted four additional robustness checks to test whether our results are stable with regard to certain alterations in our modeling strategy. First, we used hierarchical logit models instead of hierarchical linear probability models. Second, we used jackknife standard errors to test whether our results are robust to the exclusion of individual countries. Third, we included additional country-level control variables in the models. For the models using replacement rates as the welfare state measure, we included GDP per capita (logarithmized), the unemployment rate, and the elderly share (share of population aged 65 or older relative to the share of population in a working age between 18 and 65) to account for different factors that determine the share of the population likely to receive welfare benefits. For the models using social expenditure, we controlled only for the unemployment rate and the elderly share because social expenditure is already standardized by the level of GDP of a country. Finally, we replicated the expenditure models with data from 2010, the same year as with our replacement rate data. As can be seen in Figs. A1, A2, and A3 in the Online Appendix, this does not substantially change the results. Only in the case of unemployment expenditure does the relationship become marginally nonsignificant when using jackknife standard errors. However, judging from the distribution of the coefficients, this does not seem to be driven by single countries and may thus be a result of lowering the already fairly limited number of countries included in our study.

Overall, our empirical results confirm existing findings in the literature and lend support to the crowding-in thesis (Kääriäinen and Lehtonen 2006; Kumlin and Rothstein 2005; van Oorschot and Arts 2005; Rothstein 2001). Generous welfare states foster social integration by enabling volunteering. Using a group- and area-specific approach to the study of welfare and volunteering, we find that this positive relationship is particularly pronounced in social groups that are targeted by specific welfare state programs. If, for instance, replacement rates for pensions are high, the retired and the elderly are more likely to volunteer. This supports previous claims in the literature that welfare state programs should unfold their enabling force particularly for those who benefit from them (Stadelmann-Steffen 2011; Visser et al. 2018). At the same time, we find weaker but nevertheless substantial and significant relationships between welfare state policies and volunteering, even for those who were not targeted by them in most models. Overall, this leads us to conclude that both mechanisms seem to play a crucial role in explaining volunteering. The assumed cultural mechanism is supported by the finding that even among those not targeted by the welfare state policies, these policies lead to a crowding-in effect, increasing voluntary engagement. The resource mechanism also seems to be true, seeing as those targeted by the welfare state policies benefit substantially and significantly more from them than those who were not targeted by the welfare state policies.

5 Conclusion

This paper analyzes the link between welfare state expenditures and volunteering as a key aspect of social integration, taking a nuanced look at group-specific and policy area–specific relationships. Social integration refers to the principles by which individuals in a society are linked to each other (Lockwood 1964). The spectrum of volunteer activities includes involvement in sports, hobby, and leisure clubs; unpaid work in the social, health, or cultural spheres; voluntary holding of political office; and mutual aid among neighbors or acquaintances bound by friendship. It would be difficult to imagine society without volunteering without at the same time accepting a painful loss of diversity and, above all, of the quality of social life. The voluntary commitment of citizens is a precious and almost priceless commodity that sets large parts of daily processes in motion. Volunteering, it is often said, is the glue that holds society as a whole together (Stadelmann-Steffen et al. 2010).

By analyzing data from the European Values Study 2017/2018 (EVS 2020) in 23 countries combined with data on welfare state expenditures and generosity on the national level, we show that generous welfare states are positively linked to volunteering and, thereby, foster social integration, lending support to the crowding-in hypothesis. By breaking up overall measures for the welfare state and considering target groups of different welfare state programs, we also find evidence for the resource mechanism. For some societal groups targeted by specific welfare state programs, such as the retired or elderly for pension welfare or the less educated for unemployment welfare, our results show higher levels of volunteering. This indicates that the targeted groups indeed benefit most from the welfare state programs in these areas. However, the resource mechanism cannot be confirmed for all areas under investigation. We do not find the expected group-specific relationship for the unemployed. For this group, the mechanism of resource compensation seems to be undermined by mechanisms of social isolation and stigmatization. As a result, extensive unemployment assistance does not contribute to greater social integration of the unemployed through volunteering. Another difficulty in this regard may be a social desirability bias in our measure for unemployment status—only a rather small share of respondents indicated that they were unemployed. For the other groups that we examined, however, the resource mechanism is confirmed by our analyses: Those targeted by the welfare state policies benefit most from them in terms of their civic engagement.

In summary, welfare state policies are positively related to volunteering for all citizens (cultural mechanism), but they particularly free up time, money, and cognitive resources for those who are actually or potentially targeted by them (resource mechanism). However, the resource mechanism does not seem to work for the unemployed.

In evaluating our approach, three limitations should be kept in mind. First, for reasons of data availability, our study was carried out in the European context with only a limited number of countries. Further research is necessary to assess the generalizability of our findings beyond the European context. Second, our dependent variable is a general measure of volunteering that does not allow for distinguishing between formal and informal volunteering or between different segments of volunteering, such as volunteering in the social sector. Thus, we do not know exactly for which kind of activities individuals use their resources freed up by the welfare state. In this regard, future studies should disentangle crowding-in and crowding-out effects between different forms of volunteering (see also Stadelmann-Steffen 2011). Third, we can present only correlational evidence given the cross-sectional, observational nature of our data. Longitudinal studies on volunteering, such as the example by Visser et al. (2018), as well as experimental evidence—for instance, in the course of basic income lotteries—would be suited to further tease out the causal mechanisms between welfare state spending and volunteering.

Despite these limitations, our findings carry important implications for social integration research and for societal debates about welfare state expenditures and welfare state retrenchment. First, when we examine the role of institutions in crafting volunteering as a cornerstone of social integration, it is important to take a nuanced stance on this issue. The general expectation is that institutions can diminish inequalities and compensate for a lack of resources on the individual level. Therefore, it is important to evaluate whether they meet this expectation by studying group-specific effects. Future research should take further steps in this direction by including additional targeted groups and area-specific policies, such as families and welfare state support for child and elderly care. Welfare state programs, the institutional arrangement studied in this paper, can contribute to greater equality in social integration only if they are able to free up time and cognitive resources for those who are less likely to be involved, given their individual characteristics. This should be kept in mind when the role of institutions is studied scientifically and also when the creation of institutional arrangements to foster social integration is discussed in society. Second, while we focused only on expenditures and income replacement to measure welfare state policies, it might be valuable to take a closer look at nonmonetary aids and their potential to foster social integration in future research. This would provide further guidance to policy makers in how to design and implement effective social policies that benefit not only the targeted groups but also society as a whole. Finally, our study underlines the effectiveness of welfare state expenditures in crafting social integration. This is an important finding that can inform debates about the future of welfare states, which are likely to regain momentum in the aftermath of the COVID-19 pandemic and the consequences of the Russian invasion of Ukraine. Even wealthy European countries do not have unlimited financial resources and will have to make decisions about their priorities, given current and future societal challenges such as climate change, migration, and the energy crisis. In debates about public policy priorities, it is important to consider the importance of the welfare state in the creation of social integration. Our research shows that welfare state measures are able to mitigate inequalities in volunteering by freeing up resources for lowly educated and elderly or retired persons. Thus, welfare state retrenchment could unfold unintended consequences in terms of social cohesion. Yet particularly in difficult times of crises and insecurity, societies have to keep an eye not only on their financial resources but also on their social resources. It has to be noted, however, that the stimulation of voluntary work is always viewed with a critical eye. Fears are also repeatedly voiced that volunteer work can displace existing jobs, not least for reasons of the austerity policy, which can pursue unpaid voluntary work as a cost-reducing instrument. However, such a policy will almost certainly contribute to social disintegration in the long run.