The author can be reached at dvdduin@gmail.com. Brian Burgoon is Professor of International and Comparative Political Economy at the University of Amsterdam (B.M.Burgoon@uva.nl). Theresa Kuhn is Associate Professor in Political Science at the University of Amsterdam (T.KuhnLancee@uva.nl).
Favouring Social Europe
over The National
Welfare State
How Dependence on Financial Markets
Influences Support for Social Security
David van der Duin
10581111
Supervisor: Brian Burgoon
Second reader: Theresa Kuhn
Research Master Social Sciences
University of Amsterdam
1 Following the euro crisis, debates in political science have considered the need of European social security policies, and the sufficiency of support among the public for such measures. While quite a few scholars have investigated the foundations of support for European social security in general, people who are dependent on the financial market for their income have remained outside the scope of this research. However, prior investigations show that being exposed to the financial market influences one’s political attitudes. Therefore, this study investigates how financial dependence influences attitudes toward social security measures at the national and the European level. It is hypothesized that people who are dependent on the financial market for income are opposed to both national and supranational social security measures. Moreover, in the case of the latter, this group is expected to prefer less generous forms of social security in particular. However, supranational security measures can be expected to be to some extent appealing to people who are dependent on financial market, because these measures are likely to reduce the volatility of the financial market and in that sense reduce the risk exposure of those who are dependent on this market. Therefore, it is expected that if this group of people is confronted with the choice between supranational and national social security measures, they would prefer the former relative to the latter. These expectations are tested using data from a novel survey that includes a conjoint experiment designed to test support for supranational social security, as well as round 8 of the European Social Survey.
Introduction
Since the 90’s of the previous century, the European Union (EU) has tried to expand its area of conduct into what used to be the domain of the national welfare state. Time and again, the European commission has tried to subject national social policies such as industrial relation regimes to an international framework (Leibfried and Pierson, 1995a; ibid., 1995b). This creation of European social policy has however ever since been a frustrated effort (Streeck, 1995; Leibfried and Pierson, 1995b).
Recent developments in political science have attempted to shed light on this problem by investigating the base of support for European social policy among European citizens. These investigations have for example tried to explain cleavages among the European population regarding the bailout measures during the Euro crisis (Bechtel, Hainmueller and Margalit, 2014) and (the prospects of) concrete forms of European social policies, such as a European wide social benefit scheme (Gerhards, Lengfeld and Häuberer, 2016) or an unemployment benefit scheme (Dolls and Werfhöfer, 2018). Like research on support for European Integration in a broad sense (Hooghe and Marks, 2004), these investigations have highlighted that cultural values are strong determinants of attitudes toward the EU, in addition to demographic factors such as age and education (Baute et al., 2018a; Baute, Meuleman and Abts, 2019; Gerhards,
2 Lengfeld and Häuberer, 2016; Bechtel, Hainmueller and Margalit, 2014; Kuhn, Solaz and Elsas, 2017).
This study however focusses on socioeconomic position. While this and related factors such as dependency on social policy are generally found to be strongly associated with national variants of social policy (Meltzer and Richard, 1981; Rehm, 2009; Rehm, Hacker and Schlesinger, 2012), these factors seem to be less important in explaining European social policy. However, one group of citizens that has remained outside the scope of these investigations are the owners of financial instruments and people who otherwise draw their income from financial sources. Recent investigations have shown that this group of people can diverge substantially in their preferences for social policy from people who have a different income source, sometimes in surprising ways (Kaustia, Knüpfer and Torstila, 2016; Jha and Shayo, 2018). Moreover, this group of actors has a profound influence on policymaking, especially under conditions of complete capital mobility as in Europe (Bearce, 2003; Watson, 2007). Therefore, it is important to understand how people exposed to the financial market position themselves toward supranational forms of social policy, and how this position is formed as such.
This research aims to shed light on this issue by investigating the support for national social security and supranational-European social security among people who are dependent on the financial market for the largest share of their income. This first involves exploring support for national-level assistance among financially dependent people versus other European residents. Thereafter, it investigates support for a European Benefit Scheme (EUBS), a scheme currently being discussed by European politicians that would insure European countries for high unemployment benefit expenditures during an economic crisis, as well as a European wide basic income scheme (EUSBS). Thereafter, support for these different levels is compared.
The central question asked is: To what extent and under what conditions do people who are dependent on the financial market for their income support or oppose supranational social security measures, and to what extent is this position toward supranational social security similar to their attitudes toward national forms of such measures? In addition, the following three sub-questions are posed to answer this main question: (1) What specific features or design elements of an EUBS cause support or opposition among people that are dependent on the financial market for income?, (2) How are these attitudes toward specific policy elements among people dependent on the financial market different from people who derive their income from other sources? and (3) To what extent does support for or opposition to supranational social security measures among people dependent on the financial market differ from their attitudes toward national social security measures?
3 The second section will develop a theoretical framework that synthesises insights from macroeconomy and political economy to form concrete expectations regarding the relationship between financial market dependence and support for national and supranational social security. It is conceivable that financial dependence does not relate to attitudes toward social security because cultural values instead of economic motives drive these attitudes. However, multiple strands of literature oppose such a proposition. First, the political economy literature that has investigated the relationship between financial market performance and fiscal policy tells us that financial market actors have a strong preference for tight fiscal policies that keep inflation and taxes low (Scharpf, 1987; Mosley, 2000; Akitoby and Stratmann, 2008). Second, micro-level analyses have shown that exposure to the financial market through house ownership (Ansell, 2014) and the ownership of financial assets (Kaustia, Knüpfer and Torstila, 2016; Aiken, Ellis and Kang, 2017; Jha and Shayo, 2018; Kerner, 2018) can have a negative influence on support for social security and move one’s ideology toward the right. However, while these investigations tap the topic of interest here, they either investigate the relationship between financial market dependence and political attitudes in a general sense (Kaustia, Knüpfer and Torstila, 2016; Jha and Shayo, 2018), or their approach is insufficient to draw conclusions about the relationship investigated here (Ansell, 2014).
In any case, these findings lead to the general expectation that financial market dependence creates opposition toward social security. However, if social security would be introduced on the EU-level, as with an EUBS, this might actually indirectly reduce the risk exposure of people dependent on the financial market. Therefore, these people are expected to prefer supranational forms of social security relative to national forms.
These expectations will be tested using two datasets. The first is a novel dataset including 13 countries that is specifically designed to measure support for an EUBS by means of a conjoint experiment. In the survey, respondents were asked to rank and rate six ‘policy packages’ which randomly differed on six characteristics. This allows an analysis of the relationship between these features and support for an EUBS with a fair claim to causality (Hainmueller, Hopkins and Yamamoto, 2014). In addition, this dataset allows the identification of people who are for the largest share of their income dependent on the financial market. Second, the findings based on this first dataset will be buttressed by analysing round 8 of the European Social Survey (ESS). While this dataset does not allow an as detailed analysis of support for supranational social security as with the EUBS data, it covers more countries and generally draws larger samples within those countries, allowing for an assessment of generalisability of the findings based on the EUBS data as well.
4 The results show that while the analysis fails to find a relationship between financial dependence and attitudes toward national social security, people who are dependent on financial market for their income are actually more supportive of an EUBS than are people with a regular source of income. Moreover, they have one diverging preference regarding the characteristics of such a scheme, namely that the its level of generosity should be low. In contrast, they are not more, or less, supportive of an EUSBS than people with a regular source of income.
The rest of the paper is structured as follows. The next section develops the theoretical framework and hypotheses, explaining that financially dependent people can be expected to oppose social security, but that they might have a relative preference for the supranational organisation of such measures. The third section describes the data used and how it is operationalised. The fourth section provides the formal discussion of the empirical results, while the fifth section discusses these results in their substantive detail. The sixth section concludes.
Theoretical framework
According to classic conceptions of welfare state attitudes, the main lines of cleavage are formed on the basis of income and socioeconomic position (Korpi, 1983; Meltzer and Richard, 1981). This argument entails that people with a high income oppose welfare state measures, as they have to pay the highest taxes to fund welfare state institutions, while they do not or hardly enjoy its benefits. Conversely, people with a low income pay less taxes and benefit more from the welfare state, and therefore they support it.
Recent advances in political science have shown that this conception is too simplistic. Scholars belonging to the ‘revisionist’ school (e.g. Moene and Wallerstein, 2001) argue that welfare state institutions should be conceived of as an insurance device. Therefore, risk exposure, for example by having a high chance of becoming unemployed in the future, is also an important source of support for the welfare state (Rehm, 2009). Moreover, individual support for different domains of the welfare state can also be explained by the extent to which a policy in a certain domain would cover that individuals’ risk exposure. For example, someone can support social policy targeting housing more or less than social policy targeting unemployment, depending on the extent to which that individual’s risk exposure is reduced by that particular policy (Rehm, Hacker and Schlesinger, 2012).
5 Both of these two perspectives suggest that actors who are dependent on the financial market oppose welfare state policies. As these policies are generally targeted at the sick, old and unemployed, they do little to redistribute income to or reduce the risk exposure these people, while the latter nonetheless pay for such measures through taxes. However, to the author’s knowledge, neither practitioners of the classic conception of support for the welfare state nor the revisionist school have empirically investigated the relationship between financial market dependence and support for social security. Therefore, we will now turn to two adjacent sets of literature that allow us to form expectations about this relationship. First, it will be discussed how the effects of different forms of macroeconomic policy shape the preferences of financial market actors.
Macroeconomic dynamics
Since many decades, capital mobility has been increasing (Watson, 2007), and within the EU there has virtually been a complete removal of capital barriers. Under these conditions, the effects of fiscal and monetary policy have taken centre stage in domestic politics. This political economy of who wins and who loses from different forms of macroeconomic policy is informative for the current investigation insofar as it predicts the preferences of financially dependent people in an indirect manner, that is, assuming that financial market actors are aware of these dynamics and behave rationally accordingly.
For example, loose monetary policy and expansionary fiscal policy can be in the interest of actors and sectors with domestic interests, while the macroeconomic effects of these policies, such as inflation and exchange rate depreciations, create losses for exporting economic actors or anybody who depends on a stable exchange rate for economic activity (Frieden, 1991; Kenworthy, 2002). Although the euro has removed exchange rates within the EU, similar dynamics operate for those macroeconomic indicators that remain. Actors that operate intensively in the international markets prefer openness and have an aversion toward expansionary fiscal and loose monetary policy as these can deteriorate the performance of markets and the competitive position of a country. This in turn creates domestic cleavages with regard to such government measures: while domestically oriented actors are relatively indifferent toward inflation and government deficits, these same measures are worrisome for internationally oriented economic actors (Bearce, 2003; Boix, 1998; Frieden and Rogowski, 1996).
6 Those dependent on the financial market for income are to be counted among these internationally oriented actors. A stockowner, for example, has reason to worry about any policy that might negatively influence the competitive position of the companies that he owns, such as tax increases. In addition, those who own bonds or other forms of credit should have a particular aversion toward inflation, as this reduces the value of their credit (Scharpf, 1987).
Mosley (2000) shows that these expectations are not just hypothetical. When deciding in what government bonds to invest, bond market actors are first and foremost concerned with inflation and with fiscal deficits. As high inflation rates reduce the returns from credit, and as fiscal deficits increase government debt and therefore the government’s incentive to inflate this debt, the investors interviewed by Mosley mentioned a specific preference for a maximum of 2 percent inflation. However, how governments achieved this low rate of inflation is of less importance. In fact, the investors were relatively unconcerned with taxation policy. Mosley (2000) buttresses these qualitative findings with a quantitative analysis of financial market performance that shows that the effects of other potentially important indicators, including the taxation regime, are of minimal importance compared to inflation and the fiscal deficit in explaining bond rates. This analysis also showed an indifference toward the size of government expenditures.
These findings are in line with more recent studies that show that what is important for financial market performance is not the size of government expenditures, but instead the means by which this expenditure is funded. More specifically, while financial markets penalise governments for debt-financing fiscal expenditure by increasing bond premiums, this penalisation occurs less when fiscal expenditure is financed through taxes (Akitoby and Stratmann, 2008; Lierse and Seelkopf, 2016). As such, bond market actors do not necessarily have a problem with high taxation, but are instead mostly concerned with excessive deficits, debt and inflation.
It should be noted that the category of people of interest to the research presented here is broader than the bond market actors researched by Mosley (2000) and Akitoby and Stratmann (2008). As will be discussed below, the vast majority of the people investigated here earn financial-market income on a private basis, i.e. without being employed in the financial sector. We can expect this to cause a stronger sense of responsibility: while for professionals their performance on the financial market is the main source of income, private owners depend on this performance for both their income and their stock of wealth. In that sense, failing financial markets can be expected to have stronger consequences for the people studied here than for professionals.
7 In addition, the dynamics that Mosley (2000) and Akitoby and Stratmann (2008) explicate are slightly different in the case of stock markets. The relationship between stockownership and fiscal policy preferences is ambiguous: fiscal policy might stimulate the value of stocks in the short run, but it negatively influences the performance of stock markets in the long run. More specifically, fiscal expenditures can lead to higher real interest rates by increased bond issuing and crowding-out effects, and higher interest rates tend to dampen corporate profits. Moreover, stockowners are more concerned with the level of taxation than bond holders (Mosley and Singer, 2008). In this sense, fiscal policy is worrisome for stockholders. As such, while the macroeconomic policy preferences of bondholders and stockowners are not completely aligned, they do not oppose each other.
In any case, this relationship between macroeconomic policy and financial market performance is a first step in shaping general expectations regarding the relationship between financial market dependence and attitudes toward social security policies. While it seems plausible that the specific preferences depend on the portfolio composition of an owner of assets, anyone dependent on the financial market for income has a reason to oppose such social security measures, as these will either increase inflation, interest rates and/or taxation.
To refine these expectations, we now turn to recent microlevel analyses of the relationship between financial market dependence and political behaviour.
Microlevel analyses
The analyses conducted on the relationship between financial market dependence and political attitudes can be grouped into two categories. The first concerns the relationship between house ownership and welfare state attitudes, while the second concerns direct financial market exposure through the ownership of financial assets. Both are informative for the purpose of this research: the first investigates welfare state attitudes specifically, but looks at a different yet somewhat similar explanatory factor. The second looks at the same explanatory factor that is of interest here, but only investigates political attitudes in a broad sense.
The political dimension of private housing might not be apparent at first sight. However, private housing has been a crucial target of conservative governments since at least the 80’s of the previous century. Boix (1998) discusses that the policies of these governments, such as mortgage relief, are geared specifically towards an asset owning middle class. Moreover, these policies, as well as privatizations, also increase asset ownership among the population (Biais and Perotti, 2002; Kaustia, Knüpfer and Torstila, 2016; Kerner, 2018). If the number of citizens
8 owning assets of substantial value increase in the population, it is likely that the political attitudes and voting behaviour of this population shift toward policies beneficial to asset owners. In fact, this expectation was an explicit motivation of the conservative reforms enacted by the Thatcher government (Boix, 1998), the Bush administration and the governments of many Latin-American countries (Kerner, 2018).
Investigating the relationship between support for social security and house prices, Ansell (2014) has shown that these expectations were right. He argues that when the value of a house increases, the income of its owner increases as well, even if the illiquidity of a house makes the flow of this income rigid. Subsequently, this increase in the value of a house leads to a reduction in support for welfare measures through four mechanisms: first, higher asset value leads to higher taxation; second, the value of assets may cause people to lose their eligibility for welfare transfers; third, income through asset value reduces demand for redistribution (conform Meltzer and Richard, 1981); and fourth, the increased value of a house can serve as self-insurance. Using data for the US and the UK, Ansell (2014; also see André et al., 2018) shows that during the previous two decades house price increases indeed reduce support for social security measures. Additionally, Ansell (2014) also investigates whether income from stocks and investment shows the same relationship with welfare state attitudes. However, in this analysis his methodology is inconclusive, as either the dependent variable is a very broad measure of social security (whether the government should implement a full-employment policy), or the identification of financial income is binary – that is, whether the respondent has any stocks or investments. As will be discussed in more detail below, the investigation presented here omits these shortcomings by looking at a group of people that are for the largest share dependent on the financial market for their income.
Within the second category of micro analyses, several investigations have shown that being directly involved in the financial market through the ownership of financial assets can profoundly influence political attitudes. Kaustia, Knüpfer and Torstila (2016) show for example that increased stock ownership due to demutualisation of Danish telecom services increased the vote share of right-to-centre parties in areas subject to such a demutualisation by 1.7 to 2.7 percent. The authors suggest that such an effect cannot be attributed to the increase in wealth generated by the demutualisation, as the effect was absent among citizens who, also due to the demutualisation, became owners of similar but non-tradable assets. In other words, the change in voting choice cannot be explained by the direct material gains that might result from this vote. Instead, the authors suggest that financial asset ownership makes voters expose themselves to different sources of information, which in turn changes their political attitudes.
9 They make this claim plausible by also showing that owners of financial assets are more likely to read media oriented at financial markets.
An experimental investigation of the effect of financial market exposure among Israeli voters by Jha and Shayo (2018) points in the same direction and gives a more specific assessment of the mechanism causing the effect that Kaustia et al. (2016) find. In the study by Jha and Shayo, a treatment group of Israeli citizens were given financial assets worth up to either 50 or 100 dollars and were encouraged to actively engage in trading. This treatment substantially increased the propensity to vote for parties advocating peace in Israel’s conflict with Palestine. Notably, if these effects would occur across the entire population, they would be substantial enough to tilt the election results in favour of parties advocating peace. In addition, and similar to Kaustia et al. (2016), Jha and Shayo (2018) show that this change in behaviour is not necessarily driven by a material incentive, as treated individuals that had been divested before the election show an equally strong preference for a peaceful outcome. Instead, the authors suggest that increased sociotropic concern with the economy should be considered as the main driver, a suggestion which is buttressed by the increased awareness of financial market dynamics and performance among respondents in the treatment group, the convincement among those respondents that peace will have positive effects on the economy in general and their increased likeliness to rank the economy as the most important election issue.
In sum, these prior investigations clearly show that being exposed to the financial market influences one’s political attitudes. As such, being dependent on the financial market for income is likely to reduce support for national social security measures for two reasons. First, social security measures have a negative effect on financial market performance by either increasing inflation or taxation, and thereby on people who are dependent on financial assets for their income (Scharpf, 1987; Mosley, 2000; Mosley and Singer, 2008). Second, these people are unlikely to benefit from these measures, as generally they are targeted at the unemployed, the sick and the old, and because the store of financial assets possessed by these people can serve as a private insurance for economic hardship (Ansell, 2014; Rehm, 2009). This leads to the following hypothesis:
Hypothesis 1: National social security measures are less likely to be supported by people who
are more dependent on the financial market for their household income than by people who are less dependent on the financial market for their household income.
10
The European level
In this section, it will be discussed whether we can expect a relationship between financial market dependence and attitudes toward supranational social security measures as well. First, a brief review of the literature investigating attitudes toward the EU and European welfare institutions is in order. Remarkably, research on support for European integration, broadly conceptualised, shows that explanations based on the economic self-interest of individuals generally hold little bite. Instead, cultural values such as identification with one’s nation (Hooghe and Marks, 2004), ideology, altruïsm and cosmopolitanism (Bechtel, Hainmueller and Margalit, 2014; Kuhn, Solaz and Elsas, 2017) are shown to be key determinants.
Similar dynamics showed themselves in attitudes toward fiscal assistance to debt-ridden countries during the euro crisis, as these cultural values generally were the most important causes of support for the financial bailouts among European citizens (Bechtel et al., 2014; ibid., 2017; Kleider and Stoeckel, 2016; Lengfeld, Schmidt and Häuberer, 2015). However, citizens with a low socioeconomic position partly form an exception, as among these citizens a left-wing ideology relates to opposition toward the bailouts, in contrast to the support expressed by the left-wing citizens with a higher socioeconomic position (Kleider and Stoeckel, 2016). Moreover, and closer to the topic of interest here, Bechtel, Hainmueller and Margalit (2014) hypothesize that possessing financial assets increases support for the European financial bailouts among German citizens. Although the effect is small, this notably is the only hypothesis along the lines of economic self-interest that is supported by their analysis. In a later contribution, these authors (ibid., 2017) also show that stockowners oppose higher burden sharing by the German public more than non-stockowners.
However, attitudes toward supranational welfare institutions are different in nature than attitudes toward European integration in general or the financial bailouts: citizens differentiate between dimensions of the European Union, such as European citizenship, member-state solidarity and interpersonal solidarity (Baute et al., 2018b), and have different attitudes toward policies related to these dimensions (Baute et al., 2018a). Nonetheless, even support for far reaching European social policies tends to be substantial, and at least higher than the measures directed at indebted countries during the euro crisis (Genschel and Hemerijck, 2018; Lahusen and Grasso, 2018). For example, a majority of German, Polish and Spanish citizens support a uniform social welfare system (Gerhards, Lengfeld and Häuberer, 2016).1 While the same
cultural values as discussed above still seem to form the most important sources of support, an
11 investigation of attitudes toward a European Unemployment Benefit Scheme (similar to the EUBS investigated here) and a Sovereign Insolvency Procedure (SIP) by Dolls and Werfhörer (2018) found that in addition to these cultural values, income levels – especially high income levels – have a substantial impact on opposition (in the case of an EUBS) and support (in the case of an SIP) for these measures.
As such, at first glance the literature on attitudes toward European integration and policies suggests that egoistic economic motivations will do a poor job at explaining attitudes toward European policy. However, these general expectations sometimes do not hold for individuals with particular socioeconomic positions. More specifically, people with a particularly high (Dolls and Werfhörer, 2018) or low income (Kleider and Stoeckel, 2016), and people that are engaged in the financial market (Bechtel, Hainmueller and Margalit, 2014; 2017), seem to form an exception.
Especially the exception formed by the latter group motivates the expectation that financial market dependence influences one’s attitude toward supranational social security measures, even if socioeconomic factors are generally found to be less relevant. Drawing on our knowledge of the relationship between such a socioeconomic position and attitudes toward national welfare state institutions, this leads us to expect opposition toward supranational social security institutions among people who are dependent on the financial market for income.
This relationship is again likely for two reasons. Regardless of the particular form in which such policies, such as an EUBS, would be implemented, they will cause an increase in government expenditure used to support consumption by the unemployed. As such, social security measures are essentially a demand side policy. Therefore, it is likely that they will increase inflation, government debt, or a combination of those (Boix, 1998; Scharpf, 1987), and thereby cause opposition among people who are dependent on the financial market. Second, most social security measures, but an EUBS in particular, are targeted at citizens with labour as their primary source of their income. Therefore, these policies do not have the goal of reducing the risk exposure of citizens deriving (part of) their income from financial instruments (Rehm, 2009). Conversely, as owners of assets they will have to pay taxes for such a scheme. Moreover, their asset ownership is also already functioning as a form of self-insurance (Ansell, 2014). In short, while the introduction of supranational social security measures is likely to cause inflation, and thereby incur negative effects on the economic position of people dependent on the financial market, these people also do not benefit from these measures in any direct way while they do have to pay for it. This leads to the following hypothesis:
12
Hypothesis 2: European level social security measures are less likely to be supported by
people who are more dependent on the financial market for their household income than by people who are less dependent on the financial market for their household income.
In addition, the mechanisms proposed above also allow us to hypothesise how elements of an EUBS, as a specific EU-level social security policy, are related to the level of opposition. A companion study to this paper (Vandenbroucke et al., 2018) has shown that generally, the level of generosity of an EUBS is the main source of support for such a scheme. However, it can be expected that among people who are dependent on the financial market for income, this level of generosity has the opposite effect. This opposition can take the form of relative opposition, meaning that financially dependent people are less supportive of a more generous scheme compared to people with a regular source of income, or absolute opposition, meaning that while generosity relates to support among people with a regular source of income, it relates to opposition among financially dependent people.
This opposition can be expected to be caused through two different mechanisms. The first mechanism operates through a lack of benefits for financially dependent people. This mechanism can be expected to result in indifference regarding the level of generosity among this group. In other words, as financially dependent people are unlikely to benefit from an EUBS, the level of support among these people remains the same at different levels of generosity of the scheme. As generally an increase in generosity increases the support for an EUBS among people with a regular source of income, this entails that financially dependent people show less support relative to people with a regular source of income for a scheme that is more generous. More specifically, while this mechanism predicts relative opposition, it does not predict absolute opposition. This leads to the following hypothesis:
Hypothesis 3a: The higher the generosity of a European Unemployment Benefit Scheme, the
less likely people who are more dependent on the financial market for their household income are to support this scheme than are non-financially dependent people.
The second mechanism operates through the indirect costs of the scheme for financially dependent people. This can be expected to result in absolute opposition among this group. More specifically, this mechanism entails that higher generosity levels make the scheme more prone to inflation, endangering the value of financial assets. Moreover, financially dependent people are likely to shoulder some of the costs related to this higher level of generosity. From this
13 perspective, it can be expected that financially dependent people oppose an EUBS more the more generous it is. The difference with the first mechanism is that in this case opposition is not only relative to people with a regular source of income, but that financially dependent people are also opposed to an EUBS in absolute terms. In other words, this mechanism predicts that an EUBS with a higher generosity level is less likely to be supported by financially dependent people than a less generous EUBS. This leads to the following hypothesis:
Hypothesis 3b: The higher the generosity of a European Unemployment Benefit Scheme, the
less likely people who are more dependent on the financial market for their household income are to support this scheme.
However, at this point it should be noted that the line of reasoning explained above does not consider European-level policies as different in nature than their national counterparts. The political economy literature that informs the relationship between financial market dependence and support for national welfare state institutions suggests a negative relationship between the two, and the research so far conducted on support for (the prospect of) EU counterparts of such institutions tells us that it is plausible that financial market dependence is one specific socioeconomic position that holds some bite in explaining opposition toward supranational welfare state institutions. Nonetheless, there are good reasons to expect that the dynamics causing support or opposition toward European Union-level policies are different from those that cause support or opposition toward national-level policies. More specifically, while international social security measures are of little interest to owners of financial assets in a direct way, they are of interest to them in an indirect way. The incoherence of fiscal policies of EU member states in combination with monetary union currently creates economic imbalances which make the member states prone to severe economic shocks (Stiglitz, 2016; De Grauwe and Yi, 2017). The Euro crisis has demonstrated the serious consequences that such shocks can have for the financial market and therefore the economies of all EU member states. The international insurance of unemployment benefits is highly likely to reduce and smoothen the impact of economic shocks (Dolls, Fuest, Neumann and Peichl, 2018). Obviously, national social security measures have a similar stabilising effect on most markets, but under the conditions of tight international integration, especially of financial markets, national social security measures do little to reduce volatility of, and thereby the risks stemming from, being dependent on these financial markets. Conversely, supranational security measures are likely
14 to reduce this volatility, and in that sense they reduce the risk exposure of the owners of financial instruments as well.
In order for these macroeconomic dynamics to actually reduce opposition toward an EUBS among people who are dependent on the financial market for income, these people would have to understand the relationship between these dynamics and their personal economic position, and act upon this understanding. Yet the investigations by Kaustia, Knüpfer and Torstila (2016) and Jha and Shayo (2018) show that this is plausible. Even minor increases in financial assets already make people more aware of how policies affect the economy. In addition, Bechtel et al. (2014) found that stock ownership among German citizens actually increased support for the financial bailouts, even if such fiscal expenditures could be costly for stock owners. Moreover, stock owners were not more sensitive to a haircut for private investors as a part of such a bailout package than other people: even a haircut of 50% had an effect of virtually zero on opposition toward the bailouts among stockholders (ibid., 2017). While it might of course be argued that sovereign default would be even worse for these people, these findings nonetheless point in the direction that the owners of financial assets consider the relation between their own position and economic policy with more insight than a simple egoïstic perspective suggests. In other words, it seems to be the case that individuals exposed to the financial market take the effects of policies on macroeconomic dynamics into account when considering how these policies will ultimately affect them.
Therefore, supranational social security measures might be appealing to people that are dependent on the financial market for their income. This makes schemes such as an EUBS different from national welfare state policies and leads to the expectation that these people are not as opposed to supranational social security relative to national social security. In other words, if confronted with the choice between for example an EUBS and a national form of social security, people that are dependent on the financial market are expected to choose an EUBS. This leads to the following hypothesis:
Hypothesis 4: Supranational social security measures are more likely to be supported relative
to national social security measures by people who are dependent on the financial market for their household income.
15
Data and measurement
These expectations will be tested using two datasets. The first stems from a survey specifically designed to measure support for EUBS policy proposals by means of a conjoint experiment. This allows for a detailed investigation of not only to what extent people would support such a measure, but also what characteristics of such a policy increase or decrease this support. This survey has been conducted in 13 European countries (see below), in each of which a representative random sample of approximately 1500 respondents has been drawn. These countries have been selected in order to represent variation within the EU on a variety of dimensions, most notably economic development, welfare state type, economic performance over the last decennium and EMU membership. The fieldwork has been conducted by means of the online panes of survey company IPSOS during October and December 2018. The survey was accessed on personal computers and mobile devices, and translated into the countries’ main language(s). Quotas for the following demographic categories have been used in order to ensure the representativeness of each country sample: age, gender, education and region. This representativeness has also been validated by comparisons with national demographic data. For most countries, the discrepancy between any of the demographic categories in the national data and the sample was less than 2 percent, and no more than 4 percent for any of the categories in any of the countries, with the single exception of low-educated individuals in Hungary. Remaining discrepancies are resolved with weights, which are applied in all the analyses presented below.
Additionally, round 8 of the European Social Survey (ESS) will be used. This is the only other publicly available dataset that allows financial market dependence and attitudes to both national and European-level social security measures to be operationalised. While it does not allow an as detailed analysis of such measures as the EUBS dataset does, it does cover more countries and generally draws larger samples within those countries.
Independent variables
Ideally, financial market dependence would have been operationalised using the total income from financial assets. However, to the author’s knowledge there are no datasets available that both include such a measure and allow the operationalisation of attitudes toward European social security institutions. Therefore, financial market dependence will be operationalised by identifying respondents for whom income from financial assets is the main source of income in their household. This can either be ‘income from investments, saving, other financial
16 instruments’ or ‘income from rents and property.’2 In the ESS, the same question is asked, but
these two categories are a priori combined into the same answer.3
This operationalisation is for several reasons suitable for the purposes of this research. First, in comparison to measuring whether a respondent owns any financial assets, as for example Bechtel et al. (2014) do, it allows for a clearer estimation of the effect of financial market dependence as only respondents who are for the largest share of their income dependent on the financial market will be included. The downside of this strength is that the variable is still a crude measure of financial market dependence. A second reason why this operationalisation is suitable for the purposes of this research, is that by considering the income of the entire household, respondents who are indirectly dependent on the financial market through the income of their partner are also included, while the reverse is the case for respondents who are less dependent on the financial market due to their partner’s alternative income source. This is a preferable characteristic, as income sources from one’s partner might reduce a respondent’s actual dependence on the financial market.
For the EUBS data, respondents who have investments and financial instruments as their main source of income, and respondents who have rent and property as their main source of income, will be combined into one group. This combination is warranted as both have a source of income that makes them less likely to need, and be eligible for, social security (Ansell, 2014). Moreover, both groups of respondents are subjected and therefore dependent on the same financial market dynamics. As the financial crisis of 2007 has shown, there is a strong link between financial and housing markets. Therefore, financial market dynamics also expose property owners to the same type of risk exposure as the owners of financial instruments. As such, we can expect the interests of both groups of people to align in the sense that economic stability is in their interest. Nonetheless, whether the results presented below are different when these groups are included separately will be discussed in the robustness section.
Table 1 shows the distribution of respondents that are dependent on the financial market for income per country, in percentages per dataset. In both datasets, the group is relatively small: in the EUBS dataset there are 234 financially dependent respondents, in the ESS dataset there are 222 financially dependent respondents. This low figure in the ESS data relative to the EUBS data is remarkable, especially since the sample size of the latter is smaller. Part of this difference
2 “Please consider the income of all household members and any income which may be received by the
household as a whole. What is the main source of income in your household?”
3 The summary statistics in table A.1 and A.2 in the appendix provide descriptive information about this
17 can however be explained by the country composition of both datasets: the EUBS includes almost all of Europe’s most economically advanced countries, while the ESS contains relatively more countries that are less economically advanced. In any case, as the number of respondents within this category is rather low, loss of cases within this group, for example due to missing values on variables, is prevented as much as possible throughout the analyses.
As the EUBS contains disaggregated information regarding the income sources of investments and financial instruments on the one hand, and property and rent on the other, we can compare each of these subgroups per country. Generally, the first category (financial instruments and investment) is somewhat larger than the second (rent and property), although for Germany and Italy this figure is reversed. Hungary has an exceptionally low number of respondents who are dependent on the financial market.
Table 1. Distribution of financially dependent respondents by country
Country Financial instruments Rent and property Total (EUBS) Total (ESS) Country Financial instruments Rent and property Total (EUBS) Total (ESS) Austria 0.72 0.39 1.12 0.20 Ireland 1.52 1.06 2.58 0.33 Belgium 0.46 0.20 0.66 0.68 Lithuania . . . 0.28 Czechia . . . 0.18 The Netherlands 0.86 0.07 0.93 0.89 Denmark 0.66 0.20 0.86 . Poland 0.80 0.40 1.20 0.18 Estonia 0.53 0.07 0.60 0.30 Portugal . . . 0.71 Finland 0.99 0.79 1.78 1.35 Slovenia . . . 0.15 France 0.27 0.66 0.93 0.43 Spain 0.73 0.53 1.26 0.41 Germany 0.53 0.86 1.39 0.84 Sweden . . . 0.32 Hungary 0.20 0.13 0.33 0.06 United Kingdom . . . 1.89 Italy 0.67 1.20 1.86 0.88 Complete sample 0.69 0.50 1.19 0.57
Generally, the size of the combined subgroup is per country relatively similar in both datasets. Some notable exceptions are Austria (a difference of 13 respondents), Ireland (30) and Poland (15).
In addition, statistics describing how the demographic characteristics of people dependent on the financial market compare to those of other people show some noteworthy features. First, those dependent on the financial market tend to have a somewhat higher education level than others. Second, only a small part of the people dependent on the financial market is mainly involved in paid work, and instead they are more often unemployed, or involved in education,
18 community or military service, or housework, relative to respondents with a regular source of income.. Third, there seems to be a propensity for those dependent on the financial market for income to be at either end of the income distribution, or, in other words, to have either a low or a high income (i.e. being in one of the lower or upper three deciles of the income distribution). A more detailed investigation shows that this pattern is to a large extent structured along the lines of the different main activities: those with a below-average income are more likely to be in education, community service, housework, to be unemployed or sick or disabled, while those with an above-average income are more likely to be involved in paid work, or to be retired. Figures visualising these statistics can be found in the appendix (figures A.1 to A.4). These figures show that these characteristics are reassuringly similar across both databases, suggesting that this pattern is not the result of bias in the sampling process.
Lastly, it should be noted that the vast majority of financially dependent respondents is not employed in the financial sector. The ESS dataset includes information regarding the occupation of respondents and their partners according to the ISCO coding standard. This information shows that out of the 222 respondents in the ESS data who are for their household income dependent on the financial market, only eight are themselves, or have a partner that is, employed in the financial sector. This strongly suggests that most of the respondents identified here as dependent on the financial market are operating on a private basis.
Dependent variables
The first dependent variable, support for national social security, will for both datasets be operationalised with the following question:
“On a score of 0-10, how much responsibility do you think governments should have to ensure a reasonable standard of living for the unemployed? 0 means it should not be governments’ responsibility at all and 10 means it should be entirely governments’ responsibility.”
This question has the desirable property that it focusses on unemployment, while it is ambiguous concerning the means through which such a reasonable standard should be achieved. As such, it taps one of the core elements on national social security provision. The analysis of this dependent variable will be conducted with an ordered logistic regression.
The second set of hypotheses, concerning support for supranational social security, is operationalised mainly with the EUBS dataset. At the core the EUBS survey is a conjoint
19 experiment. Such an experiment is used to analyse stated preferences by asking respondents to choose between and/or rate their preference for two profiles. Each profile differs on a number of attributes, or dimensions. By randomly drawing the actual content of these dimensions shown to respondents from a prepared set, one can estimate the average effect of each of these dimension values on the rating or rank of the profile separately (Hainmueller, Hopkins and Yamamoto, 2014).
In the EUBS data, the profiles are policy packages, and each package has six dimensions which can show different characteristics. The ‘replacement rate’ dimension shows different values of the European contribution to the national unemployment benefit expenditures, at either 40, 60 or 70 percent of the last wage, covering the first 6 months of unemployment. In terms of the practical implications of the policy, this dimension would entail both a common floor for the level of unemployment benefits in all participating countries and the amount of European subsidisation to national unemployment benefits. The ‘training and education’ dimension shows whether there are conditions that countries in need must fulfil to obtain the support, with either ‘no conditions’ or the condition that ‘A country can only participate if it offers education and training opportunities for all its unemployed citizens.’ The ‘taxation dimension’ shows the long-term impact on the taxes that people will have to pay for their unemployment insurance; with either no long-run impact; a 0.5 percent increase in taxes for everyone in the respondents’ country; or a 1 percent increase for only the rich in the respondents’ country. The ‘administration’ dimension shows who will administer the program, either displaying ‘The European Union’ or ‘National governments.’ The ‘eligibility criteria’ dimension shows whether there are conditions for unemployed people in participating countries, and can display that the unemployed must apply for at least one job per week, and accept any suitable job offer, or lose the benefit; that the unemployed must accept any suitable job offer or lose the benefit; or that there are no conditions. Lastly, the ‘cross-border redistribution’ dimension shows whether some countries may receive more support from the programme than they pay in, displaying that in the long run countries cannot receive more support from the programme than they paid into the programme; that in the long run countries can receive more support from the programme than they paid into the programme; or that in the long run poor countries will receive more support from the programme than they paid into it, while rich countries will receive less support from the programme than they paid pay into it. The different dimension values are completely randomised across iterations. In addition, the dimension order is randomised across respondents, with the exception of training and education dimension and the between-country redistribution dimension, which always occur together and
20 in that order. Due to this randomisation, the conjoint experiment allows a fair claim of causality regarding the relationship between the policy characteristics and support for a policy package (Hainmueller et al., 2014). An overview of the dimensions of the conjoint experiment can be found in table A.3 in the appendix.
The experiment was introduced with a neutral framing that explains the purpose of the policy.4 This framing explicitly stresses the macroeconomic goals of the policy, namely the stabilisation of the European economies. Subsequently, respondents were asked to rate three pairs of policy packages, resulting in a total of six policy packages rated per respondent. They were asked to respond to each paring in two different ways. First, respondents were asked to choose between the two packages presented to them, resulting in binary choice information.5 Second, respondents were asked to rate each package on a 5-point scale ranging from strongly in favour to strongly against, resulting in an absolute level rating.6 This latter variable can be
used as an ordinal scale variable, but can also be dichotomised. The analyses presented below will use a dichotomisation of the rank variable, with neutral values included as opposition. However, robustness of the results across the other operationalisations will also be discussed.
In order to test hypothesis 2, a count measure of the number of policy packages supported will be created using the ordinal measure described above, and accordingly an ordered logistic regression will be used to analyse this dependent variable. Notably, this measure does not take the package characteristics into account. As the package characteristics are endogenous to this independent variable, this introduces some noise. However, it can safely be assumed that due to the randomisation of these package characteristics, this noise is to a large extent cancelled out on the aggregate level. Furthermore, robustness of the results over model specifications that do take the package characteristics into account is discussed in the robustness section.
Hypothesis 2 will also be tested with round 8 of the ESS data. Here, the question is asked to what extent respondents support an EU-wide social benefit scheme which would guarantee a minimum standard of living for all poor people in the EU and which would be financed more by the richer EU countries than by the poor.7 This variable is answered on a 4-point scale ranging from ‘strongly against’ to ‘strongly in favour,’ and will therefore be analysed with an ordered logistic regression. Admittedly, this operationalisation is less ideal than the operationalisation in the EUBS dataset, for at least three reasons. First, in contrast to an EUBS,
4 The text of the frame can be found in the appendix.
5 “Which one of the following two options for this European programme do you prefer?” 6 “How much are you in favour or against option 1[/2]?”
21 the policy proposed in the question is far removed from what is currently feasible on the European level, both economically and politically. Second, the proposal is ambiguous about its primary goal and target, i.e. how large the basic income has to be to ensure a minimum standard of living and how poor people have to be in order to be eligible for the policy. This might cause differences in interpretation to drive results instead of fundamental opinions. Third, because this scheme would be for all poor people, and not just the unemployed, this operationalisation is based on a broader conceptualisation of supranational social security than the operationalisation for the EUBS dataset. These characteristics are likely to cause a divergence between the results of the analysis for both datasets, although it is uncertain in which direction this divergence will be.
In order to test hypotheses 3a and 3b, the influence of the policy design on support for the EUBS (whether a package has been rated with ‘in favour’ or ‘strongly in favour’) is taken into account by including a dummy for the different dimension values (excluding a base level for each dimension). The sensitivity of people deriving the main source of their income from financial instruments, property and investment to the generosity of the scheme will be assessed by interacting financial market dependence with the dimension values of interest. Specifically, two of the EUBS dimensions in the experiment can be considered as determining its generosity. The most direct dimension is the level to which the policy would insure national unemployment benefits. However, the strictness of eligibility criteria can also be considered as determining the scheme’s generosity, as stricter eligibility criteria are likely to reduce the number of European citizens eligible for benefits.
According to both hypothesis 3a and 3b, a negative interaction between financial dependence and higher insured replacement rates and looser eligibility criteria is expected. However, according to hypothesis 3a, the net difference between financially dependent respondents and respondents who draw their income from a regular source is expected to be zero, cancelling out the dimension’s effect on support for an EUBS among financially dependent respondents. Conversely, according to hypothesis 3b, the net difference between both groups is expected to be negative the higher the replacement rate and the looser the eligibility criteria are, meaning that the scheme is more opposed by financially dependent respondents if it has these characteristics.
It might be argued that an increase in taxation imposed by the scheme could also relate to stronger opposition among financially dependent people. We can expect people dependent on the financial market for income to be averse to increases in taxation resulting from an EUBS, as an increase in taxation has a direct impact on their own earnings. However, in this regard
22 they would be similar to other people, and as such it would be ambiguous whether this preference could be ascribed to their dependence on the financial market. Moreover, the influence of the costs of the scheme on the preferences of this group of people through its effect on macroeconomic performance is also ambiguous. On the one hand, an increase in taxation might negatively influence financial markets due to feared negative effects on the competitive position of Europe’s economy. On the other hand, the taxation level is of less concern to financial market actors than inflation (Mosley and Singer, 2008), and these actors in fact prefer taxation over debt-financed spending (Akitoby and Stratmann, 2008).
Lastly, hypothesis 4 will be operationalised by testing the difference found between the relationships postulated under hypothesis 1 and 2 for both datasets. More specifically, it will be investigated whether the difference between support for national and supranational social security measures is larger (and positive) among those who are dependent on the financial market for their household income than among those who are not. This will be investigated by first combining the models testing hypothesis 1 and 2 into a seemingly unrelated regression. This technique allows an estimation of the correlation of individual residuals across estimations, taking into account that an individual’s level of support for national social security could be related to that same individual’s level of support for supranational social security. This technique is by some considered to be correct and unbiased if these residuals are indeed correlated (Srivastava and Giles, 1987). Thereafter, the same models are combined into a bivariate ordered probit regression, which does not allow an exact estimation of the correlation of the individual residuals, but does take into account the ordered nature of the dependent variables.8 This latter model will be used to compare the effects of financial dependence across the models, but robustness across other specifications is also reported.
Control Variables
The analyses include a set of control variables throughout. First, standard demographic variables such as age, gender and education level (8-level ISCED coding) are controlled for.
Income will be included to measure wealth effects as this has been shown to influence attitudes
toward social security (Meltzer and Richard, 1981; Rehm, 2009; Ansell, 2014), and might subsequently confound the relationship investigated here because financial market dependence
8 A probit model is used in this case because, to the author’s knowledge, algorithms for bivariate ordered logistic
regression have not been written for conventional statistics packages. It should also be noted that while the bivariate ordered probit regression cannot calculate the exact correlation of individual residuals, it does take this correlation into account in its estimations.
23 might be positively correlated with income levels. Note, however, that this confounding relationship cannot be in the opposite direction, as in this case the nature of the main independent variable excludes the possibility that financial instrument ownership is itself the result of a high income. Income is originally measured on the household level and subsequently equivalised according to the household size and composition.9
Moreover, binary variables indicating three types of main activities are included in the analyses: (1) being in education, (2) unemployed or (3) retired. The reference category for these variables are respondents who have a paid job, are sick or disabled, are performing community or military service or housework. The first and third category indicate that a respondent is not in the labour market, and therefore it is likely that these respondents are less interested in social security. Contrarily, respondents who are currently unemployed are likely to be more supportive of social security. In line with this expectation, a binary variable indicating recently unemployed (during the previous five years) is also included in the analyses, as this is likely to have a positive effect on support for social security measures (Hacker, Rehm and Schlesinger, 2013). Only respondents who are currently employed are included in this variable in order to prevent overlap with the unemployed variable.
Furthermore, ideology will for the EUBS data be controlled for as one’s position toward business regulation by the government, coded on a 5-point scale ranging from strongly in favour (1) to strongly against (5). For the ESS data, this is operationalised as a respondent’s self-placement on the left-right scale ranging from 0 (left) to 10 (right). Moreover, the party family of the party voted for by the respondent during the previous elections is included as well. The categories for this variable are drawn from the CHES 2014 dataset.10 Parties are categorised into left, liberal, conservative, Christian democratic and miscellaneous parties. Radical right-wing parties and nationalist parties are included in the conservative category. Moreover, respondents who did not vote during the previous elections are included as a separate category as well to prevent unnecessary loss of data. Ideology and partisanship might confound the relationship between financial market dependence and support for (supra)national social security measures because this dependence moves political attitudes to the right (Kaustia, Knüpfer and Torstilla, 2016) and because people with a left-wing attitude are less likely to own financial assets in the first place (Kaustia and Torstila, 2011).
The analyses of support for supranational social security will also include support for
redistribution to take into account that support for supranational social security might depend
9 For this equivalisation the adjusted OECD scale is used. 10 See Polk et al., 2017.
24 on a generic preference for governmental redistribution, regardless of the level on which it is organised. For the EUBS data, this is operationalised as one’s position toward income equalisation by the government, which is coded on a 5-point scale ranging from strongly in favour (1) to strongly against (5). For the ESS data, this is operationalised by whether the respondent thinks that for a fair society, differences in living standards should be small, the values of which range from agree strongly (1) to disagree strongly (5).
Moreover, for all of the analyses concerning forms of social security organised at the European level (supranational social security), one’s position toward Europe is controlled for as well. This concept is operationalised with two variables: (1) exclusive national identity, and (2) the respondent’s position toward EU membership. The operationalisation of these variables differs slightly between the two datasets: for the EUBS data, exclusive national identity is coded as 1 when respondents only see themselves as related to their national identity, instead of any mix between national and European identity, while for the ESS data, this variable is coded as 1 when an emotional attachment to the EU of 0, on a scale from 0 to 10, is reported.11 Position
toward EU membership is for the EUBS data operationalised as whether the respondent thinks that European membership is a ‘good thing’ or a ‘bad thing’, with neutral respondents included in the ‘bad thing’ category. For the ESS data this variable is operationalised according to whether the respondent would vote to remain in the EU or to leave.
Lastly, the EUBS survey also included an attention check at the end of the survey, which in a series of items asked respondents to choose the value of 7 for the item ‘focus.’ In addition, the conjoint experiment allows a second strategy for identifying inattentiveness among respondents, as it is possible to provide inconsistent answers during this part of the survey. That is, as respondents had to choose one package, but rank both packages separately, it is possible for them to rate the chosen package lower than the non-chosen package. It is important to take possible inattentiveness into account because the conjoint experiment confronts respondents with a relatively complicated task, and therefore inattentiveness might introduce substantial bias in the results. For this reason, a measure has been created which indicates attentiveness when a respondent passed the attention check and did not provide more than one inconsistent answer during the conjoint experiment. The analyses using a dependent variable based on the conjoint experiment are run with and without inattentive and inconsistent respondents. It should however be noted that this exclusion forms a substantial sample cut (20.5 percent of respondents) which therefore also reduces the efficiency of the estimations. This is especially
11 There are no divergences in the results presented below when a less conservative cut-off point (2) is chosen for
25 problematic considering the number of financially dependent respondents. Therefore, analyses using a dependent variable which is not based on the conjoint experiment include attentiveness as a control variable, and deviation in the results after exclusion of inattentive respondents is reported where relevant.
Results
In this section, the empirical results of the analysis are discussed in terms of their formal meaning. A more substantive discussion, in which these results will be interpreted in light of the theoretical framework of the analysis, will follow later.
Support for national and supranational social security
Table 2 shows the results of the ordered logistic regressions of support for national social security and supranational social security on financial dependence and the control variables. For models 1 and 2 the dependent variable is the extent to which the government should provide a reasonable standard of living for the unemployed, using the EUBS and ESS data respectively. This table, as well as the tables that are to follow, do not show the country level fixed effects. First, the control variables for model 1 will be discussed. These have effects in the expected directions, which are mostly significant: the proportional odds of being in a higher category of support for national social security versus a lower category decrease as education or income increase and as one’s ideological position (or the party that one intends to vote for) moves to the right, while these proportional odds increase with age. Moreover, respondents who are currently in education or who are retired (nonsignificant) are on average less supportive of national social security than those who have a paid job, are sick or disabled, are performing community or military service or housework (the reference category). The reverse is the case for men (compared to women, nonsignificant) and respondents who are currently unemployed or who have been unemployed in their recent past. The party families are significantly associated with support for national social security: in comparison to respondents who voted for parties in the ‘miscellaneous’ category, respondents voting for left parties are more supportive of national social security, while respondents voting for liberal, conservative and Christian democratic parties (insignificant) are less supportive. In addition, respondents who did not vote are more supportive, although this effect is not significant. Lastly, attentive respondents, who passed the attention check and did not provide more than one inconsistent
