Who Blames the Poor?: Multilevel evidence of support for and determinants of individualistic explanation of poverty in Europe

The article analyses the support for and determinants of individualistic explanations of poverty among Europeans. The analyses are conducted using multilevel logistic regression models. Findings suggest that the level of support for the individualistic explanation of poverty varies between European nations. Hence, the results contradict the analysis of Alesina and Glaeser (2004, Fighting Poverty in the US and Europe. A World of Difference), which gives us a very homogenous picture of European attitudes towards the poor. Among the determinants of poverty perceptions, welfare regime, short-term economic growth and social expenditures as well as individual-level demographic factors, perceived economic hardship, political affiliation and egalitarian values are associated with the individualistic attribution for poverty. In general, the findings hold critical implications for future research to develop further dynamic measurement of contextual level explanatory factors.

While cross-national differences in attitudes towards the role of government and income redistribution have long held a prominent place in the social science literature, surprisingly little is known about cross-national differences in public perceptions of the causes of poverty. Yet, the importance of public attitudes towards the poor for welfare policies is widely acknowledged. From a wider perspective attitudes towards the poor can be seen as a part of the moral economy of the welfare state (Mau 2003). Attributions for poverty also have implications for the viability of and support for anti-poverty policies (Kluegel and Smith 1986; Bullock et al. 2003). Hence, notions relating to the reasons for poverty are closely linked to the discussion about deservingness (van Oorschot 2006), which means that an examination of the public explanations of poverty contributes to an understanding of the popular cultural context of welfare rationing.

In their influential book Fighting Poverty in the US and Europe Alesina and Glaeser (2004) suggest that one of the reasons behind the transatlantic differences in redistributive policies – the ‘American exceptionalism’ (Lipset 1996) – is the difference between the world views of Americans and Europeans. They demonstrate that Americans are more likely than Europeans to believe that they live in a land of opportunity where anyone who stays poor is too lazy to pull themselves up by their own bootstraps. Thus, they refer to the dominant ideology thesis (Huber and Form 1973) that undergirds the American stratification ideology in which the principle of equal opportunities and belief in personal responsibility for one's social fate are fundamental elements. Accordingly, dominant ideology legitimises inequalities and predicts that individualistic beliefs about the causes of poverty should predominate.

Alesina and Glaeser emphasise that attributions for poverty have an effect on the level of welfare spending and reflect deeper cultural and ideological forces that have ensured the relatively greater strength of the right in the United States and the left in Europe. However, their analysis regarding the attributions for poverty is limited. They focus only on the mean values reported for the European Union (and the United States) in the 1983–1997 World Value Surveys, and the range of countries selected for analyses is very limited (Alesina and Glaeser 2004: 183–88). However, individual European nations differ not only in regard to their cultural and historical legacies but also their welfare policies, a fact that has led to distinctive patterns of policy outcomes, for instance in terms of social inequalities and the incidence of poverty. Thus, we argue that their analyses give a too simplistic and too homogeneous picture of the Europeans.

There is therefore a need for a detailed analysis of European attitudes towards the poor. The aim of the study is to analyse the variation between countries and welfare regimes in terms of perceptions of the individualistic cause of poverty and to explore contextual and individual-level determinants of such perceptions. The article also makes some methodological contributions. First, while the functionality of multilevel modelling has been emphasised in the general literature on comparative welfare state attitudes (Gelissen 2001; Blekesaune and Quadagno 2003; Mau and Burkhardt 2009; Dallinger 2010), only a very small proportion of prior studies on attributions for poverty have applied multilevel techniques in order to examine systematically the role of individual and country-level explanatory factors as determinants of poverty attributions (see Brady 2009). This study, however, applies multilevel logistic regression analysis and focuses on both individual and country-level characteristics. Second, while earlier studies have concentrated only on cross-sectional country-level circumstances, there is greater theoretical justification for examining how changes in country-level circumstances affect public opinion. Therefore, the article examines the effects of context variables by focusing on the level of and changes in the country-level factors. The article thus makes a methodological contribution by applying multilevel techniques with a more ‘dynamic’ approach with respect to country-level explanatory factors.

Pioneering studies on lay explanations of poverty by Feagin (1972, 1975) categorised reasons for poverty a priori into three basic categories: (1) individualistic reasons which emphasised the behaviour of the poor; (2) societal or structural reasons which focused on external societal and economic factors; and (3) fatalistic reasons which placed responsibility on luck and fate. Later, many factor analytic studies have given empirical support to Feagin's categorisation (Feather 1974; Lepianka 2007; Niemelä 2008).

Prior cross-national studies have focused on attributions for poverty from different points of view, and have therefore distinguished between types of explanations for different purposes (van Oorschot and Halman 2000; Gallie and Paugam 2002; Lepianka 2007; Lepianka et al. 2010). This study applies Albrekt Larsen's (2006) approach by focusing only on the ‘individual blame’ type of explanation for poverty. Based on a combination of the welfare regime theory and the theories of deservingness criteria, he established the theory that the institutional structure of each welfare regime frames the way the public perceives the poor and the unemployed. As a proxy of dimensions deduced from the deservingness literature he focused on the individualistic explanation of poverty, which ‘refers to a situation that the poor are in control of their neediness’ and argues that ‘the poor are persons with deviant characteristics compared to “ordinary” citizens’ (Albrekt Larsen 2006: 67). Earlier findings have demonstrated that the individualistic explanation varies the most between the Western capitalist countries (van Oorschot and Halman 2000; Lepianka et al. 2010).

The literature on welfare attitudes has shown that different patterns in attitudes can be identified between different welfare state regimes and welfare traditions (Taylor-Gooby 1985; Svallfors 1997; Andress and Heien 2001). Following Albrekt Larsen's (2006: 47–56) line of reasoning concerning the institutional logic of welfare attitudes, it can be argued that in the marginal welfare-policy-dominated liberal regime, the poor will be asked to fulfil much harder deservingness criteria than in a universal social democratic regime (also van Oorschot 2006). In the Anglo-Saxon countries, in which welfare programmes are more marginal and less generous, explanations of poverty are more often based on individual morality, reflecting the strength of the dominant individualistic ideology, than they are in the Continental Western European and Nordic welfare states – where external and structural explanations prevail (also Alesina and Glaeser 2004; Lepianka 2007). Previous findings have not found clear differences between the Mediterranean and the Continental Western European countries (van Oorschot and Halman 2000).

Comparative studies have also emphasised differences between Western and Eastern Europe. West Europeans are more inclined to support structural explanations for poverty than their East European counterparts. On the one hand, the ‘individual blame’ type of explanation is generally more popular in Eastern Europe than in the West. In fact, studies on post-socialists countries show that individual blame coincides strongly with social blame, which illustrates the split-consciousness of public perceptions. On the other hand, fatalistic reasons receive the lowest support in Eastern European perceptions (Kreidl 2000; van Oorschot and Halman 2000; Lepianka 2007; Lepianka et al. 2010). Consequently, we hypothesise that cross-national differences in poverty attitudes can to some extent be explained by the welfare tradition and by distinctive historical and cultural legacies.

There are only a few studies on attributions for poverty in which country-level explanatory factors are included in the analysis (Gallie and Paugam 2002; Albrekt Larsen 2006; Lepianka 2007; Lepianka et al. 2010). However, some assumptions about the importance of country-level variables can be made based on the general comparative welfare attitude literature.

While the country groupings refer not only to welfare state characteristics but also to cultural and historical legacies, the level of social expenditures refers more directly to the size of the welfare state. In addition, while the welfare regime is a rather constant measure of welfare state characteristics, changes in social expenditures allow us also to test the extent to which changes in welfare state characteristics are associated with poverty attributions. Prior findings indicate that the level of social expenditures has a negative effect on public support for the individualistic explanation of poverty (Albrekt Larsen 2006: 74–5). Accordingly, the hypothesis is as follows:

We assume that, besides the welfare state characteristics, wider economic and social conditions are associated with welfare state legitimacy and attributions for poverty. Increased financial strain is associated with stronger support for state responsibility for economic provision (Blekesaune 2007). In regard to explanations of poverty, there was a marked decline of support for the individualistic explanation in the majority of European countries between 1989 and 1993 as economic conditions deteriorated, followed by an increase between 1993 and 2001 when economic conditions improved (Gallie and Paugam 2002: 21–24). Thus, it seems that in years of economic trouble people blame more likely structural factors beyond an individual agency than individuals themselves. Related to economic conditions, during periods of high unemployment, people express more positive attitudes towards welfare policies, in particular those aimed at the unemployed (Blekesaune and Quadagno 2003; Blekesaune 2007; Pfeifer 2009). Studies on public perceptions of the causes of poverty have shown that people living in countries with high unemployment are more likely to endorse social than individualistic explanations (Albrekt Larsen 2006: 74–79; Lepianka 2007: 118–21). Similar findings have been reported concerning the incidence of poverty (Lepianka et al. 2010). The increased visibility of poverty at the country level reduces the likelihood of choosing individual laziness as an explanation of poverty (Lepianka 2007: 118–21). Economic downturns can lead to a high level of unemployment and poverty, which can affect public attitudes towards the poor through several mechanisms (Blekesaune and Quadagno 2003: 418). Increased economic insecurity at the national level can increase awareness of the risk of unemployment and poverty. Moreover, in circumstances of high unemployment and poverty, poverty tends to rise to a more prominent position on the public and political agenda (Kuivalainen and Niemelä 2010), which, in turn, can increase the tendency for poverty to be explained by reference to external rather than internal factors. Finally, rising unemployment or poverty increases the probability of individual experience of material destitution and exposure to the poor via increasing formal and informal contacts with the poor. This mechanism appears to be an individual-level process and will be discussed in greater detail in the next section.

Besides economic determinants, ethnic fractionalisation, too, has been claimed to have an effect on welfare attitudes. The argument is that increasing immigration and ethnic diversity weaken solidarity and drive reductions in social spending especially if welfare use is associated with a subordinate minority (Alesina and Glaeser 2004). This assumption is not entirely supported by empirical studies. While empirical findings do indicate that immigration reduces public support for welfare state redistribution and social welfare spending, this association is not particularly strong (Mau and Burkhardt 2009; Eger 2010). In addition, differences in immigration rates between European countries have no influence on relative solidarity towards migrants (van Oorschot 2008). Finally, in regard to attributions for poverty, Albrekt Larsen (2006: 74–75) found that ethnic fractionalisation does not have an impact on public perceptions of the causes of poverty. Thus, the hypothesis is as follows:

Still, proposing strong hypotheses as to the influence of immigration on poverty attributions is not easy. Very little has been done to single out migration/ethnicity-related characteristics that could account for cross-country differences in attributions for poverty. In addition, it is a limitation of previous studies concerning welfare attitudes that they represent a static approach to study the effect of country-level characteristics. They measure, for instance, the incidence of poverty, unemployment, the size of the economy or immigration in one cross-section. However, theoretically it is more justified to explore how changes in country-level characteristics are associated with public opinion, because it is reasonable to assume that history influences people's attitudes. This is justified not only in the case of migration/ethnicity-related issues but also in regard to country-level economic and social explanatory factors in general.

Most of the previous studies on individual-level determinants of poverty perceptions have focused on individuals’ socio-demographic characteristics. These have included such determinants as gender, age, race/ethnicity, education and occupational status (e.g., Bullock 1999; Cozzarelli et al. 2001; Niemelä 2008). Apart from race and ethnicity in the United States (Gilens 1999), results have shown that socio-demographic variables explain very little and that they do not reveal clear or consistent patterns. They can, however, explain attributions for poverty indirectly by shaping people's perception of their interest in various beliefs. We agree with Lepianka (2007: 35) that equivocal results and the indirect nature of the effects of socio-demographic variables make them particularly difficult to test and it is difficult to form any solid hypotheses as to whether they are associated with perceptions of the causes of poverty.

From a self-interest perspective, there are studies that support the assumption that those who perceive themselves to be poor are more likely to agree with the external cause of poverty than people well above the poverty line (Saunders 2002: 155–56; Niemelä 2008).

Yet, there are also contradictory findings indicating that personal experience of poverty might lead to individualistic explanations. For instance, claimants of public welfare seem to hold negative views about other recipients (Bullock 1999). In addition, differences between income groups have shown that the individualistic explanation of poverty is not at its most prevalent among the highest income groups, but rather among people with average incomes (Feagin 1972).

In the previous section we assumed that the incidence of poverty at the country level is associated with attributions for poverty. The individual-level mechanism behind this assumption is that rising poverty increases not only the probability of individual experience of poverty, but also the awareness of poverty and exposure to the poor. Increasing visibility of poverty at country and/or community level as well as increasing media depictions and informal contacts increase awareness of poverty and the mechanisms behind it. Prior findings have proven that these factors are important predictors of attributions for poverty (Lee et al. 1992, 2004; Lepianka 2007). For example, the visibility of homelessness at the community level increases the likelihood of people placing the blame for poverty on society rather than the individual (Lee et al. 1992). Thus, we can hypothesise as follows:

Another line of research on attributions for poverty has emphasised the role of other beliefs, attitudes and values over that of self-interest. The role of political affiliation/ideology, in particular, has been identified as an important determinant of perceptions of poverty. Political conservatives or those who place themselves on the right of the political spectrum are more likely to endorse individual attributions of poverty, while political liberals or those on the left of the political spectrum attribute poverty to structural forces (Kluegel and Smith 1986; Zucker and Weiner 1993; Cozzarelli et al. 2001).

Finally, attitudes towards the welfare state and social equality have been proved to be linked to attributions for poverty. Pro-welfare attitudes are positively associated with structural reasons and negatively correlated with individualistic perceptions of the causes of poverty (Kluegel and Smith 1986; Niemelä 2008). In a similar vein, structuralism is positively correlated with social legitimacy, whereas individualism is positively correlated with perceived dishonesty (Bullock 1999). Hence, in regard to values we can hypothesise:

The purpose of the article is to analyse to what extent Europeans support the individualistic explanation of poverty. In regard to determinants of public explanations of poverty, the study will analyse the effects of individual- and country-level characteristics. From the methodological point of view, the analysis will give special emphasis to country-level factors. The article explores how much the results differ when the roles of country-level economic and social factors are measured against changes in, rather than the cross-sectional level of, country-level variables.

The study utilises the Eurobarometer 67.1 survey from 2007 comprising aggregate data at the country level. The analyses cover the population of 26 European countries (residents aged 15 years or over). Hence, they comprise all of the countries in the survey except Bulgaria and Croatia. They are excluded from the analysis due to missing individual-level information and lack of reliable country-level data. The basic sample design applied in all countries is that of a multi-stage, random sample. In each country, a number of sampling points were drawn with probability proportional to population size and to population density (for a more detailed description of data, see European Commission 2007).

The respondents were asked the following question measuring public perceptions of the causes of poverty: Why in your opinion are there people who live in need? Here are six options: which is closest to yours?

1. Because they have been unlucky

2. Because of laziness and lack of willpower

3. Because there is too much injustice in our society

4. It is an inevitable part of progress

5. None of these

6. Don't know

Because the focus of this article is to analyse the individual blame explanation, the dependent variable was constructed as follows: (1) Because of laziness and lack of willpower, (0) other explanations. The ‘don't know’ category was treated as missing data.

According to the theory of institutional logic of welfare attitudes (Albrekt Larsen 2006) the institutional structure of each welfare regime influences or frames the way the public perceives the poor. Thus, the examined countries are grouped according to the traditional division between English-speaking, Continental Western European, Mediterranean and Nordic countries (Bonoli 1997; Gelissen 2001). In addition, while we acknowledge that the East-central European countries do not form a specific type of welfare regime (Cerami and Vanhuysse 2009), prior findings on attributions for poverty have emphasised the difference between the East and West of Europe. Therefore, in order to analyse our hypothesis (h1a), this article treats transition countries as one group (also Dallinger 2010; Lepianka et al. 2010; van Oorschot 2006). Hence, the examined countries are grouped as follows: (1) English-speaking (Great Britain and Ireland), (2) Nordic (Denmark, Finland and Sweden), (3) Continental Western European (Austria, Belgium, Germany, France, Luxembourg and The Netherlands), (4) Mediterranean (Spain, Greece, Italy, Portugal, Cyprus and Malta), and (5) East-central European (Czech Republic, Estonia, Hungary, Lithuania, Latvia, Poland, Slovenia, Slovakia and Romania).

The description of other independent variables is summarised in Appendix 1. At the individual level, socio-demographic characteristics (gender, age, socio-economic status), perceived economic hardship, awareness of poverty at the community level, political affiliation and the social values held by individuals are analysed. With regard to country-level independent variables, the analysis will be executed separately for independent trend and cross-section variables. The independent variables are real GDP growth rate, incidence of poverty, unemployment rate, share of total expenditure on social protection in GDP, and as a measure of ethnic heterogeneity, population share of non-nationals in 2007. In order to analyse changes in country-level variables during the period between 2000 and 2007, slopes of the regression line were estimated from each variable.

The empirical section starts by presenting the results of descriptive analyses. The main method used in this study is multilevel logistic regression analysis. Logistic regression is used in situations where the dependent variable is based on two values. Multilevel analysis allows the inclusion of independent variables at two levels (for details see Hox 2002). The analysis is performed by means of the SAS Proc Glimmix procedure and all of the models employed are so-called random intercept models at the level of individuals and countries.

Due to the fact that with only 26 countries, the number of cases at the contextual level is rather small the multilevel analysis starts with a bivariate analysis where the model contains only one independent variable at a time. Based on the bivariate analysis, those independent variables which are statistically significantly associated with the dependent variable are selected for use in the multilevel models. The multilevel models will be expressed as two tables (the first with independent country-level trend variables and the second with cross-sectional country-level independent variables). The analysis will be carried out as follows: (1) empty model, (2) model with individual-level independent variables, (3) models with individual-level independent variables and one country-level variable at a time, and (4) the last model, which will include both individual and country-level variables at the same time. The tables will include both a fixed effects and a random effects part. The fixed effects part will include odds ratios and statistical significances of independent variables, while the random effects part will indicate the variance between countries and how it changes between different models.

Figure 1 shows the proportion of people in each country who support the idea that poverty is caused by laziness and lack of willpower among the poor. The results support the assumption that there is a clear difference between the East and West of Continental Europe (h1a). Also the expectation that support for individual laziness as a reason for poverty is lowest in the Nordic regime is corroborated (h1b). However, against the assumption, citizens of the Mediterranean regime are more willing to support the individualistic explanation than citizens of the English-speaking nations.

The results also show that there is some variation within the welfare regimes. The variation is greatest in the Mediterranean regime (SD=7.43). While the individualistic explanation is supported strongly in Malta, Portugal and Cyprus, the support is weak in Greece, Italy and especially in Spain. Substantial variation is also found in East-Central Europe (SD=6.66), in which similarly weak support for individualistic explanations is found in Hungary and Romania. Also the two English-speaking countries (SD=4.50), Ireland and Great Britain, differ from each other. In fact, support for the individualistic explanation is lower in Ireland than for example in Finland, Belgium or Austria. Thus, the Irish figures do not fit particularly well within the welfare regime hypothesis. In the Nordic regime (SD=3.74), Sweden has the lowest support for the individualistic explanation in Europe whereas the Finnish figures are more in line with the Continental Western European countries and actually higher than for instance in The Netherlands, France or Germany. The lowest variation within a single regime is among the Continental Western European countries (SD=3.37). Yet, support for the individualistic explanation is very low especially in The Netherlands.

To sum up, the results indicate that the support for the individualistic explanation varies extensively between the European countries. With some exceptions Europe can be divided into two groups of countries. One group consists of the low-individual-blame Nordic and Continental European countries, while the other group is comprised of the high-individual-blame East Central European and Mediterranean countries. The most notable exceptions here are Italy and Spain as well as Hungary and Romania, where the results are more in line with the Continental Western European regime. In addition, while Ireland belongs to the relatively low-individual-blame group, Great Britain is clearly a high-individual-blame country.

In order to explore which factors explain the cross-national differences, multilevel analysis starts with a bivariate analysis (not reported here). It indicates that the individual-level independent variables are associated with attributions for poverty. On the other hand, many of the country-level factors are not connected to perceptions. Contrary to hypothesis h4, differences in attributions for poverty are not explained by the incidence of poverty or by the unemployment rate. In addition, the analysis gives support to hypothesis h5, which posits that there is no association between the population share of non-nationals and the individualistic attribution for poverty. These variables will therefore not be included in the further analysis.

The left hand side of Table 1 indicates that there are differences between European countries in terms of perceptions of the causes of poverty. The empty model shows that the variance between countries is 0.22 (SE=0.06). Model 1 indicates that individual-level variables do not explain between-country variance. On the contrary, they increase between-country variance from 0.22 to 0.27 (SE=0.08). This makes us confident that country-level independent variables are important for discovering more about cross-national differences. Model 1 shows that men are more likely than women to put the blame on the poor themselves. It also reveals that middle-aged persons are less inclined to support the individualistic explanation than younger or older age groups. Support for the individualistic explanation is strongest in the youngest age group.

In line with the self-interest hypothesis h6, those who have economic problems are less inclined to support the individualistic explanation than those who have never experienced financial problems. Socio-economic status shows that the highest support for individual-blame is among manual workers. Also other studies have shown that workers tend to be more morally rigid than other social classes (Svallfors 2006).

Contrary to the results of bivariate analysis and hypothesis h7, the awareness of poverty proves not to be a differentiator of individualistic attribution for poverty. On the other hand, and as hypothesised (h8), political affiliation and social values are associated with the individualistic explanation. Those who place themselves to the left or at the centre on a 10-point scale are less inclined to blame the poor than those who place themselves to the right. Furthermore those who do not know or do not want to reveal their political affiliation are less inclined to support the individualistic explanation than those on the right. Moreover, those who see equity and solidarity as important social values are less likely than others to blame the poor themselves.

Models 2, 3 and 4 indicate that membership in a particular welfare regime and changes in economic growth are related to support for the individualistic explanation. Model 2 shows what happens when welfare regime is included in the analysis: between-country variance decreases from 0.22 to 0.14 (SE=0.04). In line with descriptive statistics, support for the individualistic explanation is highest in the Mediterranean and East-central European countries.

Model 3 concerns the effect of changes in country-level economic conditions. When economic growth is included in the analysis, between-country variance is 0.20 (SE=0.06). The result confirms hypothesis h3, which predicts that when economic growth increases, the tendency to support the individualistic explanation increases as well. While the bivariate analysis indicated that trends in social expenditures are associated with attributions for poverty, model 4 shows that the association is not statistically significant. Thus, in regard to trends in social expenditures, hypothesis h2 is not supported.

Finally, Model 5 shows the results obtained when all independent individual and country-level variables are included in the analysis. Country-level independent variables do not explain all of the variance between European countries. Still, they decrease the variance markedly. Between-country variance decreases from 0.22 in the empty model to 0.11 in the full model (SE=0.04). Because of the correlation between contextual variables, only welfare regime remains as a statistically significant contextual variable when all country-level variables are included in the analysis. In the English-speaking, Mediterranean and East-central European regimes, people tend to blame the poor more than in the Nordic countries.

Table 2 includes three new models with cross-sectional, country-level independent variables. In line with the results of long-term economic growth, model 6 indicates that the more rapid the rate of short-term economic growth is, the more likely people are to blame the poor themselves (h3). Interestingly enough, and conversely to the results of the trends in social expenditures (Model 4), Model 7 shows that the size of the welfare state is associated with attributions for poverty: the greater the volume of social expenditure is, the less likely people are to believe in the individualistic explanation (h2).

The final model pulls together all the cross-sectional independent variables. It indicates that welfare regime is not a differentiator of attitudes (only the Mediterranean regime differs statistically significantly from the Nordic regime) when short-term economic growth and social expenditures are included in the analysis. Furthermore, short-term economic growth and social expenditures are also not associated with the individualistic explanation of poverty to a statistically significant degree. Hence, this leads to the assumption that differences between welfare regimes are a consequence of a nation's economic growth, the size of its welfare state and its stage of development.

The analysis of the article provided a detailed picture of the attributions for poverty in Europe and highlighted significant differences between European nations as well as provided important empirical findings on the individual and country-level determinants of public perceptions. The results contradict the analysis of Alesina and Glaeser (2004), which gives us a very homogeneous picture of European attitudes towards the poor. There are not only differences between the West and East of Europe, but also between the South and North. On the one hand, citizens of the Nordic and Continental European countries are less inclined than others to support the individualistic explanation. On the other hand, citizens of the Mediterranean countries and of East-central Europe are most likely to put the blame on the poor themselves. The English-speaking countries lie between these two positions. Yet, the analyses also showed quite large variation within regimes. For instance, the Irish blame the poor considerably less than the Britons.

Second, the findings show that the role of individual-level variables is surprisingly strong compared to previous empirical results. They indicate that socio-demographic factors, perceived economic hardship, political affiliation and egalitarian values are associated with the individualistic attribution for poverty.

Third, most of the country-level economic and social independent variables do not explain the support for the individualistic explanation. However, the results confirmed that welfare regime, country-level economic growth and social expenditures are associated with the individualistic attribution for poverty. Overall, the findings suggest that attitudinal differences between welfare regimes are partly a result of the economic conditions and of the size of the welfare state.

Fourth, from a methodological point of view, the study introduced a more ‘dynamic approach’ to the measurement of country-level explanatory factors by examining how changes in country-level variables are connected to attributions for poverty. On the one hand, the analysis showed that regardless of the method of measurement economic conditions are linked to the perceptions. When measuring long-term changes in the real GDP growth rate results were similar to those obtained for short-term economic growth: under situations of fast short-term economic growth, support for the individualistic perception was generally stronger than in conditions of slower growth.

Alternatively, analyses regarding social expenditures showed that while the cross-sectional level of social expenditures is negatively associated with support for the individualistic explanation, there is no statistically significant association between the 7-year trend of social expenditures and the individualistic attribution for poverty. In general, the finding holds critical implications for future research. Attempts should be made to develop further the theoretical and empirical basis of the dynamic approach, for instance by comparing the effects of long-term and short-term changes. In regard to social expenditures, we could for example assume that the 7-year period is too short to bring out true changes in welfare state characteristics.

In addition, a specific limitation of the analysis in this article is that the outcome variable used in the analysis is measured in levels, not in changes. Thus, future research should try to develop a true dynamic analysis which would include changes in country-level contextual variables as well as in outcome variables.

Finally, analyses clearly showed that support for the individualistic explanation is strongest in the East-central European countries. From the ‘worlds of welfare’ perspective, post-socialist countries do not form a specific type of welfare regime (Cerami and Vanhuysse 2009), but they share a common communist legacy and experienced a similar transition period after the collapse of the Communist regime. While the findings of the study showed clear differences between the West and East of Europe, they also indicated considerable differences within the East-central European family. Thus, the results call for more analysis of the mechanisms behind the East-central European attitudes.

This research is part of the corresponding author's research project funded by the Academy of Finland (grant number: 121779). Previous versions of the paper were presented at several international conferences and scientific meetings. The authors would like to thank all participants of those meetings. Special thanks go to Olli Kangas and Michaela Pfeifer for their valuable comments.

Johanna Kallio, PhD in Social Policy, is a post-doctoral researcher in the Swedish School of Social Science at the University of Helsinki. Her research focuses on various aspects of welfare attitudes. She has published several articles on public and decision-maker attitudes towards the public services. Her current research focuses on social workers’ attitudes towards welfare and on cross-national differences in public attitudes towards the role of government.

Mikko Niemelä, Adjunct Professor in Sociology, is a senior researcher at the Social Insurance Institution of Finland. His current research focuses on public perceptions of the causes of poverty and on cross-national differences in public attitudes towards the role of government. Other areas of research include the measurement of poverty, the intergenerational transmission of poverty and the mechanisms of the institutional change of welfare states.