Exploring the something for
nothing syndrome
Confused citizens or free riders? Evidence from Sweden
Jonas Edlund
Ingemar Johansson Sevä
Working paper 2/2012
Department of Sociology
901 87 Umeå, Sweden
Telephone: 090-786 50 00
www.umu.se
Abstract The objective of this article is to study the occurrence of political non‐attitudes in Sweden and make an attempt to identify the segment of citizens possessing a something for nothing mentality when it comes to social spending. Sweden – by many scholars regarded as the epitome of the welfare state – is used as an analytically interesting case of contrast to the available predominantly US based evidence on the subject. We argue that the impact of national political institutions on the prevalence – and social base – of the something for nothing segment is fundamental. The strategy applied to identify something for nothing sentiments is to use two different question batteries measuring social spending preferences. One battery contains “priced” items emphasizing the cost of public spending, while the other set of “unpriced” items do not explicitly connect an increase in public spending with tax raises. We first explore the patterning of attitudes across these items, thereby answering the principal question of whether the presence of non‐attitudes in general, and the something for nothing syndrome in particular, is common in the Swedish electorate. The observed attitude patterns are then linked to broader sets of attitudes related to the welfare state thereby testing whether the degree of ideological coherence is fundamentally different in the something for nothing segment. The analysis then explores the individual‐level determinants associated with different attitude patterns. The results show that political non‐attitudes and something for nothing sentiments are not dominant in the Swedish citizenry. Our analysis indicates that the last segment does not easily fit into the non‐attitudes concept, the quality of the something for nothing segment’s belief system, in terms of ideological coherence, does not represent a deviant case. Our conclusion is to view them as free riders in economically vulnerable positions, rather than ignorant and politically confused citizens. 1 Introduction In most contemporary Western democracies, political struggles over taxation and social spending priorities dominate domestic politics. Indeed, knowledge about the major political parties' preferences in these matters provides us with key information when defining the political landscape in a country. The prominent position of the welfare state in national politics is understandable. In the Scandinavian countries, the tax revenue is at present as large as nearly half of the gross domestic product. Even in less ambitious welfare states such as the USA and Canada, the total tax revenue constitutes about a third of the GDP. Needless to say, in all western countries, the lion's share of the tax revenue is used for financing social insurance programs and social services. The impact of the welfare state on the everyday life of citizens is substantial but varies in effect across social groups; some collect more social benefits than others; tax levels differ between income groups; and some social strata are more dependent on the services of the welfare state than others, just to mention a few areas where social cleavages may translate into conflicting political preferences. Making a judgement based on these premises, it is tempting to assume that the welfare state should occupy an equally prominent position in the political minds of citizens as it does for political parties. After all, rational voters should be interested in both knowing and trying to influence how tax money are collected and spent. However, this image of a typical voter does not fare well with a pile of empirical evidence ‐ mostly from the USA ‐ which suggests that citizens in general possess a rather low level of political sophistication, broadly understood. These works form an integral body of evidence in a 2 longstanding and hotly contested debate about the quality of mass belief systems – the non‐
attitudes debate – which began some fifty years ago with the seminal work of Converse (1964).1 Within survey research the two dominant criteria for analyzing the presence of non‐attitudes are: (a) the stability of attitudes over time; and (b) the structure and organization of political attitudes and, occasionally, their relationships to some a priori defined ideology.2 In this paper, the focus is on a key issue that belongs to the last criterion: the something for nothing syndrome – that there often exists a discrepancy between what citizens want and what they are willing to pay for. Although some controversy still exists, current scholarship largely agrees that the general level of political sophistication among voters is rather low. Research – still dominated by US scholars – capitalizing on these findings seems to proceed along three main avenues. Scholars on the first avenue are indebted to the theoretical heritage of Converse and aim at integrating the empirical findings into a general model of public opinion and the underlying mechanisms of attitude formation. The explanatory focus is biased towards the individual level where demographic as well as psychological predispositions are emphasized (Lewis 1982; Saris & Sniderman 2004; Zaller 1992; cf. Kinder 1983). Those travelling on the second avenue are critical towards the Converse paradigm and present a substantially more positive view on the ability of ordinary citizens to make sense of the political world, despite their lack of political sophistication (Popkin 1991; Stimson 1991; Inglehart 1990: 104‐129). According to this perspective, the scholarly emphasis on studying the political incompetence of voters has propelled a view of the American public as being less politically sophisticated than it actually is 3 (Claggett & Shafer 2010). The third avenue scholars take an intermediate position. While they agree, by and large, that the level of political sophistication is not even close to the classical ideal of informed democratic citizenship, it is argued that a “key innovation” (Brooks & Manza 2007: 147) in the study of policy‐public opinion linkages is a move away from the individual level to the aggregate level: “social aggregation” (Page & Shapiro 1992). The collective body of public opinion is supposed to behave rational and understandable (Page & Shapiro 1993; Wlezien 1995; Soroka & Wlezien 2005). Considering the amount of research devoted to map and explain qualities in mass belief systems, it is both surprising and unfortunate that research on the subject have stayed within the borders of the USA. We believe that the field would benefit from a broader perspective, where the feedback‐effects from national institutions on mass attitudes are taken into account (Mettler & Soss 2004; Svallfors 2007). This paper offers such an approach to the study of mass belief systems; that the socio‐political institutional context within which citizens are embedded may have profound effects on citizens’ understanding, interest, and awareness of politics. Assessing the impact of national political institutions on the prevalence of the something for nothing syndrome and its social base would ideally require a thorough cross country comparative design. Lacking comparative data, the chosen research design is more modest. Sweden – by many scholars regarded as the epitome of the encompassing welfare state – is used as an analytically interesting case of contrast to the available predominantly liberal welfare state (i.e. USA) based evidence on the subject. Even in the light of this small‐scale design we hope to make two contributions: to describe a 4 theoretical approach for explaining cross‐national differences in the quality of mass belief systems, and applying the perspective empirically on a key subject – the something for nothing (SFN) syndrome – in a new, fundamentally different national context. The study is based on nationally‐representative data of the adult Swedish population and uses two sets of specially designed item batteries on preferences for public spending across a range of social policy areas. The objective is to study the occurrence of non‐attitudes in Sweden and make an attempt to identify the segment of citizens possessing an SFN mentality. By investigating the relationship between social spending preferences and views on taxation we identify a number of attitude clusters. These segments vary in terms of coherence between what they want and what they are willing to pay for. In addition to identifying the SFN segment our objective is to get an informed view of this specific segment. Is the something for nothing segment mainly a group of ill‐informed voters with incoherent attitudes or do they possess the same qualities in terms of coherence as other groups of citizens supporting the welfare state? The paper is organized as follows. First, previous US findings on the quality of mass belief systems are reviewed, followed by a section discussing alternative, more rational, reasons for holding SFN preferences. Thereafter, arguments on the role of political institutions for influencing the quality of mass belief systems, including the prevalence of the SFN syndrome, are presented. Next, scattered Swedish empirical evidence are reviewed and discussed in relation to the US findings. This is followed by a description of the data, methods, and measurements applied. The empirical part is divided into three sections. First, the patterning of social spending preferences is explored. Finding a constellation of different social spending 5 preferences – where some qualify as possible candidates of non‐attitudes – the second section tests whether the degree of welfare state ideological consistency differ across the distinguished social spending preference constellations by linking them to broader sets of attitudes related to the welfare state. The third section explores the social base of each social spending preferences constellation. The main objective of sections 2 and 3 is to further our understanding of whether SFN sentiments should be regarded as a form of non‐attitudes or if there might be other, more rational, motives underlying the SFN preference. The paper ends with a summary of the main findings and a concluding discussion. Public opinion on taxing and spending – a short review of American evidence Some fifty years ago, after a careful analysis of the American citizenry, Converse (1964: 213), initiated the non‐attitudes debate by arguing that the ill‐informed, fragmented, and unstable political attitudes that he had observed "are not a pathology limited to a thin and disoriented bottom layer of the lumpenproletariat; they are immediately relevant in understanding the bulk of mass political behaviour." Some thirty years later, Bartels (1996: 194), drawing on a literature review (Sniderman 1993), summarized the situation of the field and argued that "the political ignorance of the American voter is one of the best‐documented features of contemporary politics (...)." As noted above, the concept of social aggregation has been brought forward as a fruitful tool adjusting for incoherent attitudinal structures on the individual level. But even in the light of this modification, there are still some unresolved issues. Apart from the critique delivered by Kinder 6 and Herzog (1993) and Bartels (1996), a particularly troublesome and persistent empirical finding to account for is the widespread lack of understanding of tax policy and the basic functions of the tax system and its connections to social spending that have been revealed by previous research. In fact, these are the areas where the American public seems to have tremendous difficulties to possess informed and well‐defined political preferences. In The Psychology of Taxation, Lewis (1982: 71) resignedly concludes, “we have to accept that attitudes towards taxation and public expenditure are based on something less than perfect knowledge on the part of taxpayers. More realistically, attitudes should be examined for what they are ‐ a product of myth and misperception.” Bartles (2005: 21) goes even further: “”public opinion” about tax policy is a very fragile concept (...) to the extent that it exists at all (...).” Other scholars are quite pessimistic as well. Relying on collected evidence (Lewis 1978; Citrin 1979; Beedle & Taylor‐Gooby 1983), Steinmo (1998: xi) states that “those who have carefully studied the public's attitudes, perceptions and knowledge of taxes and tax policy have generally found that citizens are indeed remarkably misinformed and/or confused.” The results reported by Roberts et al. (1994) travel in the same direction: ignorance about key tax concepts is widespread. Respondents favour the principle of progressive taxation but reject it in concrete applications. In a thorough analysis of the 1979 tax reform in California, Sears and Citrin (1985: 260) note that there is a “striking invariance in the disjunction between opinions about taxing and spending. The desire to get something for nothing is well entrenched among a substantial portion of the citizenry in all segments of society. There remain tiny extremes whose general attitudes about 7 the role of government are tightly bound to reactions to one specific effort to harness the public sector. But for most people the connections are loose.” Depending on how strictly the borderlines are defined, the SFN segment constituted between 50 percent and 80 percent of the population.3 Bartels (2005), analyzing public attitudes towards the tax reforms under the Bush administration in 2001 and 2003, found that support for these highly regressive tax reforms was positively associated with support for increased taxes on high income earners and increased social spending. Support for the reforms was negatively associated with personal income. Commenting on these rather exotic patterns, he notes that “survey results as these make it clear that most ordinary citizens are remarkably ignorant and uncertain about the workings of the tax system and the policy options under consideration, or actually adopted, in Washington” (Bartels 2005: 21). In relation to the policy‐public opinion linkage, Steinmo (1993: 195), rhetorically asks, “if citizens are poorly informed about taxes – even their own taxes – how can they be expected to provide guidance to political leaders? The answer, I submit, is that citizens provide little guidance indeed.” The main message from previous US research on taxing and spending is in the spirit of Converse: American citizens are generally ill‐informed, insensitive to tax policy implications on welfare programs, and, to paraphrase Bartels (2005: 21), “oddly disconnected from (or misconnected to) a variety of relevant values and material interests.” In relation to the SFN syndrome, this particular segment of citizens is usually interpreted as politically confused and ignorant citizens. 8 Something for nothing – confusion or rationality? The occurrence of SFN sentiments may, however, be explained by other factors than ignorance. An alternative, less widespread, explanation is that citizens are generally rational and simply assume that someone else will pay the extra taxes needed for financing the increase in social spending (Hadenius 1986: 123; Key 1961). For those still believing that citizens are reasonably well‐informed and understand complex policy issues, this argument provides a comfortable retreat.4 From a methodological as well as a substantive perspective, it is clear that the larger the share of citizens holding the SFN attitude, the harder to correctly estimate the true political support or opposition for increased social spending. Moreover, whether citizens holding the SFN mentality are in reality confused voters or have rational reasons for their position may carry different implications on political representation and support. From a non‐attitudes perspective the SFN segment would be interpreted as a volatile proportion of the electorate whose voting preferences may be particularly sensitive to the dynamics of political campaign articulations. Those having rational reasons, on the other hand, are not as likely to be open to political persuasion or as sensitive to the type of political rhetoric applied as confused voters. It is however important to acknowledge that rational SFN preferences can be driven by at least two different ideal‐typical motives/reasons (Corneo & Grüner 2002; Alesina & Giuliano 2010). The first type is the standard version in economics of the free rider ‐ someone who wants to enjoy the fruits of increased social spending but chooses not to participate, i.e., to pay the 9 additional taxes that is needed. Someone driven by the other type of motives/reasons may have the same preference for increased social spending but economic constraints prevent participation, e.g., low income earners and other groups with scarce economic resources. While both types of motives/reasons involve some form of free riding, they differ in two substantial respects. First, the option of a true choice to participate or not is open to the standard version free rider only. Second, when it comes to moral justifications for these two types of free riding, the standard version free rider tends in public discourse to be associated with selfishness and uncooperative behaviour ‐ sometimes even associated with anti‐social behaviour which is morally rebukable. Needless to say, no dominant political ideology celebrates the behaviour of the standard version of the free rider. In contrast, the second ideal‐typical motives/reasons are backed up by a left‐movement ideology articulating messages emphasizing that welfare state redistribution ‐ i.e., that those better of in the stratification system should contribute disproportionally more compared to those worse off ‐ is a fruitful avenue to achieve distributive justice.5 Unfortunately, the data set at our disposal does not contain items covering welfare state redistribution, e.g., attitudes towards the current levels and distributions of different taxes. It is therefore not possible to assess the role of distributive preferences in the analysis. However, by examining the impact of socio‐economic variables on cluster membership, we may get some clues helping us to decide which of the two ideal‐types that resonates best with the observed relationships. Finding that SFN sentiments are over‐represented in groups with limited 10 resources, e.g., low income earners and single parents, would fit the economic vulnerability version rather than the standard economic version. Conversely, empirical evidence suggesting that the SFN syndrome is under‐represented in low‐resource groups would speak in favour of the standard economic understanding of the free rider. On the role of political institutions for attitude formation and previous Swedish evidence As previously noted, the non‐attitudes debate originated in the USA, and still continues to be a salient but rather isolated US‐phenomenon. In Sweden the issue about public non‐attitudes has never been particularly salient, and accordingly there are not many studies explicitly examining the prevalence of non‐attitudes in the public. However, from a comparative perspective there are reasons to believe that the occurrence of non‐attitudes towards taxes and spending is considerably lower in Sweden compared to the USA. Borrowing insights from the historical institutionalism school (Rothstein & Steinmo 2002; Thelén & Steinmo 1992; Rothstein 1998) and the power resources approach (Korpi & Palme 1998, 2003; Korpi 1980, 1983, 2006), and research on institutional feedback effects on mass attitudes (Mettler & Soss 2004; Svallfors 2006, 2007) the argument is that cross‐national variation in non‐attitudes may fruitfully be sought among factors related to the welfare state and the rhetoric of associated organized interests – political parties, unions, and employer organizations – surrounding issues of taxation and redistributive welfare policies. The influence of political institutions is thus fundamental in our attempt to explore non‐attitudes and the SFN syndrome. We define political institutions as “the formal rules, compliance procedures, and standard operating practices that structure the 11 relationship between individuals in various units of the polity and economy” (Hall 1992: 96), or simply put: “the rules of the game” (North 1990: 3–5). Examples of institutions therefore include social security systems, tax systems, election systems and family law. More concretely, the argument takes into account the prominent levels of taxes and social spending in Sweden, and the clear grouping or organized interests along the social class axis – factors that are all conducive to the salience of redistributive politics in public and political debates. Strategies pursued by business interests and right‐wing parties aiming at cutting taxes and decreasing redistribution have routinely been challenged by left movement counter‐
arguments defending the current state of affairs by arguing the necessity of taxes to secure the quality of welfare policies and to achieve redistributive goals. Compared to the Swedish polarized environment of prosecutors and defenders, the milieu surrounding tax politics in the USA appears as less antagonistic, since virtually all organized interests tend to act as prosecutors, and when it comes to social spending rhetoric, “politicians do their best to spread the illusion of costless progress” (Sears & Citrin 1985: 259; Peters 1991; Steinmo 1993; Lipset 1996). A closely related factor of importance is the extent to which the welfare state can be considered to have a major influence on citizens’ everyday life (Kriesi 2004). Owing to the residual character of the American welfare state, experiences at the grassroots level of public welfare policies are likely to be more common among Swedish citizens than among the American citizenry. Finally, the tax system is often conceived of as very technical and difficult to grasp (Peters 1991). However, in comparison with the fairly standardized system in Sweden, the American tax 12 system appears to be substantially more complex and particularistic (Steinmo 1993). It is therefore plausible that the higher transparency of the Swedish tax system facilitates a more accurate public understanding of the functioning of the tax system. In general, scattered empirical evidence does not collide with the above suggestions. Milner (2002), applying a comparative perspective, found that countries with welfare states equalizing material as well as intellectual resources have a higher level of politically well‐informed citizens. In a comprehensive comparative study of political attitudes, Granberg and Holmberg (1988) found more coherent and tightly integrated patterns – particularly concerning welfare state related attitudes – in Sweden compared to the USA. Similar results are observed by Svallfors (2006). Edlund’s (1999a: 128) examination of attitudes towards state redistribution reports again similar findings and concludes that perceptions of progressive taxation as a redistributive instrument are far more common among Swedish voters compared to the American electorate. Other studies related to the non‐attitude thesis are in the same direction. In relation to attitudinal stability, Niemi and Westholm (1984) found that political attitudes were less stable in the USA than in Sweden. A panel analysis of attitudes towards the 1991 Swedish “tax reform of the century” indicated little support for the non‐attitudes thesis. Rather than finding fragmented attitude patterns in flux, the results showed that the structure of attitudes was characterized by stability and coherency to a significant degree on both the aggregate and the individual level. It is suggested that attitudes are sometimes being remodelled, but this remodelling often consists of marginal changes rather than a major change of form (Edlund 1999b). In terms of knowledge of the tax system, studies indicate that Swedes in general have 13 rather good knowledge about the tax system and its general rules (Vogel 1970; Hadenius 1986). Finally, in an analysis of how citizens understand the concept of progressive taxation, Edlund (2003) reports that “it is difficult to find any evidence (...) supporting the non‐attitude thesis that citizens do not understand the concept of progressive taxation.” Taken as a whole, we believe that these findings fit rather well with the presented theoretical approach which suggests that the Swedish welfare state institutions and the surrounding political environment may be exercising substantially stronger educational as well as ideological influences on the relationships between taxes and social spending in the public at large compared to the USA. To make a systematic comparison between the national‐specific Swedish and American studies on the non‐attitudes subject is difficult; differences regarding question wording, subject matter, and sample representations are not negligible. Still, we would argue that the overall pattern indicated by these studies suggests that public attitudes related to welfare state politics vary dramatically between the US and Sweden; even though it is difficult to determine the exact size of these differences. However, the few available comparative studies on the subject confirm the substantial differences between the two countries (e.g., Milner 2002; Holmberg & Granberg 1988; Svallfors 2006). Data and dependent variables The strategy applied in this paper to identify SFN sentiments is to use two different types of survey items measuring social spending preferences (see Confalonieri & Newton 1995; Lewis 1982). The first set contains “priced” items emphasizing that an increase in public spending 14 comes with a price in terms of the respondent having to pay more in tax in order to finance increased spending. The second set contains the “unpriced” survey items. Here, there is no mentioning of the respondent having to pay more in tax. These items simply ask whether the amount of tax revenue that goes to a particular social policy area should increase, stay the same or decrease. The difference between the two sets of survey questions is thus that the “priced” questions urge the respondents to consider their own tax contribution when financing increased public spending. One of few data sets containing both “unpriced” as well as “priced” survey questions about public spending is the 1998 Living Conditions Survey administered by Statistics Sweden. The sample used here is representative for the adult (18–79 years) Swedish population and consists of 5324 respondents (response rate: 76.7 per cent). The dependent variable – social spending preferences – is constructed with two sets of items and a total of ten indicators each capturing public spending preferences in relation to a particular social policy. Five policy areas are included. The first two policies attempt to mitigate the effects of more universally distributed risks: “medical and health care” (items 1A, 2A) and “support for the elderly” (items 1B, 2B). The next three policy areas are more selective in character as they target specific risk‐groups for which membership probabilities vary extensively in the population: “support for families with children” (items 1C, 2C), “employment policies” (items 1D, 2D) and “social assistance” (items 1E, 2E). The first battery – the “unpriced” items (1A‐E) – is worded: “Taxes are used for different purposes. Do you think that the amount of tax revenue that goes to the programs that I 15 mention should increase, stay the same, or decrease?” The response categories are recoded into three categories: (1) “increase”, (2) “stay the same or decrease”, and (3) “don’t know”. The second battery of “priced” items (2A‐E), explicitly encourages respondents to think of their own tax contribution when financing social policies: “Would you personally be willing to pay more in tax if the money went to any of the following areas …?” The response categories are coded into three groups: (1) “yes, absolutely/presumably”, (2) “no, presumably/absolutely not”, and (3) “don’t know”. Collapsing the ”stay the same” and ”decrease” response alternatives on the first battery is done in order to make the two batteries equivalent since the ”decrease” option does not exist in the second question battery. Furthermore, the proportion answering ”decrease” spending is generally small among Swedes making it less interesting to distinguish these respondents.6 Other variables used in the forthcoming analyses are described in the empirical sections as they appear. Method Two statistical techniques are mainly used in the article. First, we use Latent Class Analysis (LCA) to explore the patterning of attitudes across the two batteries of items thereby identifying potential SFN sentiments. LCA has the advantage of allowing nominal categories to be estimated thereby identifying “ideal types” without restricting attitude patterns to be linear (Hagenaars and Halman 1989). LCA also calculates the probability for each case to belong to any latent cluster making it possible to estimate the impact of various background variables on cluster membership probability. The software used for estimation of the latent variables is 16 Latent GOLD (Vermunt and Magidson 2005). Second, in order to assess the impact of the background variables on cluster membership probability we use OLS regression. Patterns of social spending preferences SFN sentiments are in the following analysis identified as those respondents expressing a preference for “increased spending” on items 1A‐E but not on the corresponding items 2A‐E. In order to explore the patterning of such responses across “unpriced” and “priced” items, a number of latent class models are tested against the data. Six LCA models are reported in Table 1. Small L2 and low BIC values indicate representative or “good” models. The L2 value of Model 1, the baseline model, indicates the maximum association between indicators that can be explained by any latent class model and thus represents the baseline model for model‐
comparisons. After running several explorative models we selected the 6‐cluster model, which reduced the L2 value by 71 percent, as representing the data. Adding additional clusters did not provide any valuable extra information and did only negligibly reduce the L2 value. [Table 1 about here] Which “ideal types” of attitude constellations can be identified when social spending preferences are simultaneously measured with “unpriced” as well as “priced” survey questions? 17 The characteristics of each of the six clusters are reported in Table 2. The table shows the probability (0‐100) for indicator responses “Increase” and “Don’t know” for each cluster.7 The size of each cluster is also reported in the table along with a discrepancy measure (DM) which captures the average degree of difference between “unpriced” and “priced” questions for each cluster.8 It is evident from Table 2 that four distinct clusters of social spending preferences can be identified (clusters 1‐4). In clusters 5 and 6, we find respondents with comparatively high probabilities to answer “don’t know”. The structure of the indifferent responses is not similar across these clusters but cluster 5 respondents clearly have a high probability to answer “don’t know” on all items regardless of question battery. Cluster 6 on the other hand display mixed attitudes across the different sets of indicators, these respondents tend to be indifferent on the “unpriced” indicators but not on the priced ones. However, cluster 6 is relatively small (2.6%) thereby preventing any in‐depth analysis or definite conclusions concerning these respondents. Returning to clusters 1‐4, it seems that three of these are characterized by a relatively high degree of attitudinal coherence across both types of survey questions whereas one cluster appears to lack such coherence. Cluster 1 respondents (hereafter labeled “universal only”) make a distinction between policies. They express strong support for increased public spending on health care and the elderly, but they show considerably less interest in further spending on policies related to children, unemployment, and social assistance, irrespective of question battery. Cluster 2 respondents (hereafter labeled “encompassing”) express coherent and the overall strongest support for increased social spending across both batteries. The probability of 18 favoring increased spending ranges between 75 and 99 per cent on all indicators except one among these respondents. In contrast, respondents belonging to cluster 4 (hereafter labeled “sceptics”) generally express the lowest level of support for increased social spending. Similar to respondents in clusters 1 and 2, they are coherent in their spending preferences across “unpriced” and “priced” items. The relative size of this cluster is about 9 percent which can be compared to the much larger groups of respondents in clusters 1 and 2, constituting 36 and 33 percent, respectively. The third cluster, constituting 15 per cent of the citizenry, is of particular interest since the social spending preferences of these respondents suggest that they want something for nothing and lack the coherence found within clusters 1, 2 and 4. When focusing on the first battery of indicators, 1A‐E, cluster 3 respondents show similarities with the respondents belonging to cluster 2 in their relatively strong support for increased social spending. However, this similarity vanishes completely when the second set of “priced” indicators are examined. Cluster 3 respondents now express the lowest level of support for increased spending compared to all other clusters. Respondents belonging to cluster 3 (hereafter labeled “SFN”) thus appear to dramatically reevaluate their support for social spending when faced with questions highlighting the fact that increased public spending comes with a personal price. [Table 2 about here] 19 A look at the discrepancy measure presented in Table 2 confirms the inconsistency characterizing the social spending preferences of cluster 3 respondents: the mean of the absolute difference in probabilities between each paired items is 31 for cluster 3, compared to 6 for cluster 1 and 2. The probability of favoring increased spending thus differs by 31 percentage units on average across all five pairs of indicators for cluster 3 respondents. This difference is most staggering when it comes to “support for the elderly”, indicators 1B and 2B; when asked an “unpriced” question, the probability to answer “increase spending” is 93 per cent which can be compared to the 3 percent chance of wanting to increase spending when the “priced” question is used. A considerable degree of coherence can be found among cluster 1, 2 and 4 respondents. Since these clusters constitute 37, 32, and 9 per cent, respectively, of the sample, it seems that Swedish mass opinion on public spending is characterized by coherence rather than incoherence. Considering the remaining 22 percent of the sample, the two small indifferent clusters (5 and 6) make up 7 per cent of the respondents and the SFN pattern among cluster 3 respondents constitutes about 15 per cent of the sample. [Table 3 about here] Of course, one could argue that cluster 3 respondents think that the extra taxes needed for increased spending on welfare programs should be financed by decreased spending in other 20 domains. Finding that these respondents are more prone to cut spending in other areas compared to other groups would, of course, seriously undermine any kind of SFN‐interpretation of the empirical results. However, as shown in Table 3, there is no evidence that cluster 3 respondents are more sympathetic to decreased spending on defense, public administration, research and higher education, compulsory schooling, or environmental protection compared to the respondents of cluster 1, 2 and 49. How should the ambivalent something for nothing pattern associated with cluster 3 respondents be understood; as a rational group of citizens that assume that someone else should pay the necessary extra taxes financing the increased social spending, or as ill‐informed, ignorant, and confused citizens possessing political non‐attitudes? In the next section, we will study how the six clusters identified fit an ideal‐type model of welfare state ideology. The main purpose is to determine whether citizens holding SFN attitudes deviate to a greater extent from the ideal‐type model compared to the other groups indicating less coherence and political sophistication. Similarly, we also analyze whether the degree of political interest varies across clusters, possibly indicating whether cluster 3 respondents deviate from other respondents in terms of political interest. Cluster membership and its fit to the ideal‐typical model of welfare state ideology Svallfors (1999) describes the characteristics of an ideal‐type model of support for an encompassing welfare state. First, trust in the capabilities of the welfare state institutions to solve social problems should be high. There should ideally be a strong support for high levels of 21 public spending primarily financed collectively through taxes. Public authorities should generally be regarded as best suited for organizing and delivering public services. Further, suspicion about welfare policies being abused should not be widespread in the public. Svallfors emphasizes that this should be regarded as an “ideal” model, and we can therefore not expect to find uniform support for all aspects of the welfare state among the public. However, this model can serve as a guideline when determining how an ideal‐type model of welfare state ideology should be structured. The attitude dimensions of interest when outlining the ideal‐type model are thus the following three: financing (collective vs individual), service delivery (state/local authorities vs other alternatives) and suspicion of abuse of welfare policies. In our data set, the following social policy areas are available: health care, child care, elderly care, pensions, sickness insurance, unemployment insurance, social services, housing allowances and social assistance. In sum, the typology of welfare state ideology is empirically measured by 15 trichotomized items. 10 For each item, the value 1 (ideal model support) indicate a preference for collective financing, public organization of service delivery, or low degree of suspicion of abuse of welfare policies. Conversely, we also distinguish those respondents holding consistent ideal typical attitudes against an encompassing welfare state. These are represented by the value 0 (ideal model opposition) and indicate for each item a preference for individual financing, non‐public organization of service delivery, or high degree of suspicion of abuse of welfare policies. Finally, the indifferent responses (don’t know) is represented by the value –1. All bivariate relationships across the 15 items are cross‐tabulated and allocated into four main categories. The mean share of respondents expressing consistent pro welfare state attitudes on these item‐pairs (cell: 1,1) is the first category. The second category is comprised of those 22 expressing consistent negative welfare state combinations (cell: 0,0). The third response combination, classified as “mixed”, covers the respondents expressing any combinations of the responses 1,0 or 0,1, e.g., support for collective financing of health care combined with support for non‐public service organization. Finally, the share of indifferent respondents answering “don’t know” (any bivariate combination involving at least one –1 response) constitutes the fourth category. In the analysis a distinction is made between universal (policies related to health care and the elderly) and selective (policies related to children, unemployment, housing, and social services) policy areas. Is the proportion of consistent, mixed and indifferent attitude combinations different within the SFN segment compared to cluster 1, 2 and 4 thereby suggesting that SFN respondents generally are less coherent in terms of their welfare state ideology? Figures 1 and 2 show the results of the analysis for the universal and the selective policy areas, respectively. Each cluster is represented by a bar and the bars are divided into four segments according to the different attitude combinations described above. Two segments indicate the proportion of “consistent” ideal‐typical support (consistent +) and non‐support (consistent –). The segment labeled “mixed” show the proportion expressing inconsistent attitudes and the fourth segment shows the proportion of “don’t know” responses. Figure 1 shows, for example, that the average share of consistent ideal‐typical attitude combinations (consistent +) on the universal item‐pairs is 59 per cent for cluster 3, and the share of consistent opponents (consistent –) to such an ideal‐typical welfare state ideology amount to 4 per cent. This can be compared to 62 + 4 per cent and 69 + 3 per cent for clusters 1 and 2. The average proportion of “mixed” attitudes is 26 per cent among cluster 3 respondents 23 which can be compared to 28 and 22 per cent for clusters 1 and 2. It is also evident from Figure 1 that clusters 5 and 6 display a high degree of “don’t know”‐answers compared to the other clusters indicating that these respondents tend to be indifferent regardless of whether attitudes towards social policy concerns spending, financing, service delivery, or suspicion of abuse of social policies. For the selective programs, depicted in Figure 2, we find slightly different patterns. In general, attitudes are less ideologically consistent; for selective programs citizens are more prone to see mixed responsibilities. In terms of cluster differences, empirical results are similar in both figures. Cluster 2 respondents, favoring an encompassing welfare state, show overall the highest level of ideological consistency. Again clusters 5 and 6 stand out as having a high degree of indifferent responses. However, the most important conclusion to be drawn from Figures 1 and 2 is that SFN respondents – although having a slightly higher proportion of indifferent responses – do not deviate in any obvious sense from clusters 1, 2 and 4 in terms of ideological consistency. There is thus little evidence suggesting that the SFN sentiments of cluster 3 respondents reflect general non‐attitudes. [Figure 1 + 2] When examining the consistent pattern across the clusters, it is clear that those in favor of an ideal‐typical welfare state dominate over those opposing the ideal‐typical model, even among respondents that are opposing increased spending (sceptics). While this may be seen as ideologically incoherent, we must remember that cluster 4 respondents may not be in favour of 24 welfare state retrenchment, they simply do not want welfare state expansion. We would rather suggest that this finding reflect that the share of true anti‐welfare state proponents actually constitute a quite small proportion of the Swedish electorate. Finally, in order to further investigate whether the SFN sentiments of cluster 3 respondents reflect general non‐attitudes we compare the degree of political interest across the clusters. We use an item asking whether the respondent is taking part in political discussions among friends distinguishing between those actively taking part in such discussions and those not taking part. The results shown in Table 4 again suggest that the SFN respondents of cluster 3 are quite similar to clusters 1, 2 and 4 in terms of political interest/activity. [Table 4 about here] In summary, there is little suggesting that the SFN pattern identified among cluster 3 respondents reflects general non‐attitudes. Cluster 3 clearly does not stand out from clusters 1, 2 and 4 in terms of attitudinal coherence and general welfare state ideology. Nor is the SFN respondents of cluster 3 less politically interested than the respondents of clusters 1, 2 and 4. The analysis has also revealed two clusters possessing non‐attitudes, clusters 5 and 6. These respondents are clearly indifferent when it comes to social spending, financing, service delivery and suspicion of abuse of welfare policies. 25 Associations between structural determinants and cluster membership The aim of the following analysis is to study how individual background characteristics and economic factors are associated to the different clusters. The SFN segment will be focused and contrasted primarily with cluster 1, 2 and 4. Cluster 5 is also included but due to the small size and ambiguous characteristics of cluster 6 we have omitted this cluster from further analysis. The explanatory variables used have all been identified by previous research as important predictors for social spending preferences and include gender, age, class, household composition, employment sector, income, and economic problems. The class variable distinguishes 6 categories according to the EGP classification (Erikson & Goldthorpe 1992). Employment sector separates public employees from those working in the private sector. Income is measured as quartiles according to household income. Whether or not the respondent experience economic problems are measured by the following survey question: “Going by your current household income, are you able to make ends meet?” A number of OLS regressions will estimate the impact of these variables on the probability to belong to the various clusters. Initially, traditional background characteristics describing the structural position of the respondents are included in Table 5. The bivariate estimates as well as the estimates derived from the multiple regressions are displayed in the table. Overall, the results found in Table 5 for clusters 2 and 4 indicate the significance of gender, class, and employment sector and are in line with previous research on social spending preferences (see e.g. Edlund 2006; Svallfors 1996; Svallfors 1999). For example, an increase in cluster 2 membership probability is associated with belonging to the working class and public 26 sector employment. The opposite relationship between these structural determinants and cluster membership probability can be observed for the “sceptics” belonging to cluster 4. These respondents express the lowest levels of support for increased spending among all clusters and are typically men belonging to the service class. Belonging to cluster 1, favoring increased spending on universal policy areas but not on selective ones, is associated with being male and self‐employed. However, these clusters are of less interest in the subsequent analysis and serve primarily as contrasting cases for comparisons against the SFN sentiments of the cluster 3 respondents. [Table 5 about here] Which individual background characteristics are associated with the SFN segment? Table 5 shows that the SFN attitudes of cluster 3 respondents are primarily determined by gender, age and household composition. Interestingly, the typical determinants predicting welfare state support such as social class and employment sector does not affect the probability to belong to cluster 3. Women have a significantly higher probability belonging to cluster 3 compared to men. When it comes to age the pattern identified suggests that the youngest and oldest age groups are significantly overrepresented in cluster 3 compared to the middle aged groups. The variable “household composition” is further indicating that “singles with children” are the ones having the highest probability to belong to cluster 3. 27 The patterns associated with cluster 3 membership lean toward an interpretation of the segment as possibly being in economically vulnerable positions. Those of young or old age might well be more exposed to economic uncertainty compared to the middle aged, as might also be the case for women compared to men. This picture is further strengthened when the impact of household composition is examined. Here “singles with children” stand out as having about 7 per cent higher probability to belong to cluster 3 compared to “couples”. The next step in the analysis is therefore to incorporate two economic variables in order to examine whether the observed gender, age, and household composition differences, have something to do with differences in economic vulnerability. New models for clusters 1‐5, incorporating two variables measuring income as well as whether respondents are experiencing economic problems, are presented in Table 6. One hypothesis could be that the gender, age and household differences disappear when “economic vulnerability” is sufficiently measured and included in the model. When examining Table 6 it is evident that the differences along the lines of gender and age are not significantly reduced controlling for the economic variables. However, the variable “household composition” is no longer statistically significant indicating that differences between respondents according to their household composition were mainly due to economic differences. [Table 6 about here] Although the economic situation of the respondents cannot explain differences between men 28 and women or between different age groups, Table 6 clearly shows that the occurrence of economic problems distinctly increases the probability to belong to the SFN segment. The difference between those respondents experiencing economic problems and those having no problems making ends meet is about 9 per cent. In a bivariate model (not shown) the variable “income” is statistically significant, but do not reach significance when “economic problems” is included. Having a low income thus impacts on cluster 3 membership probability, but it is mainly a matter of low income categories experiencing economic problems. Overall, the results of the regression analyses lean towards an interpretation that SFN sentiments should rather be understood as a consequence of economic hardship than as free riding in the standard economic version of the concept. Concluding discussion This paper has examined public preferences for social spending with an explicit focus on the presence of the something for nothing syndrome and non‐attitudes. The analysis shows that political non‐attitudes is not dominant in the Swedish citizenry; indifferent respondents admitting that they do not have any specific preferences constitute about 7 per cent of the electorate. Another group of respondents – constituting 15 per cent of the sample – that may fall within the limits of the non‐attitude concept is those holding something for nothing attitudes; thus leaving out nearly 80 per cent of the respondents. The true estimate regarding the prevalence of non‐attitudes might even be smaller because the empirical results indicate that those holding SFN preferences do not easily fit into the non‐attitudes concept. The quality 29 of the SFN segment’s belief system, in terms of ideological coherence, does not represent a deviant case compared to other groups. Furthermore, it is also difficult to view them as free riders in the standard economic version of the concept. As indicated by the individual‐level structural determinants associated with the SFN pattern, it seems more appropriate to view them as free riders in economically vulnerable positions. Another implication of the findings in this paper, which fare rather well with previous Swedish empirical evidence, is that socio‐political context may matter for the presence of non‐attitudes. Compared to previous US evidence, referred to above, reporting very low levels of coherent and informed attitudes and rather large fractions of different types of non‐attitude patterns, the results reported in this paper, along with previous Swedish evidence, suggests that national political institutions and associated organized interests may influence the degree of public understanding of the functioning of the tax system and other welfare state institutions. In our view, to focus on such macro‐social factors in the search for explanations of the “something‐for‐
nothing” syndrome and the non‐attitudes phenomenon in general provides a fruitful alternative/complement to approaches arguing for a stricter focus on socio‐psychological traits and abilities on the individual level. Indeed, individual circumstances matter, but so do national contexts. 30 References Alesina, A. & Giuliano, P. 2010. Preferences for Redistribution. In Besin, A., Benhabib, J. & Jackson, M O. (eds.). Handbook of Social Economics. Amsterdam: North Holland. Bartels, L M. 1996. Uninformed Votes: Information Effects in Presidential Elections. American Journal of Political Science 40 (1): 194‐230. Bartels, L M. 2005. Homer Gets a Tax Cut: Inequality and Public Policy in the American Mind. Perspectives on Politics 3: 15‐31. Beedle, P. & Taylor‐Gooby, P. 1983. Ambivalence and Altruism: Public Opinion about Taxation and Welfare. Policy & Politics 11 (1): 15‐39. Brooks, C. & Manza, J. 2007. Why Welfare States Persist. The Importance of Public Opinion in Democracies. Chicago: The University of Chicago Press. Citrin, J. 1979. Do People Want Something for Nothing: Public Opinion on Taxes and Government Spending. National Tax Journal 32: 113‐129. Claggett, W J M. & Schafer, B E. 2010. The American Public Mind: The Issues Structure of Mass Politics in the Postwar United States. Cambridge: Cambridge University Press. Corneo, G. & Grüner, H P. 2002. Individual Preferences for Political Redistribution. Journal of Public Economics. 83: 83‐107. Confalonieri, M A. and Newton, K. 1995. Taxing and spending: Tax revolt or tax protest? In Borre, O. & Scarbrough, E. (eds.) The Scope of Government. Beliefs in Government 3: 121‐148. Oxford: Oxford University Press. 31 Converse, P E. 1964. The nature of belief systems in mass politics. Ideology and discontent: 206‐
261. Edlund, J. 1999a. Attitudes Towards Tax Reform and Progressive Taxation: Sweden 1991‐1996. Acta Sociologica 42 (4): 337‐355. Edlund, J. 1999b. Progressive Taxation Farewell? Attitudes to Income Redistribution and Taxation in Sweden, Great Britain and the United States. In Svallfors S & P Taylor‐Gooby, eds. The End of the Welfare State? Public Responses to State Retrenchment. London: Routledge. Edlund, J. 2003. Attitudes Towards Taxation: Ignorant and Incoherent? Scandinavian Political Studies 26 (2): 145‐167. Edlund, J. 2006. Trust in the capability of the welfare state and general welfare state support: Sweden 1997‐2002. Acta Sociologica 49: 395‐417. Erikson, R. & Goldthorpe, J H. 1992. The Constant Flux. Oxford: Clarendon Press. Granberg, D. & Holmberg, S. 1988. The Political System Matters: Social Psychology and Voting Behavior in Sweden and the United States. Cambridge: Cambridge University Press. Hadenius, A. 1986. A Crisis of the Welfare State? Stockholm: Almqvist & Wiksell international. Hagenaars, J A. and Halman, L C. 1989. Searching for Ideal Types: The Potentialities of Latent Class Analysis. European Sociological Review 5: 81‐96. Hall, P. 1992. The Movement from Keynesianism to Monetarism: Institutional Analysis and British Economic Policy in the 1970s. In Structuring Politics. Historical Institutionalism in Comparative Analyses, (eds.) Steinmo, S., Thelen, K., & Longstreth, F. Cambridge: Cambridge 32 University Press. Inglehart, R. 1990. Culture Shift in Advanced Industrial Society. Princeton, N J: Princeton University Press. Key, W O. 1961. Public Opinion and American Democracy. New York: John Wiley. Kinder, D & Herzog, D. 1993. Democratic Discussion. In Marcus, G E & Hanson, R L (eds.) Reconsidering the Democratic Public. Pennsylvania: The Pennsylvania State University Press. Kinder, D R. 1983. Diversity and Complexity in American Public Opinion. In Finifter, A W. (ed.) Political Science: The State of the Discipline. Washington: American Political Science Association. Korpi, W. & Palme, J. 1998. The Paradox of Redistribution and Strategies of Equality: Welfare State Institutions, Inequality, and Poverty in the Western Countries. American Sociological Review 63: 661‐687. Korpi, W & Palme, J. 2003. New Politics and Class Politics in the Context of Austerity and Globalization: Welfare State Regress in 18 Countries 1975‐1995. American Political Science Review 97: 426‐446. Korpi, W. 1980. Social Policy and Distributional Conflict in the Capitalist Democracies. A Preliminary Comparative Framework. West European Politics 3: 296‐316. Korpi, W. 1983. The Democratic Class Struggle. London: Routledge and Kegan Paul. Korpi, W. 2006. Power Resources and Employer‐Centered Approaches in Explanations of Welfare States and Varieties of Capitalism: Protagonists, Consenters, and Antagonists. World 33 Politics. 58 (2): 167–206. Kriesi, H. 2004. Stability and Change of Opinion: The Case of Swiss Policy against Pollution caused by Cars. In Saris, W E & Sniderman, P. (eds.) Studies in Public Opinion: Attitudes, Nonattitudes, Measurement Error, and Change. Princeton: Princeton University Press. Lewis, A. 1978. Perceptions of Tax Rates. British Tax Review 6: 358‐366. Lewis, A. 1982. The Psychology of Taxation. Oxford: Martin Robertson. Lipset, S M. 1996. American Exceptionalism: A Double‐edged Sword. New York: W. W. Norton & Company. Mettler, S & Soss, J. 2004. The consequences of public policy for democratic citizenship: Bridging policy studies and mass politics. Perspectives on Politics 2: 55‐73. Milner, H. 2001. Civic Literacy: How Informed Citizens Make Democracy Work. Hanover NH: University Press of New England. Niemi, R & Westholm, A. 1984. Issues, Parties, and Attitudinal Stability: a Comparative Study of Sweden and the United States. Electoral Studies 3: 65‐83. North, D. 1990. Institutions, institutional change, and economic performance. Cambridge: Cambridge University Press. Page, B & Shapiro, R. 1992. The Rational Public. Fifty Years of Trends in Americans’ Policy Preferences. Chicago: University of Chicago Press. Page, B & Shapiro, R. 1993. The Rational Public and Democracy in Marcus, G E & Hanson, R L (eds.) Reconsidering the Democratic Public. Pennsylvania: The Pennsylvania State University 34 Press. Peters, G. 1991. The Politics of Taxation. Oxford: Blackwell. Popkin, S L. 1991. The reasoning voter: communication and persuasion in presidential campaigns. Chicago: University of Chicago Press. Roberts, M L., Hite P A. & Bradley C F. 1994. Understanding Attitudes Toward Progressive Taxation. Public Opinion Quarterly 58 (2): 165–190. Rothstein, B. 1998. Just Institutions Matter. The Moral and Political Logic of the Universal Welfare State. Cambridge: Cambridge University Press. Rothstein, B. & Steinmo, S. 2002. (eds.). Restructuring the Welfare State: Political Institutions and Policy Change. New York: Palgrave. Saris, W E. & Sniderman, P M. 2004. Introduction in Saris, W E & Sniderman, P M (eds.). Studies in Public Opinion. Attitudes, Nonattitudes, Measurement Error, and Change. Princeton: Princeton University Press. Sears, D O. & Citrin, J. 1985. Tax Revolt. Something for Nothing in California. Enlarged Edition. Cambridge: Harvard University Press. Sears, D O. & Citrin, J. 1982. Tax revolt: Something for nothing in California. Cambridge: Harvard University Press. Sniderman, P M. 1993. The New Look in Public Opinion Research in Finifter A W (ed.). Political Science: The State of the Discipline II. Washington DC: The American Political Science Association. 35 Soroka, S N. & Wlezien, C. 2005. Opinion–Policy Dynamics: Public Preferences and Public Expenditurein the United Kingdom. British Journal of Political Science. 35: 665‐689. Steinmo, S. 1993. Taxation and Democracy. New Haven: Yale University Press. Steinmo, S. 1998. Tax Policy. Cheltenham: Edward Elgar. Stimson, J A. 1991. Public Opinion in America: Moods, Cycles, and Swings. Boulder CO: Westview Press. Svallfors, S. 1996. Välfärdsstatens moraliska ekonomi: välfärdsopinionen i 90‐talets Sverige [The Moral Economy of the Welfare State: Public Opinion on Welfare in Sweden in the 1990s]. Umeå: Boréa. Svallfors, S. 1999. Mellan risk och tilltro: opinionsstödet för en kollektiv välfärdspolitik [Between Risk and Trust: Public Support for Collective Welfare Policies]. Umeå: Univ. Svallfors, S. 2006. The Moral Economy of Class: Class and Attitudes in Comparative Perspective. Stanford: Stanford University Press. Svallfors, S. 2007. Introduction in Svallfors, S. (ed.) The Political Sociology of the Welfare State: Institutions, Social cleavages, and Orientations. Stanford: Stanford University Press. Thelen, K. & S Steinmo. 1992. Historical Institutionalism in Comparative Politics. In Steinmo S, K Thelen & F Longstreth, eds. Structuring Politics. Historical Institutionalism in Comparative Analysis. Cambridge: Cambridge University Press. Welch, S. 1985. The "More for Less" Paradox: Public Attitudes on Taxing and Spending. The Public Opinion Quarterly 49: 310‐316. 36 Wlezien, C. 1995. The Public as Thermostat: Dynamics of Preferences for Spending. American Journal of Political Science 39: 981‐1000. Vermunt, J K. & Magidson, J. 2005. Latent GOLD 4.0 User’s guide. Belmont, MA: Statistical Innovations Inc. Vogel, J. 1970. Aspirationer, möjligheter och skattemoral. SOU 1970:25. Stockholm: Allmänna förlaget. Zaller, J R. 1992. The Nature and Origins of Mass Opinion. New York: Cambridge University Press. 37 Notes 1
For an overview of the non‐attitudes debate and its tremendous influence on the field in the USA, see Kinder (1983). 2
Originally, the concept of non‐attitudes had a quite specific meaning and was mainly used as explaining attitude instability over time. The idea was that most citizens do not hold any views on major political issues but when asked about them in surveys they express opinions anyway in order to avoid embarrassment of being ignorant. In this paper, the term non‐attitudes is used as an umbrella concept covering both criteria, as both can be seen as different types of empirical responses reflecting the same phenomenon (cf. Saris & Sniderman 2004). A closely related issue is the degree of political knowledge and awareness, which may be seen as a cause of expressed non‐
attitudes, rather than as a dimension of the non‐attitudes concept. 3
But see Welch (1985) suggesting that the SFN segment may be considerable smaller if alternative sources of financing increased spending are presented to the respondents. 4
We have to admit that we do not believe that such an interpretation would fare well with the data presented in Sears and Citrin (1985) and Bartels (2005) but from the analytical perspective of this paper, the argument is interesting. 5
Although there might be disagreement among scholars of social policy whether or not it is appropriate to denote the latter type of motives/reasons as free riding influences, we would defend the use of the term free rider by arguing that the term, in this paper, is used in a value‐laden free way; it is used for the sole purpose to analytically capture those that have rational motives/reasons – which may differ depending on socio‐economic position and political ideology – for expressing SFN preferences. 6
The percentage of “decrease spending” responses on items 1A‐E is: 1.7; 1.0; 3.5; 11.9 and 10.5. 7
The probability for the third response category, “Decrease or stay the same”, is given by subtracting these probabilities from 100. For example, for cluster 2 respondents, the probability for indicator response “increase” on the first indicator 1A (medical and health care) is 93 per cent; the corresponding probability for answering “don’t know” is 1 per cent. The probability for the indicator response, “decrease or stay the same”, is therefore 6 per cent (100 – (93+1) = 6). 8
DM = (|1A‐2A| + |1B‐2B| + |1C‐2C| + |1D‐2D| + |1E‐2E|) / 5 9
These items are included in the same battery as the “unpriced” items constituting the dependent variable, 1A‐E, and the wording of the questions is thus identical: “Taxes are used for different purposes. Do you think that the amount of tax revenue that goes to the programs that I mention should increase, stay the same, or decrease?” 10
The number of social policy areas varies between the three attitude dimensions. Financing: health care, child care, elderly care, pensions, sickness insurance and unemployment insurance; public service provider: health care, elderly care, child care and social services; suspicion of abuse: health care, sickness insurance, unemployment insurance, housing allowances and social assistance. 38 Table 1. Public spending preferences. Swedes aged 18–79 years (n = 5185). Model fit for 6 Latent Class models. BIC (L2) df L2‐reduction % L2 Model 1 1‐cluster 26191.6 –17987.4 5165 0.0 Model 2 2‐cluster 16129.1 –27870.2 5144 38.4 Model 3 3‐cluster 11121.4 –32698.3 5123 57.5 Model 4 4‐cluster 9485.6 –34154.4 5102 63.8 Model 5 5‐cluster 8467.2 –34993.3 5081 67.7 Model 6 6‐cluster 7602.6 –35678.2 5060 71.0 Table 2. Public spending preferences. Probability (0‐100) for indicator responses Increase|Don’t know by cluster membership. 6‐cluster model from Table 1 (n = 5185). Cluster
Indicators 1A 1B 1C 1D 1E 1 2 3 4 5 6 78|01 78|01 36|03 46|04 10|07 93|01 95|01 78|04 80|02 41|11 89|02 93|02 55|06 59|04 19|16 19|02 20|02 11|01 18|03 03|05 46|38 48|38 26|48 21|56 07|67 24|68 23|76 10|79 19|70 02|87 2A 2B 2C 2D 2E Cluster size % DM 90|00 92|01 53|02 41|02 27|02 37.2 6 99|00 99|00 96|01 84|02 75|03 32.2 6 02|00 03|00 04|00 07|00 02|00 14.7 31 12|01 13|01 08|01 08|02 02|01 8.7 3 27|70 24|75 10|87 05|88 00|92 4.6 8 40|00 41|03 25|04 22|06 18|04 2.6 18 % in tot. sample 76|05 77|05 49|08 55|08 21|14 68|03 69|04 53|05 45|06 35|06 Table 3. Proportion of respondents proposing “decrease spending” on various public policy areas by cluster membership (%). C1 C2
C3
C4
C5 Defense 45.1 46.2
42.6
46.7
22.8 15.4
Public administration 57.3 44.1
54.5
60.4
20.7 23.0
Research and higher education 5.1 1.9
4.3
5.3
2.1 26.5
Compulsory schooling 0.5 0.1
0.3
2.9
0.4 0.0
Environmental protection 3.8 1.3
3.9
8.2
3.0 0.7
(1927) (1671)
(760)
(452)
(236) (135)
(n) C6
Table 4. Political interest by cluster membership. Proportion of respondents participating actively in political discussions (%). C1
C2
C3
C4
C5 C6
Not active 27
29
26
18
47 38
Active 73
71
74
82
53 62
(1925)
(1672)
(761)
(452)
(233) (135)
(n) Table 5. Social spending preferences. Cluster membership probability (0 – 100) by individual structural location. OLS regression. Cell entries are unstandardized regression coefficients. Variable values equal to 0 are reference categories. Cluster 1 Cluster 2 Cluster 3 Cluster 4 Cluster 5 Universal only Encompassing SFN Sceptics Indifferent Bivar. Mult. Bivar. Mult. Bivar. Mult. Bivar. Mult. Bivar. Mult. Gender Men 4.12 3.04 –1.83 1.06 –4.50 –5.50 4.78 4.07 –1.79 –1.51 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 Women Age 3.76 4.17 –0.17 –2.94 –1.16 0.16 0.43 0.09 –2.65 –1.26 18‐29 years 30‐49 years 4.24 4.98 5.24 2.57 –4.98 –6.69 2.71 2.51 –5.40 –2.02 3.76 2.41 6.84 5.05 –5.80 –5.40 2.10 1.13 –5.80 –2.69 50‐64 years 65‐79 years 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 Class 6.54 7.02 –8.58 –8.79 0.88 2.86 4.13 3.48 –3.36 –5.05 Self‐employed 5.24 4.02 –8.86 –11.89 –3.58 –0.69 11.87 10.96 –4.67 –3.17 Service class I Service class II 3.59 3.11 –1.33 –3.57 –0.21 1.18 4.62 4.47 –4.98 –4.02 5.06 4.95 –2.05 –2.75 0.72 0.91 –0.52 –0.23 –3.63 –3.37 Routine non‐manuals Skilled workers –1.76 –2.50 4.92 4.08 0.28 1.69 –0.06 –0.87 –2.55 –1.95 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 Unskilled workers Household comp. 11.15 11.42 –5.96 –2.63 –6.19 –7.45 1.35 1.36 –1.37 –2.65 Couples Couples with children 10.26 8.22 –4.89 –2.51 –5.93 –4.21 1.97 0.59 –2.73 –2.80 8.21 8.39 –10.18 –6.53 –3.79 –5.02 2.17 3.01 1.39 –1.30 Singles Singles with children 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 Employment sector 3.48 2.60 –5.16 –5.75 0.42 2.11 1.17 0.70 –0.52 –0.42 Private 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 Public Intercept 18.80 41.94 24.71 1.76 10.12 (n) (4808) (4808) (4808) (4808) (4808) Estimates in boldface = p < 0.05. Table 6. Social spending preferences. Cluster membership probability (0 – 100) by individual structural location and economic vulnerability. OLS regression. Cell entries are unstandardized regression coefficients. Variable values equal to 0 are reference categories. Cluster 1 Cluster 2 Cluster 3 Cluster 4 Cluster 5 Universal only Encompassing SFN Sceptics Indifferent Gender Men 2.45 0.14 –4.90 4.33 –0.66 Women 0.00 0.00 0.00 0.00 0.00 Age 18 – 29 years 5.36 –6.62 0.80 0.33 0.22 30 – 49 years 5.50 –0.30 –6.16 2.55 –0.62 50 – 64 years 2.48 2.41 –4.64 0.87 –1.21 65 – 79 years 0.00 0.00 0.00 0.00 0.00 Class Self‐employed 6.59 –9.00 3.56 2.83 –4.28 Service class I 1.65 –11.34 1.74 9.57 –2.54 Service class II 1.83 –4.06 2.65 3.36 –2.85 Routine non‐manuals 4.26 –3.16 1.42 –0.71 –2.43 Skilled workers –3.05 4.01 2.02 –0.89 –1.81 Unskilled workers 0.00 0.00 0.00 0.00 0.00 Household composition –3.52 –0.83 –1.57 Couples 8.13 –2.13 Couples with children 6.12 –2.22 –2.11 –0.64 –2.08 Singles 7.33 –5.54 –3.25 1.46 –1.49 Singles with children 0.00 0.00 0.00 0.00 0.00 Employment sector Private 2.44 –5.47 2.44 0.34 –0.64 Public 0.00 0.00 0.00 0.00 0.00 Income High 4.56 –2.12 –2.42 1.87 –0.96 Medium/high 3.58 3.45 –1.54 –1.57 –2.01 Low/medium 4.93 1.90 –1.27 –1.80 –1.93 Low 0.00 0.00 0.00 0.00 0.00 Economic problems No difficulties making ends meet 8.28 2.47 –8.83 1.77 –2.56 Some difficulties making ends meet 6.81 –0.93 –4.97 1.44 –1.28 0.00 0.00 0.00 0.00 0.00 Difficulties making ends meet Intercept 11.75 43.47 27.30 2.98 10.14 (n) (4524) (4524) (4524) (4524) (4524) Estimates in boldface = p < 0.05.