= = = = = = Dissatisfied Democrats, Policy Feedback, and European Welfare States, 1976-2001 Staffan Kumlin = = = = = = = = = = QoG WORKING PAPER SERIES 2010:4= = THE QUALITY OF GOVERNMENT INSTITUTE Department of Political Science University of Gothenburg Box 711 SE 405 30 GÖTEBORG March 2010 ISSN 1653-8919 © 2010 by Staffan Kumlin. All rights reserved. Paper prepared for presentation at ECPR Joint Sessions of Workshops, Münster, April 2010 (workshop 19: “The Interrelationship between Institutional Performance and Political Support in Europe”). Dissatisfied Democrats, Policy Feedback, and European Welfare States, 1976-2001 Staffan Kumlin QoG Working Paper Series 2010:4 March 2010 ISSN 1653-8919 Abstract: Are citizens’ level of satisfaction with the functioning of democracy affected by welfare state-related policies and outcomes? Three-level analyses of Eurobarometer surveys from three decades suggest that generosity in unemployment benefits (but not pensions, sick pay, or income inequality) helps explain over-time within-country variation in satisfaction with democracy. This effect is relatively stable across individuals with different interests and values, but is conditioned by unemployment rates. Specifically, the results support a “visible costs hypothesis” predicting weaker generosity effects when more people are out of work. In conclusion, the long-term rise in unemployment in Western Europe may have assisted in the birth of dissatisfied democrats directly, as well as indirectly by disarming the previously legitimizing force of unemployment protection policies. Staffan Kumlin Department of Political Science University of Gothenburg staffan.kumlin@pol.gu.se 1 “The welfare state” and “political support” are among the more scrutinized topics in comparative research on advanced ind ustrial de mocrac i e s . However, while impressi v e bodies of work about each keep accumulating (see Dalton 20 04; Newton 2006; Carnes and Mares 2007) they have rar ely been expl icitly connected in em pirical analyses. This is unfortunate as the two literatures are nevertheles s linked by common patterns and themes. One of the most per v asive ones is the possibility of decline; that is, simult aneously downward trends in certain aspects of political support as well as signs of welfare state retrenchment. As for politica l support— o r “trust” as it is so metim e s called— t i m e s series includi n g the 1980s indicated no universal and lasting trend (Klin gemann and Fuchs 1995). But recent work on longer time-spans and more encompas sing data suggest “public scepticism about politicians and governm e n t officia l s is spreadi n g to virtual l y a ll the advanced in dustrial democracies” (but see Torcal and Montero 2006; Dalton 2008:243 ) . At the same time, however, normativ e support for the idea of democracy with its atta ched norms, rights, and procedures has stayed strong. To capture this current ambi valence—en dorsement of democratic principles but dissatisfaction with their implementation—scholars have coined expressi ons like “dissatisfied democrats” or “critical citizens ” (e.g. Klingema n n 1999; Norris 1999). Moving to welfare state research, sch olars have addressed the nature and impact of “permanent austerity,” (Pierson 2001) a situation in which it is difficult to finance previous commitments to public services and income replacement system s. While the policy responses are unlikely to involve radical welfare backlash, moderate adap tati on and cost containmen t efforts within existing systems now seem more common. As prophesied by Pierson (2001:417), “neither the alternatives of standing pat or dism antl i n g are likely to prove viable in most countrie s . Instead, as in most aspects of politics, we should expect st ron g pressures to move towards more cen trist— and therefore more incremental—res ponses. Those seeking to generate significant cost 2 reductions while modernizing particular aspects of social provision will generally hold the balance of political power.” Several policy changes ha ve been register e d . Korpi and Palme (2003) investigated net replacem ent rates in the public insuran c e systems for sicknes s , work acciden t , and unemplo yment for 18 OECD cou ntries, and found that “th e long gradual increase in average benefit levels characterizi ng developments up to the mid-1970s has not only stopped but turned into a reverse” (Korpi and Palme 2003:445 ; c.f. Allan and Scruggs 2004 ). Similarly, results indicating gradual service deterioratio n and in creas i ng resource-scarcity (rath er than radical system chan ge) have been reported in comparative studies of public services such as edu cation and health care (Clayton and Pontusso n 1998). Fina lly, adding injury to insult, income inequali t y (Brandolini and Smeeding 2008) and unemploymen t rates (Ca m eron 2001) increased in many countries during this period, especially in Western Europe. At the same time, support for welfare state policies a ppears rath er stable at high levels (for recent overview s , see Kumlin 2007b; Svallfor s 2010b), an d is found to strengthen where unem ployment and inequality are increasi ng (Borre and Scar brough 1995; Blekesaune and Quadagno 2003; Finseraas 2009). This leads to the bas ic query deal t with in this paper. Does less generosity in widely popular welfare state policies, and deterioration in the outcomes th at they are mean t to affect, generate dissatisfied democrats ? The next section reviews and critiques past rese arch on “government performance” and political support. Many studies have dealt with th e apparently weak impact of macroeconomic performance. However, I argue that su ch negative results—and the explanati ons offered—are not automa tically valid for welfare state-related outputs and outcomes. The subsequent section discusses a smaller accu mulation of studies on we lfare state performance and political support. These concl u de that perf ormance may affect polit ical support. However, th ey also display various featur e s callin g for furthe r inve stigation. For example, one strand of evidence comes from 3 historical case studies; these are valuab le but cannot simultaneou s ly gauge effects of, and interactions between, performance factors. Ot her studies exami ne effects of subjective dissatisfaction rather than “actual/objective” pe rformance; such analyses are worthwhile but open to suspicions of endogeneity. Finally, few have simultaneously analyzed welfare-related outcomes (s uch as inequ ality) and welfare state pol i ces (such as benefit generosity). In contrast, drawing on the emerging literature on “policy feed back” (e.g. Soss and Schram 2007) I raise the possibility that welfare state ramificati ons for democratic dissatisfaction may reflect attention to policies themselves, rather than to policy outcomes. I then proceed to three-l evel analyses Euroba rometer surveys across 25 yea rs (Schmitt et al. 2005). These suggest generosity in unemployment bene fits (but not pensions , sick pay, or income inequality) helps explain over-time within-country variation in “satisfaction with democr acy,” while controll i n g for macroeco n o m i c factors. This positive effect is relatively stable across individuals with different interests and values , but is conditioned by unemployment rates. Specifically, the results support a “visible costs hyp othesis” predicting weaker generosity effects when more people are out of work. Dissatisfied democrats and macroeconomic performance: a weak relationship? This study belongs to a broader theoretical fa mily emphasizing policy outpu ts and outcomes as explanations of general political support (shor thand: “government performance”) (see Eas t on 1975). Most empirica l studies have examined macroeconomic factors while ignoring other types of perform a n c e . Still, far-rang i n g conclus i o n s have be en drawn, reflecting a general scepticism about the explanatory value of performance in es tablished democraci es. For example, Dalton (2004:126-7) studied macr oeconomic performance and political trust over time in a large number of advanced industrial democracies and conclude d that “The empirical analyses […] demonstrate the limitatio ns of the performance model […] eco nomic performance, whether measured in 4 objective or subjective terms, does not seem to be a significant contributor to the long-term decline in political support during the later twen tieth century (see also Listhaug 1995). A common explanation for the relative unimpor tance of (macr oeconomic) performance factors highlights repeated exp eriences of electoral acco u ntability. Such experiences teach ci tizens the value of punishing and rewarding incumbents at the polls, rather than blaming the political system more generally. Further, essentially pos itive experiences of accountab ility generates affectively based “diffuse support.” This type of support is by definition insensitive to short-term fluctuations in performance. In McAllister’s (1 999:20 3 ) formulati o n “The politica l economy of confidence of democratic institutions is […] stri ctly limited. This conclusi o n underlie s the gradual transformation that has taken place in the esta blish ed democracies, where the frequency of national elections has slowly generated a reservoi r of popular support for democratic institutions, with citizens drawing a clear distinction between th e institutions of the state on the one hand, and the party and leaders elected to conduct public policy on the oth er.” Other politica l science subfield s , however, increa singly see electoral accountability as fragile. Recent research on “economi c voting,” for exampl e, concludes that accountability is highly unstable and variant across time and space (for recent overvi e w s , see Anderso n 2007; Duch 2007; Lewis-B e c k and Stegmai e r 2007). Specifi c a l l y , the economy has stronger effects on the vote under “clarity of responsi b i l i t y ” (Powell an d Whitte n 1993; Anders o n 2000; Taylor 2000; Bengtsson 2002; Nadeau, Niemi, and Yoshinaka 2002) and strong “competency signals” (Duch and Stevens on 2008). Unfortunately, most institu tional and contextual con ditions conducive to these values —such as single-party majority government, long period of incumbency, clear government alternatives, centralized governme n t etc.—are unu sual not least in Europe. Conversely, Taylor (2000) found that that the economy has weaker effects on democratic dissatisfaction under the same rare con ditions. Thus, citizens dissatisfied with government 5 performance appear to value accounta bility bu t get frustrated with the functi oning of the democratic system when this value is obfuscated, i.e. when they cannot hold any specific actor to account. Accountability may be particularly frag ile in the welfare state domain. Several studies find that policymakers build blame-avoidance in to the desi gn of retrenchment policies themselves (e.g. Pierson 1994; Lindbom 2007). For example , r etrenchment is more likely to occur in an incremental and hard-to-detect fashio n, in areas were client organisations are weak, or where retrench m e n t can occur as the result of non-deci s i ons (i.e. failing to adjust benefit ceilings and floors upwards with inflation) and opaque tinkerin g with eligibility criteria. Moreover, studies of voter behavior conclude that accountability in th is domain is (even) weaker and more variable than in the macroeconomic realm. One study find s that dissatis f action with salient and supported services such as education and health care had significant effects on government voting in only four of nine analyzed West European elections. Th ese effects were system ati cally conting ent on institutional clarity of responsibility, and usua lly weaker than the impact of macroeconomic perceptions (Kumlin 2007). Similarly, actual cuts in replacemen t rates are inconsequential for government survival in Western Europ e unless cu ts are very large and recen t, or extensively covered in election camp aigns (Armingeon and Giger 2008). In sum, then, it seems unsafe to simply a ssume positive experien c e s of well-fun c t i o n i n g accountability in the welf are state domain. The clinical distinction between specific incumbents and “politic i a n s ” and “politic s ” more generall y se ems at least potentially pr oblematic. Therefore, experiences of poor welfare state performance ma y be generalised beyond incumbents even in established democracies. The next section re views research on this particular topic. Dissatisfied democrats, policy feedback, and the welfare state 6 Political behaviour research is often divided into three paradigms. There is the “sociological” tradition, focusing on group socializati on and communication; there is the “psychological” tradition looking more to individual values and identifications; and there is the “economi c” tradition con centrating on self-interest and rati ona lity. In spite of obvious differences, all three highlight factors largely exogenous to political inst itutions and public policies. There has been less room for “policy feedback,” that is, the possibility that the groups, values, and interests etc. are in turn shaped by results of previous democratic pr ocesses. As Mettler and Soss (2003:1) argue in a programmatic article, “aside from some notable ex ceptions, political science has had little to say about the consequences of public policy for demo cratic citizen s hip” Quite such a harsh verdict may no longer be fair, however. Policy feedback id eas have recently been applied in empirical studies on political participation (e.g. Soss 19 99; Mettler 2002; Campbell 2005; Soss and Schram 2007), welfare attitud e s (e.g. Mau 2003; Jæger 20 06; Larsen 2007; Svallfor s 2010a), and social capital (e.g. Kumlin and Rothstein 2005) But only a handful of studies have examined how welfare state-related performance and policies affect general confidence in democrati c processes and institutions. Interestingly, the studies that exist assign greater weight to “performance” compared to the literature on macroeconomi c variables. One line of inquiry is offered by case studies of specific countrie s and historic a l phases. These often conclude that dramatic increases in the proport i o n of “dissa tisfied democrats” is preceded by a whole pack age of poor performan ce involving a recession, rising unemployment and budgetary imbalances , unpopular public sector cutbacks, an d rising inequality and poverty rates. Newto n (2006:860) examined th e four most extreme cas e s of trust decline among established democracies– Finland, Sweden, New Zealand, and Japan. A caref ul analysis of sequences of events revea l ed that “It is striking that all four countries experienced real pr oblems of economic and political performance […] real world problem s caused citizen s to revise their 7 political opinions, and when these wer e (partially) solved in Finland, Sweden and New Zealand, the political mood became more posit iv e and support i v e ” (c.f. Holmber g 1999). Case studies are clearly valuable, but have difficu lties disentangling effects of different kinds of performance. Are citizens reacting against macr oeconomic problems per se, against welfare state retrenchmen t, or against the social outcomes of recessio n s and retrench m e n t , such as growing inequality an d poverty? Do such factors interact with each oth er so as to increase or suppress each oth er’s impact? Moreover, case studies run the risk of emphasizing the peculiarities of a situation. Dalton (2004:46-7) notes tha t “the nati on al literatures often link the trends to the unique historical experiences of the nation. In Britain, for example, the decli ne is linked to economic str uggles of the nation; in Canada, it is linked to the fractious regional conflict; in Austria to the collapse of the Social-Liberal cons ensus.” Case studies, then, may reveal the impact of a dramatic country-specific crisis, but may not do justice to performance factors under s tood in a more systematic sense. Other studies have gauged the individu a l - l e v e l impact of subjective evaluatio ns of performance. Analyzing ei ght European countries, Huseby (2000) found that negative eva luations of performance in elder care, health care, job- an d social security all negatively affect attitudes towards the functioning of democracy, but not sup port for democrat i c principl e s (see also Miller and Listhaug 1999; Roller 1999). 1 Similarly, examining fifteen countries Kumlin (2007a, 2009) report that dissatisfaction with health and educatio n services hampers national political trust in all examined co untries, and breeds euroscepticism in most of them. Public service dissatisfaction has stronger effects in these regards than dissatisfaction with the economy. 1 Simil a r findi n g s were reache d by Miller and Listha u g using Norweg i a n and Americ a n data, and by Roll er who found that former Eas t Germ ans’ comparisons between the com munist and postc ommu nist welfare stat e were unf latt ering for the latte r, whi c h in tur n had nega tive cons equenc es for politi cal trust. 8 Studies of subjective eval uations add pieces to the puzzle but also leave questions open. Do performance evaluations really drive mistrust in a causal sense or are th ey merely projections of attitudes such as political mistrust ? Are evaluati ons systema tically driven by “actual/objective” trends or best understood as “constructions” inspired perhaps by political discourse and idiosyncratic interpretatio ns with little basis in common pattern s across time and countries. Only a few studies have examined the impact of “actual/objecti ve” performance variables. Huseby (2000) found that actual performance plays a role, but was forced to use social spending indicators, which have serious and well-know n drawbacks (see Esping-Andersen 1990; for a recent discus sion, see Scruggs 2008). Anderson an d Singer (2008) reported that greater disposable income ineq uality affect trust negatively among countries from both western and central/ eastern Europe , contro l l i n g for indivi d u a l - l e v el variab les. They also controlled one type of macro performance (inequality) for another (macroecono mic conditio ns). 2 Interesti ngly, macroeconomic conditions came out entirely insignificant once inequality level s were accou nted for. The latter study raises two final issues to be ta ken on board. A methodological remark is that cross-country variation at a single point in time intr oduces institutional, political, cultural variation, much of which cannot be controlled fo r. There should be considerable leverage in analyzing also within-country variation, especially as such variation can be substantial for welfare state outpu ts and outcom es (Brandolini and Smeeding 2008). A second iss ue is that pas t studies 2 The re ar e just a few oth er studies that sim u ltan eous ly consider per for man ce in sev e ral domain s . Among them are Huseby (2000) and Kornberg and Clarke (1 9 92). This is unfortu na t e as perform a n c e across different polic y domai ns is likely to corr elate. Thus, cons idering their effects un der contro l for each other is necess ar y to avoid spurio u s inter pr e t a t i o n s and to reac h a fair er verd ict on the “pe rform ance model.” For ex ampl e, macr oe conom i c downturns are to som e ex ten t like l y to be corr e l at e d with many other type s of gove rnm en t perf or ma nce nega tively. Al so, both welf are state gener osity and inc o me inequali ty are likely to be aff ec ted at least in sev e re rec essions. 9 concentra te on policy outcomes rath er than on polit ically contr olled instruments meant to affect such outcomes. This is true for studies of ma croeconomic performance such as unempl oyment and growth rates as well as for Anderson and Singer’s study of inequality. The latter schol ars, however, touched on the distinction in choosing to study “inequality in disposable incomes—that is, posttransfer incomes —rather tha n market inco mes or wea lth […] beca use they are shaped by both the market and the state and thus should be closer to how voters evaluate democra tic institutions than pretransfer levels of income would be” (2008:5 7 8 ) . Still, one may object that disposable inequality blur market inequ ality and ma r k et-correcting policies. We are still left with the question of whether citizens “read off” welfare state performance by perceiv i n g aggrega t e d outcomes su ch as inequal i ty (Anderso n and Singer 2008), or mainly by observing more directly the politicall y controlled policies that affect such patterns? An answer was proposed by Soss and Schram (2007) in their case study of America’s AFDC/T A N F reform . This reform introd u c e d , am ong other things, stronger work incentiv e s and stricter eligibility criteria for recipients. Imagined “policy feedback” effects on atti tud es towards reci pients and welfare were part of it s political rational e. However, “Work requirements and time limits may be popular, but they did not generate more positive images of poor people, welfare recipients, or welfare itself.” The proposed explanation for absent feedback is that while policies received massive media attenti on at their inception, subsequent policy effects on recipient beh aviour (viewed by many experts as a success) did not. Lacking attention to outcomes, th e public was rather affected by information and symbols surrounding policies themselves, rather than by societal policy outcom es. This argument, then, reinforces the need to simultaneous ly consider distant policy outcomes as well as actual policies. What now? Contribution, data and measures 10 T h i s paper sustain s the emergin g researc h progra m on welfare state performance, while aiming at progre s s in severa l regard s . It adopts a large- N approach that models over-time within-co u ntry variation and allows simultaneous cons ideration of different types of performance variables. We avoid subjective evaluations and empl oy measures of actual/objective welfar e variables. Among these variables one finds polices as well as outcom es. Specifically, I combine contextual data on perfor mance with microlevel Eurobarometer surveys. 3 Data from the most often repeated sur vey items have been compiled in the “Mannheim Eurobarometer trend file” (see Schmi tt et al. 2005). This data set is by far the most encom pas s i n g and suitable given our purposes, but it still only contains one indicator of democratic dissatisfaction: “On the whole, would you say tha t yo u are very satisfied, fairly satisfied, not very satisfied, or not at all sati sfied with the way de mocrac y is function i n g in [COUNTRY ] . ” Critics of this ubiquitous item point out that its precise mean ing is unclear. Does it measure overall democratic performance, trust in specific institut ions, trust in politicians, support for democratic principles, or some mixture of these (Canac he , Mondak, and Seligson 2001; Linde and Ekman 2003)? Its defenders agree but maintai n its usefulne ss as an overall measure of subjective political support (Klingemann 1999; Anderson 2002; Blais and Gélineau 2007). Given the interest here in broader generalizations to attitudes th a t transcen d specific incumbents, my position is that finer distinctions would be des irable but are not absolu tely crucial as long as the measure captu r es much of the broad catego r y of general dissati sfaction with the fu nctioning of politics and democracy. 3 E ur o b a r om et er surv ey s are biannual opinion polls conducted on behalf of the Europe an Commission in all member states. The Mannheim Tr end File was provided by The Central Arch ive for Empirical Social Research (ZA) in Cologne 11 The macro data were taken from the Quality of Government Institute’s Social Policy Data Set (Samanni et al. 2008). In turn, this data set dr aws annual unemployment levels from OECD data as taken from the “Com parative Political Data Set 1960-20 0 6 ” (Armin g e o n et al. 2008), and GDP growth levels from Eurostat ( http:// ec.europa.eu /eurostat ) and “Penn World Table” (Heston, Summers, and Aten 2002). GINI coefficients meas urin g disposab l e income inequali t y come from the Luxembourg Income Study (www.lisproject.org; for an analysis, see Brandolini and Smeeding 2008) . Welfare state policies, finally, are represented by three variab l e s from Lyle Scruggs’ “Comparative Welfare Entitlements Dataset” (see Scruggs 2008). These variables represent the perhaps mos t ambitious attempt to tra ck welfare state develop ment across time and cou ntries. They are ava ilable on an annual basis for 11 Eurobarometer co untries (Austria, Belgium, Denmark, Finland, France, Germany, Ireland, It aly, the Neth erlands, Sweden, and the UK). Immediately inspired by Esping-Andersen’s (199 0) decommod i fication index, the three indices register a number of generosity aspects of unempl oyment benefits, sick pay, and pensions. These aspects inclu de replacement rates for different household types, qualifying period length, benefit duration, waiting days, coverage ratios, and (for pensions) minimum and standard replacement rates, as well as proportion of retirees receiving benefits. The resulting indices, displayed in Figure 1, take all such features into account (see Scruggs 2008 for details). The annual nature of these data all ows us to tak e advantage of the over-time component of the Eurobarometer surveys in ways tha t are explained next. Findings: The importance of unemployment benefit generosity Table 1 displays estima tes of multilevel models with three hierarchically nested levels: individuals nested in years nested in countries. The dependent var iable is satisfaction with democracy, ranging from 1 to 4, with higher values indicati n g gr eater satisfaction. Model 1 is an “empty” variance 12 components model that lacks independ ent variable s . Instead, it only estimate s a universa l intercept tog ether with one random error term for ea ch of the three levels (i=individuals, j=years, k=countr i e s ) . 4 Model 1: Satdem i j k = α + e ijk + u jk + v k Model 1 is interesting as the variation of the erro r terms hint at the hierarchi cal causal origins of satisfaction with democr acy. Of course, a precond it ion for pursuing contextual effects is variation at the parti cu l ar level in question. Characteristica lly for survey data, much of the overall variation can be attributed to individual-level factors (SD= .779 ) . Still, ther e is signific a n t variatio n across countries (.218) as well as across years within countrie s (.156). This three-le v e l nature of satisfaction with democr acy means applying a “flat” single-l e v e l OLS model here give seriousl y biased coeffic i e n t s and standar d errors (Hox 2002; Steenbe r g e n and Jones 2002). [TABLE 1] Model 2 introduc e s independ e n t variable s , in cludin g individu a l - l e v e l control variable s , macroecono mic performance aspects such as unemployment rate and growth, income inequality, as well as the welfare ben efit generosity indices. 5 Model 2: Satdem i j k = α + Individual level cont rols i jk + β Unemployment j k + β GDPgr owt h j k + β GINI j k + β Year j k + β Pension generosity j k + β Sick leave gene ros i t y j k + β Unemployment bene fit generosity j k + e ijk + u jk + v k 4 Multi l e v e l model s were es tima t e d using STATA ’ s xtmix e d comma n d us ing the unstructured var ianc e- covari ance option. 5 As the uni ts at level 2 are ti me poin ts I also include a li nea r time vari able at this lev el. 13 A key observ ation is that unemployment benefit ge nerosity, but not pensions and sick leave, has a significant positive effect on democratic satisf a c t i o n , even control l i n g for individ u a l - l e v e l variables and macroecono mic performance. As for latter factors, Model 2 suggests both unemployment and GDP growth have signif icant effects in the expected directions. Model 3 adds LIS data on disposable income ineq uality. This specification plays several roles. First, it tests if Anderso n and Singer’ s (200 8) cross-sectional rela tionship between incom e inequality an d democratic satisfaction can be foun d also in over-time within-country variation. Second, we want to com pare the imp act of ge nerosity policies with those of a central distributional outcome. Now, the prize to be paid for Model 3 is a reduction in the num ber of cases as the inequality data are not ava ilable for nearly as many time points as the others (beginning in the early 1980s in several countries). 6 To at least avoid losing countrie s model 3 drops the most unusual individual-level controls. 7 The results show that unemployment benefit gene rosity retains its effect also under control for income ineq uality. 8 Pensions and sick leave contin ue to be insi gnificant. Moreover, we see no significant effect of inequality variation. Thus, the negative effects reported by Anderson and Singer (2008) are not present here. Finally, their results suggested macroeco nomic performance is 6 Ano the r poi n t conce rni ng the num ber of cases con cer n s the fact that models 2 and 4 only 11 countries at the hi ghes t level. Admit te d ly, this is a bit on the low si de but should work as long as we are not inter ested in estima ting effe cts of varia ble s at this lev el. Howev er , to be on th e safe side I have also estima t e d all models as tw o-le v e l models (indiv i d u a l s in count r y - y e a r s ) contr o l l i n g for cou ntry dummies . This ope rati on yielded the same main obs ervati ons and conclusions. 7 Analys es incl uding the m (based on 9 cou ntries; no t sho wn here) indic ate that result s and int erpr e tati ons remai n larg ely the same. The exce pti on is that the im pa c t of une mpl oy men t levels remains subs ta n tively strong it drops below statis t i c a l signif i c a n c e . 8 I have als o es tima ted model s including GDP/ca pita bu t drop ped this variabl e as it had no imp ac t whatsoev er controlling for the vari ab l es that are included in Table 1. 14 w h o l l y inconse q u e n t i a l under contro l for inequality. I find this to be the case for growth but not for unemployment whi ch , if anything, takes on a s lightl y larger effect in Model 3. Thus, a higher unemployment rate negati vely affects democratic satisfaction even controlling for whether the individual is unemployed or not. Individual-level variation in contextual effects of generosity and inequality? An objection to table 1 is that overall contextual effects could mask variation among individuals. In fact, a cou ple of studies do suggest that these could be contin gent on political values as well as economic interests. As for values, Anderson and Singer (2008) uncovere d a cross-le v e l interact i o n between macro-level inequality and individual-lev el ideology, with political trust being more affected by inequality among leftist citizens; still, there was a significant negative effect also among non-lef t i s t s . General left-ri g h t positio n is relevant, not just for inequality, but for the evaluation of welfare generosity. Thus, below I follow Anderson and Singer’s example and let this variable interact also with the impact of generosity on democratic satisfactio n. As for interests, Oskarson (2007) found that a sim ultaneous co mbination of high individual “social risks” and welfare state retren chment is especial l y conduciv e to “politic a l alienati o n . ” Now, the Eu robarometer trend file does not allow tapping risks and interests with great precision. What we can do, however, is analyz ing interactions with broader demographic variables such as income (in the cas e of in equality), unemployment (for unemployment generosity), and age (for pensions). It is possible to formulate different expectations on such intera ctions. On the one hand large majorities in most of these countries have suppor ted at least basic state responsibility for welfare policies throughout the studied period (Edlund 2009). This would imply that most groups react (more or less) positively to greater benefit genero sity. On other hand, issues of welfare benefits 15 and income distributions also pit interests and id eo logies against each oth er. After all, some groups benefit more than others, and some suppor t welfare policies more than others. Taken to its extreme, this could even imply that, say, un empl oyment generosity effects have different signs depending on interests or values; i.e. that more generosity simultaneously makes, say beneficiaries or leftists more satisfied with democr acy, and ta xpayin g non-bene f i c i a r i e s , perhaps with a rightist value orientation, less so. But I find rather few traces of such empirical drama. Effects are only inconsistently and very mildly structured by values and interests. This conclusion was reached by adding, one at the time, multiplicative cross-level interactions to Models 2 and 3 (not shown in tables, reported in text). The general form of the equations, which includ e level 2 variatio n in slopes of the level 1 interaction variable, can be expressed like this: Cross-leve l interaction models: Satdem i j k = α + Individual level cont rols i jk + β Unemployment j k + β GDPgr owt h j k + β GINI j k + β Year j k + β Pension generosity j k + β Sick leave gene ros i t y j k + β Unemployment bene fit generosity j k + β Individual-level variable i j k X Cont ext ua l variab l e jk + e ijk + u jk + v k + u j; β Indiv i d u a l - l e v e l vari a b l e First, adding a cross-level interaction between left ist ideology and inequali t y to model 3 yields no significant coefficient for this multiplicative term (p-value for b ineq ua l i t y x unemp lo ye d =.98). Neither is the impact dependan t on the income of responde n t s (p-value for b ineq ua l i t y x inco me =.34). By the same token, the positive impact of unemployment gener osity is stable across broa d interest an d value groups. For example, an interaction wi th leftist id eology to model 2 yields nothing significant (p- 16 v a l u e for b unemp l oy m en t gene r os i t y x left =.42). 9 The same is tr ue for being unemployed (p-value for b unem pl o y me n t gener o s i ty X unempl o ye d =.44) and for income (p-value for b unempl oy m en t gene ros i t y X inco me =.14). 1 0 Moving to pensions, there is indeed a weak tend en cy for generosity to yield democratic satisfaction among leftists (b pen si on gene r os i ty X left =.022; p=.000). Thus, in this case the insignificant overall impact in Table 1 masks a positive eff ect among an important political subgroup. But even among leftists the pension effect is only a bout half the str onger and more universal impact of unemployment genero sity. Furthermore, there is not a more positive pension effect among the old. 1 1 Moving finally to sick pay, the data set does not contain pr oxies for health risks and individual sick leave. What we can do, therefore, is to investigate if sick pay generosity is more consequ e n t i a l among leftist s . The results sugge st this is not the case (p-value for b sick pay gene ro s i ty X left =. 3 0 ). Unemployment rates and the benefit generosity effect: visible interests or visible costs? Thus far we can conclude that especial ly unemployment genero sity is a welfa re state featu r e that has systema ti cally affected democracy satisfaction across time and space in Western Europ e. Moreover, the impact is rather similar across broad ideological and socioeco nomic groups. With so much time and space, however, it is still not necessarily the case that we have a context u a l l y “monolithic” causal factor at hand. Therefore, th is sections tes ts two hypoth eses about a possible contextual source of variation in the effect. 9 Consis t e n t with Anders o n ’ s and Singer’ s (2008) coding I use a dummy that takes on the value 1 if the respo nde nt chose pl aced her s elf on one of the thre e first poin ts 1 0 Here, ther e is in fac t a ver y mild and non- significa nt ten d ency for hig her-i nc ome gr oups to rea ct somew ha t more posi tively tha n others to great e r bene f i t gene r os i ty (b un e m pl o y me n t gene ros i t y X inco me = . 0 0 3 ; p=.13). 1 1 In fac t, there is an unex pe cted but very mild tenden cy for pensio ns to mat ter less amon g the old (b ge n e ro s i ty x age =-.0 0 0 4 ; p=.00 0 ) . 17 Expressed generally, we are interested in how the prevalence of an underlying social problem/risk interacts with generosity in policies insuring citizens against it. Specific a l l y , we pit two hypotheses concerni ng unemployment generosi ty against each other. The first one draws on the reoccurring finding that unemployment an d cri sis drives up support for unemployment protection. This has typically been exp lained by a mix of increasing salience of unemployment as a personal an d societal problem, as well as incr easing sympathy for the seem ing victims of circumstance (Blekesaune and Quadagno 2003). Al l this would lead one to expect a positive interaction, with benefit generosity assuming a st ronger positive effect on democratic satisfaction at higher unemployment rates. This prediction ma y be called the “visib l e intere s t s hypoth e s i s . ” Expressed generally, as a socioecono mic risk beco mes more common and visible, the benefits of social protection against it become more sa lient and visible, and appear more just. But one may also imagine mechanisms working in the opposite direction. They can be summed up in a “visible costs hyp othesis”. The key here is that high unemployment may not only drive up welfare support and visibility, but also public expenditure and debt. This may in turn make different groups of citizens simultaneously per ceive cause for dissatisfaction. Rightists and non- beneficiaries will complain about the state budget be ing in the red and fear realized or potential tax increases. Leftists and beneficiaries, on their part, will worry about potentia l cutbacks in terms of replacement rates or benefit eligibility. All ca n be united, however , in accusin g politic i a n s for creating a welfare system that currentl y seems unaf fordable and less viable than under lower unemployment rates. Put differently, higher unem ployment may prime citizen s ’ attention to costs and problems associated with benefit generosi ty. Therefore, the normally positive effect of generosity may becom e weaker as unemployment rises. 18 Table 2 introduces a multi plicative interaction term between the unemploy ment rate and benefit generosity. The key obser vation is the significan tly negative coefficient for this term (-.005; p=.001). This is in line with the “visible costs hypo thesis” rather than the “visible interests” hypothesis. Now, this does not exclud e the possib ility that also “visible interests” mecha nisms are at work. However, the negative intera ction does su ggest that the former process is on balance more powerful . [Table 2 and Figure 2] To facilitate interpretation, the joint influence of unemployment and benefit generosity is plotted in figure 2 with other variables at their mean s. First note the gen erally negative effect of unemployment. At mean levels of generosity, a noth er 10 percent of unemployment is predicted to reduce satisfaction by about .20 along the four-poi n t scale. Judging from the tables, furthermore, this roughly equals the impact of becoming unemployed oneself. Moving to the interaction, benefit generosity has a clearly positive effect at five percent unemployment. At this level, moving from the lowest recorded generosity of around 2 (Italy during the 70s and early 80s) to the hi ghest values of around 13 (Scandin a v i a during early 90s) is predicted to enhance democratic sati sfaction by around one standard deviation (.80). As unemployment increases to crisis proportions of 15 percent, ho wever, the ef fect is predicted to have shrunk by around two-thir d s . At the apoc alyptic unempl oyment rate of 25 percent, the effect of greater welfare generosity is even pred icted to be slightly negative. Admittedly, this extrapolation is a bit of a stretch as Eu rope has mer cifully experienced few su ch situations. But it is good for heuristi c purposes as it illustra t e s the main point: the legitimacy-building role of welfare policies is found mainly when the probl ems policies alleviat e stay within normal and affordable limits. 19 Conclusions: Dissatisfied Democrats and the Nature of Welfare State Feedback This paper has linked research on the welfare state in general, and “policy feedback ” in particul a r , with research on political support. On the one hand , the findings confirm suspicions that welfare state variables can affect democratic dissatisfa ction. On the oth er hand, the findings offer progress in several respects. By examp l e, it is a pparently not onl y possible to explain historically large confidence crises wi th reference to a time specific mix of poor performance (i.e. Newton 2006). One can apparentl y also explain general wi thin-country over-time variation in democratic dissatisfaction using systematic measures of bene fit generosity. This finding, moreover, supports the causal rel evance of subjective eval uations of “personal social protection,” and the like (i.e. Huseby 2000). Perceived malperformance does not seem to be entirely endogenous to democratic dissatisfaction, or only rooted in soci al constructions and idiosyncratic urban legends. At the same time, the res ults go some way towards specifying the aspects of actual social protection th at are universally (un)imp ortant across time and space. 1 2 In particu l a r , less generou s unemployment benefits appears to be a univer sal generator democratic diss atisfaction across broad groups in Western Europe. All this is to say that studies on macroeconomic factors may have been too quick in dismissing performance-type factors more broadly conceived (i.e McAllist e r 1999; Dalton 2004). The results suggest there is no simple master variable, or policy domain, tha t subsumes all relevant performance. Rather, wha t we need seems to be simultan e o u s consider a t i o n of several policy domains (i.e macroecono mics and the welfare state) . Part of this is the simple observation that different policy domains may have dir ect main effects. A finer case in point concerns the interactive in terplay between social pr otection and the prevalen ce of underlying risks and costs. 1 2 This observ at i o n , it should be noted, is a co ntra s t to most the seriou s worries about subjecti v e perform a n c e evaluat i o n s th at are sometim e s voiced (van der Brug , van der Eijk, and Frankli n 2007). 20 On the one hand, Figure 2 implies that one of the better vaccines against democratic dissatisfaction is a combination of generous un employment benefits and low unemployment. However, generous benefits can lose their is bene ficial impact, or even contribute to a toxic mixture, if coupled with extreme unemployment an d associated costs (“visible costs hyp othesis”). Put differently, the long-term rise in unempl oyment in Western Europe has not only likely assisted in the birth of dissatisfied democrats directl y, but also indirectly by di sarming the previously legitimizing force of unemployment benefits. Findings such as these encourage more bridges b etween the vas t welfare state and political trust literatures. Future studies may wan t to look at policy areas such as health car e, elder care, and public educa tion. Indeed, Huseby (2000) found th at subjective evaluations of care for the elderly mattered mo re for political trust comp ared to sever al other subjective evalua tions. Likewise, Kumlin (2007) found that subjective evaluations of health care and education mattered more than economic ev aluations. Finally, I was not able to echo Anderson and Sin ger’s (2008) finding that disposa b l e income inequality gives birth to dissatisfied democrats. The generality of their finding is thus an open question for future research. To be fair, the deviations could be partly due to the fact that these data cover Western Eur ope over a ti mespan of several decades , whereas Anderso n and Singer’ s cross-sectional results included only recent years and some centr al and easter n European countries. Perhaps inequality did play a genuinely causal role in those countries at tha t tim e? But there is also methodol ogical and substantive roo m for doubt. Beginning with methods, Anderso n and Singer used cross-n a t i o n a l variati on across both old and new democracies. While this is a perfectly legitima te strategy it is alwa ys har d to ensure that relevant spurious macro factors are controll e d in a small but highly variab le country sam ple. On a more substantive note, 21 several studies cast doubt on the idea that “policy feedback” in the welfare state domain is driven by citizens perceiving and drawing political c onclusions from aggregated social policy outcomes (Soss and Schram 2007) such as overall inequali t y levels (see Kumlin and Svallfor s 2007). While citizens are by no means totally in the dark abou t things like wage differe n c e s (see Aalberg 2003), the results reported here suggest they are better still at monitori n g broad features of redistri b u t i v e policies themselves (such as unempl oyment bene fit generosity). This brand of policy feedback, then, may be an influential basis for citize n s ’ reasoni n g about how “fair,” “unequa l ” or “satisfactory” their welfare state has become. 22 0 5 10 15 20 0 5 10 15 20 0 5 10 15 20 1980 1990 2000 2010 1980 1990 2000 2010 1980 1990 2000 2010 1980 1990 2000 2010 Austria Belgium Denmark Finland France Germany Ireland Italy Netherlands Sweden United Kingdom Pensions generosity Sickness insurance generosity Unemployment insurance generosity Year Note: Data from the Welfare State Entitlements Dataset (see Scruggs 2008) Figure 1. Welfare state benefit generosity 23 Table 1. Multilevel models of satisfaction with democracy (three levels; ML estimation) Model 1: Model 2: Model 3: FIXED PART Individual level variables: A ge ijk .0002** -.0006*** W o m a n ij k -.02*** -.02*** Leftist ideolog y i j k -.23*** -.16*** Income i j k (cou nt r y -year z-scores) .05*** .05*** Unemployed i j k -.19*** -.24*** Political persuasion i j k ( 1 – 4) -.02*** Political discussion i j k ( 1 – 3) .07*** Low educ at ion i jk -.05*** D i v o r ce d/ se p ar a te d i j k -.10*** U r b a n resident i jk -.04*** Year level variables: Year j k (0=1980) .008** . 007** Unemployment rate j k -.015*** -.019** G D P growt h j k .01** -.004 Pension gene rosit y j k .002 -.003 S i c k leave genero s i t y j k -.01 -.01 U n e m p l o y m e nt benefit gene ros i t y j k .044*** .053*** G I N I j k (0-1) -.69 RANDOM PART Individ ua l leve l: Standa rd deviati o n of e ijk .783*** .771*** .768*** C o u n t r y -year level: Standa rd deviation of u jk .131*** .105*** .135*** Country level: Standard deviation of v k .234*** .151*** .120*** No. of countries 11 11 11 N o . of coun tr y -years 159 159 40 No. of respondents 226,236 226,236 69,442 Overall time frame 1976-2001 1976-2001 1979-2000 *p<.10 ** p<.05 *** p<.01 Notes : Unweigh t ed data from The Mann h eim Eurob a r o m et er Tren d File. The models also cont a in inter cep t s , as well as resid ua l correlatio n between erro r term s at lev el 2, the estimates of which are not disp layed here. 24 Table 2. Multilevel models of satisfaction with democracy (three levels; ML estimation) FIXED PART Individual level variables: A ge ijk .0002** W o m a n ij k -.02*** Leftist ideolog y i j k -.23*** Income i j k (cou nt r y -year z-scores) .05*** Unemployed i j k -.23*** Political persuasion i j k ( 1 – 4) -.02*** Political discussion i j k ( 1 – 3) .07*** Low educ at ion i jk -.05*** D i v o r ce d/ se p ar a te d i j k -.10*** U r b a n resident i jk -.04*** Year level variables: Year j k (0=1980) .008** Unemployment rate j k -.010** G D P growt h j k .01** Pension gene rosit y j k -.008 S i c k leave genero s i t y j k -.006 U n e m p l o y m e nt benefit gene ros i t y j k .053*** U n e m p l o y m e nt benefit gene ros i t y j k X Unemployment rate j k -.005*** RANDOM PART Individ ua l leve l: Standa rd deviati o n of e ijk .771*** C o u n t r y -year level: Standa rd deviation of u jk .101*** Country level: Standard deviation of v k .157*** N o . of countr i e s 11 No . of coun tr y -y e ar s 159 No. of respondents 226,236 Overall time frame 1976-2001 *p<.10 ** p<.05 *** p<.01 Notes : Unwe i g h te d data from The Mannhe i m Euroba r om et er Trend File. The mode l s also conta i n inte rc e pts the estimates of which are not displayed here. Unemployment rate and unemployment bene f i t gene r o s i ty are cent e red over their means. 25 1 2 3 4 S at is fa ct io n with de m oc ra cy 0 5 10 15 Unemployment benefit generosity 5 % unemployed 15 % unemployed 25 % unemployed Note: Predicted values from Table 2 Figure 2. Satisfaction with democracy and unemployment benefit generosity at different unemployment rates 26 References A a l b e r g , Toril. 2003. Achieving Justice: Comparative Public Opinion on Income Distribution. Leide n : Brill . Allan, James P., and Lyle Scruggs . 2004. "Polit i c a l Partis a n s h i p and Welfare State Reform in Advanced In dustrial Societies." American Journal of Political Science 48 (3):49 6 - 5 1 2 . Anderson, Christopher J. 2000. "Economic voti ng and political context: a comparative perspective " Electoral Studies 19 (2-3):1 5 1 - 7 0 ———. 2002. "Good Questio n s , Dubious Inferen c e s , and Bad Solutio n s : Some Further Thoughts on Satisfaction with Democra cy." Binghamton University, Center on Democratic Performance. Working Paper No. 116, Bingham t o n , NY. ———. 2007. "The End of Economic Voting? Contingency Di lemmas and the Limits of Democratic Accountability." Annual Review of Political Science 10:271 - 9 6 . Anderson, Christopher J., and Matthew M. Singer. 2008. "The Sensitive Left and the Impervious Right Multilevel Models and the Politics of Inequality, Ideology, and Legitimacy in Europe." Comparative Political Studies 41 (4):5 6 4 - 99 . Armingeon, Klaus, Marlène Gerber, Philipp Leimgruber, and Michelle Beyel er. 2008. Comparative Political Data Set 1960-2006 . Institut e of Politica l Science, Universi t y of Berne, 2008. Armingeon, Klaus, and Nathalie Giger. 2008. "Conditional Punishment: A Comparative Analysis of the Electo ral Consequences of Welfare State Retrenchment in OECD Nations, 1980- 2 0 0 3 . " West European Politics 31 (3):55 8 - 8 0. Bengts s o n , Åsa. 2002. Ekonomisk röstning och politisk kontext. En studie av 266 val i parlamentariska demokratier. Åbo: Åbo Akedemis förlag. Blais, André, and Francois Gélineau . 2007. "Wi nning, Losing, and Satisfaction with Democracy." Political Studies 55:425 - 4 1 . Blekesau n e , Morten, and Jill Quadagno. 2003. "Public Attitude s toward Welfare State Policies : A Comparis o n of 24 Nations. " European Sociological Review 19:415- 2 7 . Borre, Ole, and Elinor Scarbro u g h , eds. 1995. The Scope of Government . Oxford : Oxford Univer s i t y Press . Brandolini, Andrea, and Timothy M. Smeedin g. 2008. "Inequality Patterns in Western Democracies : Cross-Country DIfferences and Changes over Time." In Democracy, Inequality, and Representation, ed. P. Beramen d i and C. J. Anderso n . New York: Russel Sage Foundati o n . Cameron, David. 2001. "Unemplo y m e n t , Job Cr eation, and Economic and Monetary Union." In Unmployment in the New Europe, ed. N. Bermeo. Cam bridg e : Cambridg e Universi t y Press. 27 C a m p b e l l , Andrea . 2005. How Policies Make Citizens: Senior Political Activism and the American Welfare State. Princeto n , NJ: Princeto n Universi t y Press. Canache, Damarys, Jeffery Mondak, and Mitchell Seligson. 2001. "Meaning and Measurment in Cross-Na t i o n a l Research on Sa tisfaction With Democracy." Public Opinion Quarterly (65):50 6 - 2 8 . Carnes, Matthew E., and Isabela Mares. 2007. "T he Welfare Sta te in Global Perspective." In The Oxford Handbook of Comparative Politics , ed. C. Boix and S. C. Stokes. Oxford. Dalton, Russel J. 2004. Democratic Challenges, Democratic Choices: The Erosion of Political Support in Advanced Industrial Democracies. Oxford : Oxford Univ er s i t y Press. Dalton, Russell J. 2008. Citizen Politics: Public Opinion and Political Parties in Advanced Industrial Democracies. Washing t o n DC: CQ Press. Duch, Raym ond M. 2007. "Comparati ve Studies of the Econo m y and the Vote." In The Oxford Handbook of Comparative Politics, ed. C. Boix and S. C. Stok es. Oxford: Oxford University Press . Duch, Raymond M., and Randolp h T. Stevens o n . 2008. The Economic Vote: How Political and Economic Institutions Condition Election Results. Cambri dg e : Cambridg e Universi t y Press. Easton, Davi d. 1975. "A Re-Assessment of the Concept of Political Support." British Journal of Political Science 5:435- 5 7 . Edlund, Jonas. 2009. "Atti tudes toward s State Or ga nized Welfare in 22 societies: A Questi on of Convergence?" In The International Social Survey Programme 1984-2009: Charting the Globe, ed. M. Haller, R. Jowell and T. W. S. Smith. London: Routledg e . Esping-A n d e r s e n , Gøsta. 1990. The Three Worlds of Welfare Capitalism. Cambridg e : Polity Press. Finseraa s , Henning. 2009. "Income Inequali t y and Demand for Redistribution: A Multilevel Analysis of European Public Opinion " Scandinavian Political Studies 32 (1):94 - 1 1 9 . Heston, Ala n, Robert Summers, and Bettina Aten . 2002. Penn World Table Version 6.1 . Center for International Comparisons at the University of Pennsylvania (CICUP), October 2002. . Holmberg, Sören. 1999. "Down and Down We Go: Political Trust in Sweden." In Critical Citizens. Global Support for Democratic Governance , ed. P. Norris. Oxford : Oxford Univer si t y Press. Hox, Joop. 2002. Multilevel Analysis. Mahwah: Lawrence Erlbau m Associates. Huseby , Beate M. 2000. Government Performance and Political Support. A Study of How Evaluations of Economic Performance, Social Policy and Environmental Protection Influence the Popular Assessments of the Political System. Trondheim: Department of Sociology and Political Science, Norwegian University of Technology (NTNU). 28 J æ g e r , Mads Meier . 2006. "Welf a re Regimes and Attitudes towar ds Redistribution: The Regime Hypothesis Revisited." European Sociological Review 22 (2):157-70. Klingema n n , Hans-Die t e r . 1999. "Mapping Politica l Support in the 1990s: A Global Anal ysis." In Critical Citizens. Global Support for Democratic Governance , ed. P. Norri s . Oxfor d : Oxfor d University Press. Klingemann, Hans-Dieter, and Dieter Fuchs, eds. 1995. Citizens and the State. Oxford: Oxford University Press. Kornber g , Allan, and Harold D. Clarke. 1992. Citizens and Community. Political Support in a Representative Democracy. Cambridge: Cambridge University Press. Korpi, Walter, and Joakim Palme. 2003. "New Politics and Class Politics in the Context of Austerity and Globalizati on: Welfare St ate Regress in 18 Countries, 1975-95." American Political Science Review 97:425 - 4 6 . Kumlin, Staffan. 2007a. "Overloaded or Undermined? Europea n Welfare States in the Face of Performance Dissatisfaction." In The Political Sociology of the Welfare State: Institutions, Social Cleavages, and Orientations, ed. S. Svallfor s . Stanfor d : Stanfor d Univers i t y Press. ———. 2007b. "The Welfare State: Values, Poli cy Preferences, and Performance Evaluations." In The Oxford Handbook of Political Behavior, ed. R. J. Dalton and H.-D. Klingemann. New York: Oxford Univers i t y Press. ———. 2009. "Blaming Europe: Exploring the Va riable Impact of National Public Service Dissatisfaction on EU trust." Journal of European Social Policy 19 (5):40 8 - 2 0 . Kumlin, Staffan, and Bo Rothstei n . 2005. "Makin g and Breaking Social Capital: The Impact of Welfare-State Institutions." Comparative Political Studies 38:339- 6 5 . Kumlin , Staffa n , and Stefan Svallf o r s . 2007. "Soc ial Stratification and Political Articulatio n: Why Attitudinal Class Differences Vary Acr oss Countries." In Social Justice, Legitimacy and Welfare State, ed. S. Mau and B. Veghte. Aldersho t : Ashgate. Larsen, Christian Albrekt. 2007. "How Welfare Regimes Generate and Ero de Social Capital. The Impact of Underclass Phenomena." Comparative Politics 40 (1):8 3- 1 0 1. Lewis-B e c k , Michael S., and Mary Stegmai e r . 2007. "Economi c Models of Voting." In Oxford Handbook of Political Behavior, ed. R. J. Dalton and H.-D. Klingem a n n . Oxford: Oxford University Press. Lindbo m , Anders . 2007. "Obfus c a t i n g Retren c h m e nt: Swedish Welfare Policy in the 1990s " Journal of Public Policy 27 (02):129 - 5 0 Linde, Jonas, and Joakim Ekman. 2003. "Satis faction With Democracy: A Note on a Frequently Used Indicator in Comparative Politics." European Journal of Political Research 42:39 1- 4 0 8 . 29 L i s t h a u g , Ola. 1995. "The Dynamics of Trust in Politic i a n s . " In Citizens and the State, ed. H.-D. Klingemann and D. Fuchs. Oxford: Oxford University Press. Mau, Steffe n. 2003. The Moral Economy of Welfare States: Britain and Germany Compared . Londo n : Routledge. McAllister, Ian. 1999. "The Economi c Performance of Governments." In Critical Citizens. Global Support for Democratic Governance, ed. P. Norris . Oxford : Oxford Univer s i t y Press. Mettler, Suza nne. 2002. "Bringing the State Back In to Civic Engagement: Policy Feedback Effects of the G.I. Bill for World War II Veterans. " American Political Science Review 96 (2):35 1 - 6 5 . Miller, Arthur , and Ola Listhau g . 1999. "Politi c al Performance and Institutional Trust." In Critical Citizens. Global Support for Democratic Governance, ed. P. Norris. Oxford: Oxford Univers i t y Press . Nadeau, R., R.G. Niemi, and A. Yoshi naka. 2002. "A Cross-National Analysi s of Economic Voting: Taking Account of the Politic al Context across Time and Nations." Electoral Studies 21 (403-4 2 3 ) . Newton, Kenneth. 2006. "Politic a l Support: Soci al Capital, Civil Society and Political and Economic Performance." Political Studies 54:846- 6 4 . Norris , Pippa, ed. 1999. Critical Citizens. Global Support for Democratic Governance . Oxfor d : Oxfor d University Press. Oskarson, Maria. 2007. "Social Risk, Policy Dissatisfaction, and Political Alienation: A Comparison of Six European Countries." In The Political Sociology of the Welfare State: Institutions, Social Cleavages, and Orientations, ed. S. Svallfo r s . Stanfor d : Stanfo r d Univer s i t y Press . Piers o n , Paul. 1994. Dismantling the Welfare State? Cambridge: Cambridge University Press. Powell, G. Bingham, and Guy D. Whitten. 19 93. "A Cross-na t i o n a l Analysis of Economic Voting: Taking Account of the Political Context." American Journal of Political Science 37:391 - 41 4 . Roller, Edeltrau d . 1999. "Sozialp o l i t i k und de mokratische Kons olidierung. Eine empirische analyse für die Neuen Bu ndesländer." In Wahlen und politische Einstellungen in Deutschland und Österreich, ed. F. v. Plasse r , O. W. Gabrie l , J. W. Falter and P. A. Ulram. Frankfur t : Peter Lang. Samann i , Marcus , Jan Teorel l , Staffa n Kumlin , and Bo Rothst e i n . 2008. The Quality of Government Institute Social Policy Data Set. Gothenbu rg: University of Gothen burg. Available at www.qo g . p o l . g u . s e . 30 S c h m i t t , Her mann , Evi Scholz, Iris Leim, and Meinhar d Moschne r . 2005. "The Mannhei m Eurobar o m e t e r Trendfi l e 1970-20 0 2. Data Set Edition 2.00.". Scrug g s , Lyle. 2008. "Soci a l Right s , We lfare Generosity, and Inequality." In Democracy, Inequality, and Representation, ed. P. Beramend i and C. J. Anderson . New York: Russel Sage Foundati o n . Soss, Joe. 1999. "Lessons of Welfare: Policy De sign, Politica l Learning , and Politica l Action." American Political Science Review 93 (2):363 - 8 0 . Soss, Joe, and Sanford Schram . 2007. "A Pub lic Transformed? Welfare Reform as Policy Feedback." American Political Science Review 101 (1):11 1 - 2 7 . Steenbe r g e n , Marco R., and Bradfor d S. Jones. 2002. "Modeling Multilevel Data Structures." American Journal of Political Science 46 (1):2 1 8 - 3 7 . Svallf o r s, Stefan . 2010a. "Po licy Feedback, Generational Replac ment and Attitudes to Sta te Intervention: Eastern and Western Germany, 1990-2006." European Political Science Review 2 (1):119- 3 5 . ———. 2010b. "Public Opinion." In The Oxford Handbook of the Welfare State, ed. F. G. Castle s , S. Leibfr i e d , J. Lewis, H. Obinge r and C. Pierson. Oxford: Oxford University Press. Taylor , Michae l l A. 2000. "Chann e l i n g Frustr a t i o n s : Institutions, Economic Fluctuations, and Politica l Behavior . " European Journal of Political Research 38:95-1 3 4 . Torcal , Marian o , and José Ramón Monter o . 2006. Political Disaffection in Contemporary Democracies: Social Capital, Institutions, and Politics . New York: Routle d g e . van der Brug, Wouter, Cees van der Eijk, and Mark N. Frankli n . 2007. The Economy and the Vote. Cambridge: Cambridge University Press.