Introduction
In recent decades, inequality has increased massively in most countries around the world (Atkinson and Piketty Reference Atkinson and Piketty2010). Some authors even consider growing inequality as a fundamental threat to democracy (Piketty Reference Piketty2014). Progressive taxation is a highly effective tool to reduce inequality. Yet, popular demand is an essential prerequisite for an increase in tax progressivity. But when do people demand tax progressivity? Several scholars have argued that fairness arguments have a strong impact on people’s political appetite for progressive taxation (Alesina and Angeletos Reference Alesina and Angeletos2005; Tyran and Sausgruber Reference Tyran and Sausgruber2006; Durante et al. Reference Durante, Putterman and van der Weele2014; Ballard-Rosa et al. Reference Ballard-Rosa, Martin and Scheve2017). If the economic success of rich people is perceived as “undeserved”, demand for tax progressivity increases. Mass mobilisation for war has historically been the main cause of fairness-based preference shifts in favour of progressive taxation (Scheve and Stasavage Reference Scheve and Stasavage2016). However, after an initial period of very high tax rates on the rich in the postwar era, tax progressivity has decreased remarkably in the last 40 years (Kiser and Karceski Reference Kiser and Karceski2017). There are multiple explanations for this, covering tax competition (Ganghof Reference Ganghof2006; Swank Reference Swank2006; Genschel et al. Reference Genschel, Kemmerling and Seils2011; Genschel and Schwarz Reference Genschel and Schwarz2011) and the disappearance of mass warfare (Obinger Reference Obinger2012; Scheve and Stasavage Reference Scheve and Stasavage2012). So, can fairness arguments still play a role in taxing the rich in the 21st century?
This article looks at tax policy preferences in the wake of the financial crisis of 2008 in order to answer this question. I argue that fairness arguments have increased demand for progressive taxation in countries that faced a deep recession. Two factors account for this. First, the crisis has put a spotlight on risky investment behaviour in financial markets. Thus, perceptions of rich people’s wealth as individually deserved – hence, as based on their hard work and merit – have suffered. Second, the role of the state before and during the crisis has raised concerns about institutional deservingness. Crucially, the crisis has increased the salience of regulatory failure and led to discussions about large scale bailout packages. As a consequence, the perception of rich people’s economic success as institutionally deserved has suffered as well. Multilevel models combining microdata from the 2009 round of the International Social Survey Programme (ISSP) with several macroindicators support these claims. The analysis shows that demand for tax progressivity is higher in countries that have experienced a more severe asymmetric shock. Importantly, this effect stems from a strengthened impact of fairness considerations on tax policy preferences. Moreover, I use data from the 1999 ISSP round to check my model’s exogeneity assumptions. Additional robustness checks as well as placebo tests and an analysis of overtime variation in tax policy preferences provide further support for my argument.
The contribution of this article to the literature is threefold. First, the article shows that fairness arguments are still important for progressive taxation. Due to the crisis, perceptions of economic success as “deserved” have suffered and the impact of fairness beliefs on tax policy preferences has intensified. As a consequence, demand for tax progressivity is higher in crisis countries. Although mass warfare has – fortunately – disappeared in the last decades (Onorato et al. Reference Onorato, Scheve and Stasavage2014) fairness-induced appetite for progressive taxation is not dead. Other macrolevel shocks can strengthen demand for tax progressivity as well. Second, the article systematically differentiates between distinct fairness dimensions and their impact on tax policy preferences. It demonstrates how the Great Recession as a macrophenomenon triggered specific fairness dimensions whilst others remained unaffected. Thus, instead of regarding fairness dimensions en bloc, using a more fine-grained typology can help disentangle the connection between fairness perceptions and demands for public policy. Finally, the article also contributes to a growing body of literature that deals with the impact of financial and economic crises on national tax policies (Hakelberg Reference Hakelberg2016; Lierse and Seelkopf Reference Lierse and Seelkopf2016; Swank Reference Swank2016a). Whilst these studies have focussed on tax policy changes on the macrolevel, this article adds the microlevel to the discussion. By investigating how and why tax policy preferences differ between countries, it marks a first step towards better integration of the demand and supply side of tax policies in times of crisis.
The article is structured as follows. It starts by offering a short overview of the role of economic self-interest and fairness perceptions for redistributive policies. Then, it develops the theoretical argument about the impact of fairness arguments on tax policy preferences during the Great Recession. After describing the data sources and the different model specifications of the analysis, the results are presented. The final section concludes by discussing the effects of crisis-induced demand for progressive taxation on the political supply side.
Theory
Economic self-interest and fairness perceptions
The microfoundations of redistribution have been gaining increasing attention in recent years. Most studies dealing with personal attitudes towards redistributive policies look at the determinants of preferences for general redistribution and for social policy programmes. In comparison, attempts to explain preferences for tax policies have been rare. We can differentiate between two major explanatory factors for tax policy preferences: economic self-interest and fairness perceptions.
Traditionally, analyses of the impact of economic-self-interest on redistributive preferences have dominated the literature. Much work has focused on the impact of income status and socio-economic risk exposure on preferences for redistribution (Meltzer and Richard Reference Meltzer and Richard1981; Moene and Wallerstein Reference Moene and Wallerstein2001; Rehm Reference Rehm2009). Studies looking at the impact of income on tax policy preferences are based on the premise that people want to maximise their individual net income. Whilst findings clearly show that higher income leads to a lower demand for tax progressivity (Barnes Reference Barnes2015; Ballard-Rosa et al. Reference Ballard-Rosa, Martin and Scheve2017), the predictive power of income for redistributive preferences varies remarkably between countries (Dion and Birchfield Reference Dion and Birchfield2010; Beramendi and Rehm Reference Beramendi and Rehm2016; Berens and Gelepithis Reference Berens and Gelepithis2018). Authors looking at the influence of (social) risks on individual preferences argue that higher risk exposure increases demand for social insurance (Moene and Wallerstein Reference Moene and Wallerstein2001; Rehm Reference Rehm2009). For example, Rehm (Reference Rehm2011) claims that support for social policy programmes is particularly high if the common risk pool is homogeneous.Footnote 1 Progressive taxation can serve the same function of social insurance, as it mitigates the negative effect of social risks on real income by reducing income differences (Varian Reference Varian1980). Barnes (Reference Barnes2015) differentiates between preferences for the size of taxation (tax level) and the structure of taxation (tax progressivity). She finds that risk exposure leads to higher demand for tax progressivity, whilst it does not have an effect on preferences for tax levels. Although they are quite distinctive in their theoretical expectations, both income- and risk-based explanations share the same baseline assumption; individuals want to optimise their economic outcome. They do so by either maximising their current economic situation (income-based explanations) or by finding an optimal insurance coverage (risk-based explanations).Footnote 2
The general argument of studies that look at the role of fairness perceptions for redistributive preferences is straightforward: if economic outcomes are perceived as unfair, demand for a correction of these outcomes will increase (Alesina and Angeletos Reference Alesina and Angeletos2005). Hence, even the richest members of society might demand more tax progressivity if they perceive the tax system as unfair. Several studies have found strong correlations between individuals’ preferences for fairness and tax progressivity (Ackert et al. Reference Ackert, Martinez-Vazquez and Rider2007; Lü and Scheve Reference Lü and Scheve2016; Ballard-Rosa et al. Reference Ballard-Rosa, Martin and Scheve2017). Importantly, fairness beliefs are closely linked to the process that has led to the status quo (Hennighausen and Heinemann Reference Hennighausen and Heinemann2015): if the previous allocation process is perceived as fair, the socio-economic outcome will be perceived as fair, too. Thus, demand for redistributive taxation will be low (Fong Reference Fong2001; Rowlingson and Connor Reference Rowlingson and Connor2011). We can differentiate between three dimensions of fairness perceptions.
First, people are more likely to regard economic success as deserved (and fair) if they perceive success as independent from socio-economic family background. Hence, if intergenerational mobility is high, demand for market corrective, redistributive measures will be lower (Fong Reference Fong2001; Alesina et al. Reference Alesina, Stantcheva and Teso2018). However, if people believe that wealth and income levels are predetermined by socio-economic origins, the perception of deservingness will suffer. In this case, the procedural dimension – the lottery of birth – is completely based on luck. Therefore, its outcome is perceived as unfair. Unsurprisingly, fairness issues related to family background are highly salient in discourses over the inheritance tax (Beckert Reference Beckert2008).
Second, the role of individual effort and merit are important for fairness perceptions. If people attribute economic success to effort and work performance, they will perceive income differences as deserved (Rowlingson and Connor Reference Rowlingson and Connor2011; Durante et al. Reference Durante, Putterman and van der Weele2014). Contrary to the family background, the individual has an active position in the procedural dimension. Whilst family background is exogenous to an individual’s decision, work effort and performance are not. Therefore, this deservingness dimension focuses on behavioural aspects. By strategic risk-taking, an individual can even incorporate luck into the work process. If economic success is the result of risk-taking, people could still view inequality as deserved. However, this will only be the case if the risks that have been taken can actually materialise. If there is no chance of risk-materialisation, as in the case of moral hazard, the perception of success as a reward for bold risk-taking suffers.
Third, institutional circumstances matter for the perception of deservingness. If the political and economic system of a country is perceived as structurally unfair, inequality will also be regarded as unfair (Hennighausen and Heinemann Reference Hennighausen and Heinemann2015). In particular, the role of the state is of central importance. When a subgroup of the population is treated beneficially by public authorities, fairness principles are violated. In order to restore the “principle of equal treatment”, demands for policies which compensate for previously granted beneficial advantages become stronger (Scheve and Stasavage Reference Scheve and Stasavage2010, Reference Scheve and Stasavage2012, Reference Scheve and Stasavage2016).
One shortcoming of most studies on fairness perceptions is that they do not offer an explanation for why tax policy preferences vary cross-nationally and overtime. If fairness perceptions matter for tax policy preferences, why is their impact stronger in some countries than in others? And why do preferences change? The contribution by Scheve and Stasavage (Reference Scheve and Stasavage2016) is an important exception. The two authors show that mass warfare intensifies the impact of the institutional deservingness dimension. They argue that when a country faces mass warfare, richer people are treated beneficially by the state: they are less likely to face conscription and might profit economically from a higher demand for war-related goods. As a consequence, fairness-based demand for tax progressivity increases. Thus, in short, mass mobilisation for warfare causes “compensatory arguments” to tax the rich (Scheve and Stasavage Reference Scheve and Stasavage2010, Reference Scheve and Stasavage2012).
Although the work by Scheve and Stasavage helps fill a major gap in the literature on fairness beliefs, two important questions remain unanswered. First, their main independent variable – mass warfare – has disappeared. Modern wars are different to traditional wars: they are mainly extra-state or intrastate wars which are fought by smaller armies because of modern war technology (Sarkees and Wayman Reference Sarkees and Wayman2010; Onorato et al. Reference Onorato, Scheve and Stasavage2014). According to Scheve and Stasavage (Reference Scheve and Stasavage2016), the absence of mass warfare in the last decades can explain the demise of progressive taxation. However, we do not know whether this means that fairness arguments have become irrelevant for tax progressivity. The financial crisis of 2008 is a prime example of a massive asymmetric shock other than warfare which might have triggered fairness-based demands for progressive taxation. After all, both warfare and economic crises can cause fundamental structural and political changes (Widmaier et al. Reference Widmaier, Blyth and Seabrooke2007). Second, the perception of institutional deservingness might not be the only fairness dimension that is affected by asymmetric shocks. Yet, a systematic analysis that differentiates between fairness dimensions after such shocks is missing. The financial crisis provides the opportunity to look at the interplay between shocks on the macrolevel and different fairness dimensions.
To sum up, the literature that deals with preferences for redistribution has faced a recent “fairness turn”. For times of mass warfare, our understanding of why the power of fairness perceptions for tax policy preferences varies between countries has improved greatly. However, we still know little about the role of fairness arguments in the absence of mass warfare. In particular, the impact of asymmetric economic shocks might lead to an intensified impact of some fairness dimensions on tax policy preferences whilst other dimensions remain unaffected. As I argue in the next section, two characteristics of the financial crisis have increased the impact of fairness beliefs on preferences for tax progressivity: the role of financial and economic elites in the run up to the financial crisis and the role of the state before and during the crisis. In countries that were hit harder by the crisis, these two factors gained particular public attention.
Fairness arguments and the great recession
The crisis, the rich and the state
My main argument is that the Great Recession has increased demand for tax progressivity. In other words, the Great Recession as a cross-nationally varying factor on the macrolevel has increased microlevel preferences for tax progressivity. Furthermore, I claim that this effect originated in an intensified impact of fairness considerations on tax policy preferences. Fairness arguments have prominently reentered public discussions following the economic downturn in the wake of the financial crisis. The prime example of this is the Occupy Wall Street (OWS) protest movement, which has mainly targeted socio-economic inequality and particularly the role of the richest members of United States (US) society. Even the main slogan of the OWS movement, “‘We are the 99%’, straightforwardly refers to growing inequality induced by the wealth and income development”. According to Bartels (Reference Bartels2013), repealing the 2001/2003 Bush tax cuts was “the most concrete policy issue addressed (insofar as any concrete policy issue was addressed) by the Occupy Wall Street movement” (Bartels Reference Bartels2013, 63).Footnote 3 Directly referring to the OWS slogan, Paul Krugman in his New York Times column went even further and focussed on the richest 0.1% of society: “So should the 99.9 percent hate the 0.1 percent? No, not at all. But they should ignore all the propaganda about ‘job creators’ and demand that the super-elite pay substantially more in taxes” (Krugman Reference Krugman2011).
But how might the financial crisis have sparked demands for fiscal fairness? The crisis has led to an increased public salience of two factors. First, the financial crisis has sparked a public debate about its causes – prominently blaming risky financial investments. The majority of opinions in the public debate has blamed the financial and economic elites, particularly bankers, for the emergence of the Great Recession (Hellwig and Coffey Reference Hellwig and Coffey2011; Bartels and Bermeo Reference Bartels and Bermeo2014). The image of “greedy” bankers shamelessly pursuing risky financial activities to maximise personal wealth has dominated public perception. Financial market activities have been characterised as “a gambling casino” (Sinn Reference Sinn2010, 70) allowing for “skyrocketing financial speculation” (Foster and Magdoff Reference Foster and Magdoff2009, 80). This criticism has cut across political affiliations (Münnich Reference Münnich2016). Although financial speculation contributed to growing inequality before the Great Recession (Volscho and Kelly Reference Volscho and Kelly2012), it was only in the wake of the crisis that financial risk-taking gained public salience (Fourcade et al. Reference Fourcade, Steiner, Streeck and Woll2013). Thus, the discussion about the causes of the crisis has put a spotlight on practices of financial investment and the role of economic elites.
Second, the financial crisis has increased attention on the role of the state before and during the crisis (Comiskey and Madhogarhia Reference Comiskey and Madhogarhia2009; Hellwig and Coffey Reference Hellwig and Coffey2011). Discussions about the role of the state before the crisis focus on regulatory failure. The general argument is that ineffective financial regulation enabled financial market actors to take up systemic risk. The huge economic downturn in 2009 (Figure 1) revealed the external effects that came along with risky financial business models. These economic effects did not solely hit those who previously benefited from financial market practices, but hurt society as a whole. In particular, lower income groups that did not participate in risky financial investments beforehand – simply because they lacked the capital to do so – have suffered from the crisis economically. In the US, relative losses in wealth “were disproportionally concentrated among lower-income, less educated, and minority households” (Pfeffer et al. Reference Pfeffer, Danziger and Schoeni2013, 98). To sum this point up, the crisis has put a spotlight on precrisis regulatory failure which enabled financial actors to take up huge external risks.Footnote 4 Furthermore, public attention on the role of the state during the crisis has concentrated on bank bailout packages. These packages were not only expensive (Reinhart and Rogoff Reference Reinhart and Rogoff2013), but they have also been perceived by many as measures to bail out a richer subgroup of the population (Hacker and Pierson Reference Hacker and Pierson2010). Thus, rescuing struggling financial institutions with public money has become a publicly salient and highly criticised topic (Hellwig and Coffey Reference Hellwig and Coffey2011). I argue that both factors – the role of economic elites and the role of the state – have affected fairness-based demand for progressive taxation.
Average gross domestic product (GDP) growth rates of countries in the sample, 1990–2014.
Note: Data come from the World Bank (2017). Unweighted mean of the 32 countries in the sample.
Fairness dimensions and the financial crisis
To disentangle how the crisis-induced perception of economic elites and the state might have influenced attitudes towards taxation, let us refer back to the three different fairness dimensions. The impact of the lottery of birth on later economic success is a rather stable factor, independent of economic downturns. Neither the role of the rich in the run up to the crisis nor the role of the state stands in a direct connection to the perception of advantages based on family background. Therefore, we have little reason to assume that the crisis has had an impact on this fairness dimension.
Perceptions of economic success as a reward for hard work and merit were affected both by the perception of economic elites and of the state’s activities. First, discussions about the crisis’ causes put a focus on risk-taking on financial markets. When financial risk-taking becomes an important public issue, doubts that inequality simply results from economic elites’ higher work effort will increase. As a consequence, the perception of economic success as a “fruit of one’s labour” suffers and people view inequality as more unfair (Alesina and Angeletos Reference Alesina and Angeletos2005). In other words, wealth is increasingly perceived as exogenously rather than endogenously determined (Fong Reference Fong2001). Second, in principle, wealth that emerges from financial risk-taking might also be perceived as deserved. This would be the case if people view the courage to take high personal risks as an effort – and therefore as endogenously determined. If risks are entirely internalised, there would be no need for compensation. However, the crisis has shown that these risks were not completely internalised. Instead, many high risk-takers were bailed out with public money. If risks cannot materialise, risk-taking becomes a less bold endeavour. Hence, rewards resulting from moral hazard are seen as undeserved. In sum, there is good reason to assume that the impact of this behavioural fairness dimension on tax policy preferences has intensified in crisis countries.
Moreover, states’ policies before the crisis and states’ reactions to the financial crisis are related to aspects of institutional fairness. Regulatory failure in the run up to the crisis enabled financial market actors to take up risks at the expense of society as a whole. Thus, a lack of financial market regulation by the state indirectly favoured rich financial investors. Furthermore, bank bailouts also affected the institutional fairness dimension. When struggling financial institutions were rescued with public money, people may have perceived these bailouts as a beneficial treatment of a specific subgroup of the population. Bailing out risk-takers might therefore create compensatory demands (Scheve and Stasavage Reference Scheve and Stasavage2016). Hence, the role of the state before and during the crisis directly touches upon the institutional fairness dimension. The more severe the economic crisis, the more salient are discussions about the crisis’ causes and states’ reactions to it. As a consequence, the impact of the institutional fairness dimension on tax policy preferences will increase in crisis countries.
Based on these considerations, I expect that the financial crisis has caused a general shift in tax policy preferences. Issues of financial risk-taking by economic elites and the role of the state before and during the crisis have affected perceptions of behavioural and institutional fairness. Therefore, I expect that preferences for tax progressivity are stronger in countries that were hit harder by the crisis. Especially in those countries that faced the biggest asymmetric economic shocks, demand for taxing the rich should be higher. Thus, my first working hypothesis is as follows.
H1: People have a higher demand for progressive taxation in countries that have faced a more severe economic downturn after the financial crisis of 2008.
Yet, Hypothesis 1 could also follow out of pure economic self-interest in times of crisis. Most notably, a stronger economic downturn might just raise demand for insurance via taxation as it increases the risk of becoming unemployed. In addition, experiencing crisis-induced personal economic shocks might influence preferences for redistribution (Margalit Reference Margalit2013). I do not rule out that economic development has an influence on preferences for tax progressivity by changing individual socio-economic circumstances. However, my argument builds upon the influence of fairness considerations on tax policy preferences in the wake of the crisis. Therefore, we would expect that Hypothesis 1 stems from an intensified impact of behavioural and institutional fairness perceptions on tax policy preferences in crisis countries.
H2: The influence of behavioural and institutional fairness perceptions on tax policy preferences is stronger in countries that have faced a more severe economic downturn after the financial crisis of 2008.
Data and models
To test my hypotheses about the impact of the Great Recession on tax policy preferences empirically, I combine microdata from the 2009 ISSP Social Inequality IV round with several macrolevel indicators and analyse it by using multilevel modelling. In total, my sample consists of 32 countries on the macrolevel and 31,331 respondents on the microlevel.Footnote 5 My main dependent variable is the question: “Do you think people with high incomes should pay a larger share of their income in taxes than those with low incomes, the same share, or a smaller share?” Respondents could answer on a five point scale covering “much smaller share”, “smaller”, “the same share”, “larger” and “much larger share”. I recode the variable so that it ranges from 1=“much smaller share” to 5=“much larger share”. In comparison to other studies on preferences for tax progressivity, this measurement has the advantage that it does not ask people for their opinion in relation to the current tax system (thus, whether they think taxes on the rich are too high/low). Instead, it directly asks for general attitudes towards progressive taxation. I treat the values of the variable as metric.Footnote 6
To capture the different dimensions of deservingness, I include three items from the ISSP as independent variables. To cover the impact of deservingness based on family background, I use the question: “Getting ahead: How important is coming from a wealthy family?” Answers can range from 1=“Not important at all” to 5=“Essential”. Thus, the higher the variable’s values, the stronger is the perception that family background determines socio-economic success. I expect preferences for tax progressivity to be higher when the status quo is perceived as more unfair. Behavioural deservingness is measured by the question: “How well he or she does the job – how important should that be in deciding pay?”, where answers can again range from 1=“Not important at all” to 5=“Essential”. Here, higher values indicate stronger preferences for a congruence between performance and payment. I therefore expect demand for tax progressivity to be stronger as well. Finally, to measure the impact of the institutional deservingness dimension on tax policy preferences, I include the statement: “To get all the way to the top in <Respondent’s country> today, you have to be corrupt”, to which people could agree from 1=“Strongly disagree” to 5=“Strongly agree”. Admittedly, this operationalisation is far from perfect as it focusses on corruption. However, since it directly captures the perception of an important part of structural (un)deservingness in the economic and political system, it still constitutes a valid indicator for the institutional dimension. Higher values mean that the institutional set-up is perceived as more unfair. Consequently, I expect preferences for tax progressivity to increase with higher values. In sum, all three dimensions are measured on a scale from “1” to “5” and I expect all coefficients to be positive.
My main economic variable on the macrolevel, the degree to which a country has been hit by a crisis economically, is measured by real gross domestic product (GDP) growth rates in the year 2009. Data come from the World Bank’s National Account Database (World Bank 2017). GDP growth in the year 2009 is particularly suited to measuring the extent to which the economic crisis hit a country because the economic effects of the Great Recession were the most pronounced in this year. Therefore, the differences between those countries which were hit by the crisis vis-à-vis those which were relatively unaffected by it became clearest. Furthermore, economic growth rates on the country level are a very visible indicator for a general nationwide economic downturn. Figure 1 shows the average real GDP growth rate for the 32 countries in my sample. Although GDP growth already dropped from 5.1% in 2007 to 1.4% in 2008, the year 2009 marks the low point as GDP shrunk by 4.3% on average. In line with Hypothesis 1, I expect people to have higher preferences for tax progressivity in countries with a lower GDP growth in 2009. I only include those countries in the analysis in which the fieldwork exclusively took place in 2009/2010.Footnote 7
Additional to these main variables of interest, I include a battery of covariates into my models. On the microlevel, I control for several variables that are likely to influence individual attitudes towards tax progressivity. Since people with a higher income might demand less tax progressivity simply because they want to pay less taxes, I include a measurement of household income into my analysis (Kenworthy and Pontusson Reference Kenworthy and Pontusson2005). As income is not directly comparable in the ISSP, I follow common practice by looking at the relative position of income earners in a country (Barnes Reference Barnes2015; Alt and Iversen Reference Alt and Iversen2017). This is done by assigning observations to the country-specific income deciles. I expect people with higher income to be less supportive of progressive taxation.
Rehm (Reference Rehm2009, Reference Rehm2011) has made use of occupation-specific unemployment rates as a measure of economic risk. Unfortunately, occupation-specific unemployment rates are only available for a limited number of countries (∼20). This is unproblematic for Rehm’s studies as he mainly focusses on microvariables whilst controlling for multilevel structures via fixed effects. Yet, I cannot apply fixed effects models since I am primarily interested in the influence of macrovariables on attitudes towards progressive taxation (Allison Reference Allison2009; Möhring Reference Möhring2012). Thus, I use multilevel models with random effects. In these models, such a relatively low number of countries becomes problematic because type I errors are more likely (Stegmueller Reference Stegmueller2013). Therefore, occupational unemployment rates are less suitable for my analysis. In order to still control for individual risk, I use a dummy that takes the value “1” if a person is in part-time or even less than part-time employment (Rueda Reference Rueda2005; Stegmueller et al. Reference Stegmueller, Scheepers, Roβteutscher and de Jong2012). In addition, I include dummies that control for unemployment, being in education (student/school/vocational training), and retirement. Finally, a dummy for people who are not in the labour force equals “1” for those who help family members, housemen/housewives, permanently disabled, and for those who are not available on the labour market because of other reasons. For all of these dummies, the reference category is full-time employment.
To control for the effect of education on preferences towards redistribution, I include a variable that measures the highest educational degree ranging from 0=“no formal education” to 5=“university degree completed” and treat it as continuous (Barnes Reference Barnes2015; Beramendi and Rehm Reference Beramendi and Rehm2016; Häusermann et al. Reference Häusermann, Kurer and Schwander2016). Furthermore, I add a control variable for religiosity that measures the attendance of religious services, ranging from 1=“never” to 8=“several times a week”. Following studies that stress the importance of religiosity on redistributive preference (Scheve and Stasavage Reference Scheve and Stasavage2006; Stegmueller et al. Reference Stegmueller, Scheepers, Roβteutscher and de Jong2012), I expect more religious people to have a lower demand for tax progressivity. Finally, I control for age and gender (0=female; 1=male). In line with previous research on redistributive preferences (Gingrich and Ansell Reference Gingrich and Ansell2012; Schmidt-Catran Reference Schmidt-Catran2016), I expect older people to be more supportive of progressive taxation, whereas I expect men to be less in favour of tax progressivity.
On the macrolevel, I include several covariates. Since countries might already differ in economic growth prior to the crisis, I control for economic growth in 2007. To account for different levels of risks that have been taken by financial institutions in the run up to the crisis, I include a country’s average z-score from the years 2003–2008 in my analysis (Cihak et al. Reference Cihak, Demirguc-Kunt, Feyen and Levine2012). The firm-level z-score measures the financial stability of each institution. Higher values indicate a more stable financial system. It is calculated by dividing the sum of equity capital return as a percentage of assets by the standard deviation of returns. Then, the country-level averages of the firm-level z-scores are taken.Footnote 8 To take the influence of different levels of economic development into account, I control for the overall level of real GDP per capita (ln value) for the year 2009. Data come from the (World Bank 2017). In addition, I check my results for robustness by including several other macrovariables (see Table OA4 in the Online Appendix). First, different levels of inequality might influence tax policy preferences. Including inequality becomes particularly important because my income variable does not capture absolute differences in household income. Thus, I control for the market Gini coefficient (pretax and pretransfer) and the net Gini coefficient (posttax and posttransfer) as measurements of inequality (Solt Reference Solt2016). Second, tax progressivity might be more popular in countries that have a longer history of redistributive taxation. Therefore, I include the introduction year of the personal income tax (PIT) from the Tax Introduction Database (Seelkopf and Genschel Reference Seelkopf and Genschel2018). Third, since a more regressive tax system might have boosted compensatory arguments as well (Scheve and Stasavage Reference Scheve and Stasavage2016), I control for the share of total consumption tax revenues (% of GDP, year 2009) as a proxy for overall regressivity (Prasad and Deng Reference Prasad and Deng2009). Data come from Prichard (Reference Prichard2016). Finally, I control for welfare state effort by including social benefit expenditure as % of GDP for the year 2009 (IMF 2017).Footnote 9
I run several multilevel specifications with random intercepts to identify the determinants of preferences for tax progressivity. Since income as a predictor of preferences for redistribution may vary strongly between countries (Beramendi and Rehm Reference Beramendi and Rehm2016), all models include random slopes for household income. First, I calculate a minimum example that only includes real GDP growth rates. By doing so, I ensure that the effects of my main independent variable are not driven by my choice of covariates (Lenz and Sahn Reference Lenz and Sahn2017). Subsequently, I add the micro and macrovariables.Footnote 10 All individual level variables are unstandardised and unweighted.Footnote 11
Results
Main results
Table 1 presents the results of the multilevel analyses. In the minimum example (Model 1), GDP growth in 2009 has a negative and statistically significant impact on preferences for tax progressivity. Thus, respondents in countries with a lower GDP growth in 2009 have a higher demand for tax progressivity.Footnote 12 This finding is in line with Hypothesis 1 and holds when adding control variables on the microlevel (Model 2), taking average GDP growth in the first half of 2009 for those countries where the fieldwork took place earlier (Model 3), and adding further controls on the macrolevel (Model 4). A change in growth by 2 SD leads to a change in tax progressivity by 1/4 of its SD. As a comparison, this effect is nearly the same size as the effect of a change from the lowest to the highest household income group. This result is highly significant and robust to adding further control variables (Table OA4) and using multilevel generalised linear models (Table OA6).
Results multilevel models for tax progressivity
AIC=Akaike information criterion; DV=Dependent Variable.
Note: ***p<0.01, **p<0.05, *p<0.1.
Regarding the other control variables on the macrolevel, neither the coefficients of previous GDP growth in 2007, nor the ones of ln GDP per capita 2009 nor the z-score are statistically significant.
Let us now look at the microvariables.Footnote 13 All three dimensions of deservingness have positive and highly significant coefficients. Thus, demand for tax progressivity is higher if socio-economic outcomes are perceived as unfair. Since all three variables are scaled identically, we can compare their coefficients directly. The coefficients differ remarkably. The institutional dimension of deservingness has the largest effect of the three dimensions, followed by the behavioural and the family background dimension. The coefficient of the institutional dimension is nearly twice as large as that of the family background dimension. This finding indicates that the strength of different fairness dimensions varies substantially. In particular, if the political and economic system of a country is perceived as unfair, demand for correcting the economic outcomes via progressive taxation increases.
In line with other empirical studies, my results show that people with a higher income are less supportive of tax progressivity (Barnes Reference Barnes2015; Hennighausen and Heinemann Reference Hennighausen and Heinemann2015). The coefficient for economic risk – measured via part-time employment – is not statistically significant. This is in contrast to studies which look at the impact of risk on preferences for social policy. Although this might result from the operationalisation of economic risk (Rehm Reference Rehm2011), it hints at differences between social policy and taxation; in contrast to social policy, progressive taxation does not directly insure people against social risks. Therefore, the demand for social insurance via redistributive taxation (Varian Reference Varian1980) might be weaker than the demand for insurance via welfare state programmes. As expected, more religious people have a lower demand for tax progressivity (Scheve and Stasavage Reference Scheve and Stasavage2006). The same applies to people with a higher level of education. Interestingly, whereas older people have a higher demand for progressive taxation, retired persons actually want less tax progressivity. Cohort effects regarding experiences of mass warfare might be one factor that could explain why older people tend to be more supportive of tax progressivity (Obinger Reference Obinger2012; Scheve and Stasavage Reference Scheve and Stasavage2012), whereas the negative effect of retirement remains puzzling. Apart from the umbrella category of not being in the labour force, all other microvariables (unemployment and gender) are not statistically significant. Moreover, I have added dummies which measure political affiliation to the models (Table 10).Footnote 14 People with affiliations to leftist or centrist parties demand more tax progressivity than rightist voters. All other coefficients stay similar. As a comparison, a change in growth by 1 SD has the same effect on preferences for tax progressivity as being a centrist instead of a rightist voter.
Exogeneity of the crisis
The depth of the 2009 recession may not be entirely exogenous. In the following, I describe the factors that challenge the exogeneity assumption. Furthermore, I provide evidence that the effect of the financial crisis on tax policy preferences remains stable across model specifications which take exogeneity concerns into account.
First, domestic institutions and policies might mitigate the economic shock. In particular, automatic stabilisers such as social security programmes can lead to less severe economic downturns. In other words, economic shocks might be weaker in countries with bigger governments and more generous social policy programmes. To control for possible stabilisation effects, I include total government expenditure as a percentage of GDP into my model (Table 2, first column). Data come from the IMF (2017). The crisis effect remains robust.
Results multilevel models for tax progressivity in 2009 and 1999
AIC=Akaike information criterion; DV=Dependent Variable.
Note: ***p<0.01, **p<0.05, *p<0.1.
Second, the depth of the recession in 2009 could be influenced by previous economic development. Countries with a strong growth trajectory might have experienced a weaker downturn than countries that already had poor economic prospects prior to the crisis. To rule out that the crisis measure is determined by previous economic trajectories, I rerun my models by using the cumulative output gap instead of real GDP growth rates. To calculate the output gap, I estimate GDP per capita (pc) in 2009 with a Kalman smoothing procedure based on GDP pc time series from 2000 to 2008. The output gap is the difference between real and estimated values of GDP pc in 2009 as a percentage of GDP pc in 2008. Furthermore, I check the results by taking the output gap for 2010 if the ISSP’s fieldwork took place later. Columns 2 and 3 in Table 2 present the results. The findings are in line with Hypothesis 1: countries with a bigger output gap have a higher demand for tax progressivity.
Third, one might argue that countries with generally stronger preferences for progressive taxation have faced a stronger economic downturn. To rule out this possibility, I make use of the 1999 ISSP round and run a placebo test with the 1999 ISSP data and GDP growth rates from 2009 (Table 2, column 4). The results reveal that the economic downturn of 2009 was not stronger in countries where people already demanded more progressive taxation before the crisis. This finding supports my model’s exogeneity assumption. In addition, I compare the impact of GDP growth rates in 1999 on tax policy preferences to the results in 2009 for those countries that were surveyed in both rounds (Table 2, columns 5 and 6). In the wake of the crisis, the impact of GDP growth on preferences for tax progressivity is robust to using this reduced sample. In 1999, however, we cannot find an impact of GDP growth. Hence, economic development does not have an impact on tax policy preferences per se. Instead, the procedural dimension that is connected to the economic downturn – the financial crisis – is crucial in order to understand the effect in 2009.
Thus far, I have shown that crisis countries have had a higher demand for tax progressivity. However, I have not looked at changes in tax policy preferences. Due to the lack of yearly data, I focus on long-term development of tax policy preferences by looking at the changes from 1999 to 2009. I calculate each country’s weighted mean in tax progressivity preferences in both years and take the first difference. In total, this leaves me with 19 observations.Footnote 15 First, I run bivariate models to see whether the crisis in 2009 can explain differences in changes. Then, I expand this model by adding control variables. I include changes in age as a covariate since ageing societies might demand more tax progressivity. Furthermore, changing patterns of economic risk might have an effect on tax progressivity. Therefore, I include changes in unemployment and part-time work. Finally, I control for changes in religiosity to capture secularisation trends. The regression analyses (Table 3, Models 1–5) support my previous findings: across all models, real GDP growth in 2009 has a negative and statistically significant influence on the change in preferences for tax progressivity. Thus, a strong economic downturn has increased preferences for progressive taxation. In addition, I rerun the same model but replace GDP growth with a dummy variable that turns “1” when a country faced a strong economic downturn of more than 2% of GDP.Footnote 16 The results show that a strong economic downturn in the wake of the crisis has had a positive influence on support for progressive taxation (Table 3, Model 6). The effect of a major economic crisis on demand for tax progressivity is 0.2 points – again, as a comparison, this equals the effect of switching from the highest to the lowest income decile.
Determinants of change in preferences for tax progressivity 1999–2009
DV=Dependent Variable.
Note: ***p<0.01, **p<0.05, *p<0.1.
The impact of fairness perceptions in times of crisis
To find out whether the impact of the fairness dimensions on tax policy preferences is stronger in countries that faced a more severe economic downturn (Hypothesis 2), I use a cross-level interaction term between the 2009 growth rates and each of the three deservingness dimensions (Table OA5). Looking only at the interaction terms, we see that the interactions between growth and the behavioural deservingness dimension as well as between growth and the institutional deservingness dimension are negative and statistically significant. To interpret the cross-level interaction terms substantially, I calculate the marginal effects of each fairness dimension conditional on GDP growth in 2009 (Brambor et al. Reference Brambor, Clark and Golder2006). Hainmueller et al. (Reference Hainmueller, Mummolo and Xu2017) have shown that interaction effects are often interpreted in areas without common support in the data. I follow their suggestion and add histograms which show the distribution of GDP growth to the marginal effects plots. Figure 2 presents the results.Footnote 17 As expected, the coefficient for the family background dimension does not vary considerably; the impact is very similar between countries which faced a strong recession in 2009 and those which did not. For the other two dimensions, however, the coefficients differ strongly. The marginal effect for the behavioural deservingness dimension is more than twice as large in countries with a major economic downturn of 5% in 2009 compared to those with a positive growth rate of 1%. For the institutional deservingness dimension, the marginal effect increases slightly less, but still substantially by 50%. Thus, the impact of fairness considerations on tax policy preferences has intensified in countries that were hit harder by the crisis. These results are largely in line with Hypothesis 2.
Marginal effects of different fairness dimensions.
Conclusion
Can fairness arguments play a role for progressive taxation in the absence of mass warfare? By looking at the impact of the Great Recession on tax policy preferences, I have shown that different fairness dimensions are still important for shaping public preferences towards tax progressivity. The perception of rich people’s economic success as individually deserved and institutionally fair suffered as the crisis raised the salience of risky financial investments and fuelled public discussions about regulatory failure and bank bailouts. As a consequence, the impact of the behavioural and institutional fairness dimension on tax policy preferences intensified in countries with a strong economic downturn and demand for progressive taxation increased. The mechanisms during the Great Recession are somewhat similar to those during wartime (Scheve and Stasavage Reference Scheve and Stasavage2016): when society is doing badly and rich people are perceived as the ones to blame and/or profiteers of state actions, notions of undeservingness are triggered. Hence, people think it is only fair that the rich do worse as well. As a result, aggregate demand for a compression of income and wealth via progressive taxation increases. My analysis also considers that other factors can have an effect on attitudes towards progressive taxation. In fact, dominant theories about the influence of microlevel characteristics such as income and religion are supported by my results. Yet, these factors cannot fully explain why attitudes towards progressive taxation vary between countries in the wake of the crisis. Crisis-induced fairness arguments help understand this variation.
Placing my study in the discussion about progressive taxation in the last 30 years, I have shown that public opinion in the wake of the crisis pushes against the general time trend in tax policymaking. Whilst the taxation of top incomes has decreased massively since the late 1970s (Ganghof Reference Ganghof2006; Swank Reference Swank2016b; Kiser and Karceski Reference Kiser and Karceski2017), the crisis has raised political demand for progressive taxation again. By analysing preferences for progressive taxation in the wake of the crisis, this study has looked at the demand side – the very first stage of public policymaking. Yet, I have not examined actual tax policy outputs. Looking at the development of top statutory PIT rates reveals that the crisis was a game-changer indeed (Figure 3). Whilst tax rates for top incomes have decreased from 2000 to 2008, this development has reversed since the financial crisis. Thus, the data suggest that demand for progressive taxation was supplied politically. Moreover, the trend of increasing top PIT rates since the crisis has persisted. This indicates that a substantial change in tax policymaking has taken place. However, more work has to be done in order to identify the causal effect of the financial crisis on tax policies. Furthermore, we know relatively little about the responsiveness of governments towards tax policy demands. Finding out when and how politicians react to voters’ tax policy preferences is therefore a promising avenue for further research.
Top personal income tax (PIT) rate, 2000–2016.
Note: Average for 35 OECD countries. Data come from the OECD (2017).
Finally, it is noteworthy that the ISSP question about tax progressivity does not exclusively refer to highly progressive tax measures for the richest members of society. Instead, it captures a broader feeling about the idea of redistributive income taxation. While taxing income still marks the focal point of the redistributive tax state, it would be interesting to investigate whether the crisis has affected attitudes towards other taxes. For example, the idea of redistributing wealth via the taxation of inheritances has recently reentered the public and scholarly debate (Piketty Reference Piketty2014; Atkinson Reference Atkinson2015). Other highly progressive taxes such as recurrent taxes on wealth, land taxes, and capital gains taxes have also gained momentum (The Economist 2018). Finding out which role fairness perceptions have played in this development is crucial for our understanding of progressive taxation in the 21st century.
Supplementary materials
To view supplementary material for this article, please visit https://doi.org/10.1017/S0143814X18000430
Data Availability Statement
Replication materials are available in the Journal of Public Policy Dataverse at https://doi.org/10.7910/DVN/WKCZ9Z.
Acknowledgements
I wish to thank Jonas Driedger, Joseph Ganderson, Philipp Genschel, Ellen Halliday, Martin Höpner, Katy Morris, Kenneth Scheve, Laura Seelkopf, Martin Weinrich, Paula Zuluaga, workshop participants at the European University Institute and Ringberg Castle, as well as the Editor of the Journal of Public Policy and three anonymous reviewers for their valuable comments and suggestions.
