Taxing the 1 per cent: Public Opinion vs Public Policy

Abstract

Recent studies suggest that public policy in established democracies mainly caters to the interests of the rich and ignores the average citizen when their preferences diverge. I argue that high-income taxation has become a clear illustration of this pattern, and I test the proposition on a least likely case: Norway. I asked Norwegians to design their preferred tax rate structure and matched their answers with registry data on what people at different incomes actually pay in tax. I find that within the top 1 per cent, tax rates are far below (by as much as 23 percentage points) where citizens want them to be. A follow-up survey showed that this divergence is entirely driven by capital incomes being taxed too low. My results suggest that even in a reasonably egalitarian society like Norway, the rich get away with paying considerably less in tax than what people deem fair.

Research on the link between public opinion and public policy in formal democracies has established two important insights: (1) public policy often adheres to aggregate public opinion, but (2) when affluent citizens and the average citizen diverge in their preferences, government tends to follow the former and ignore the latter (Bartels 2016; Gilens 2012; Gilens and Page 2014; Jacobs and Page 2005; Mathisen et al. forthcoming; Elsässer, Hense, and Schäfer 2017; Schakel 2021). For the United States, the most studied case to date, these conclusions were not shocking given its unusually high economic inequality and the role of money in election campaigns (Ferguson, Jorgensen, and Chen 2016); Piketty 2014). What is more surprising is that similar studies have found much of the same results in other – presumably more egalitarian – western countries, such as Germany (Elsässer, Hense, and Schäfer 2017) and the Netherlands (Schakel 2021). Several cross-national studies have also concluded that public spending in the OECD overall is tilted towards the preferences of the rich (Bartels 2017; Giger, Rosset, and Bernauer 2012; Peters and Ensink 2015).

While these insights are highly valuable, knowledge of which policies are most affected by unequal political influence is still lacking; that is: if the rich hold disproportionate influence over the policy-making process, which policies deviate the most from what ordinary citizens want? Studies of political representation are usually not able to answer this question, either because they study policy changes, thus capturing responsiveness at the margins, and not overall bias (Elsässer, Hense, and Schäfer 2017; Gilens 2012; Schakel 2021), or simply because opinions and policies are measured on incomparable scales (see Simonovits, Guess, and Nagler 2019 for more on these limitations).

In this paper, the case of the income tax system is explored, with a particular focus on high-income taxation. Income taxation is a pivotal aspect of public policy as it directly affects the distribution of material goods in society. As one of the main sources of public revenue, income tax affects the government's financial wiggle room in other areas, such as social policy. Hence, it matters how states design their income tax systems and, importantly, whose preferences triumph. From the perspective of the rich, high-income taxation constitutes a potential threat, which will usually be countered with various income defence strategies. In most societies, the rich will use their vast resources to try to influence tax policy in such a way as to keep as much of their income flow as possible (Winters 2011).

The question is whether the institutions of democracy, in the face of such special interests, still manage to ensure that public policy adheres to public opinion. To shed light on this question, the case of Norway is studied. Norway is usually considered to be a well-functioning, egalitarian democracy with strong unions giving voice to the working class (Allern, Aylott, and Christiansen 2007; EIU 2020). Therefore, Norway provides a useful least-likely case because if high-income taxation is out of tune with the preferences of ordinary citizens here, one would hardly expect the situation to be much better in other affluent countries – where political influence is more unequally distributed.

To begin with (Study 1), the long-run evolution of Norway's top marginal income tax rate since the 1960s is compared with opinion surveys implemented at key moments in the timeline – just before major changes occurred. This analysis already reveals that decisions taken to substantially reduce the top marginal rate had limited, and usually very low, popular support.

Zooming in on the current tax system (Study 2), a survey was fielded where Norwegians were asked to freely design their preferred income tax structure by setting rates for a group of imaginary taxpayers with annual incomes ranging from low (c. $11,000) to extremely high (c. $11 million). Their answers were matched with registry data on what the same kinds of taxpayers actually pay in income tax. The results show that while actual tax rates in the bottom 99 per cent were quite close to where citizens wanted them to be, the tax rates in the top 1 per cent were far off. On average, citizens wanted the one-per centres to pay around 45 per cent in income tax, considerably higher than their actual tax burden, which is around 30 per cent (according to the conservative estimates used in this article). Moreover, and unsurprisingly, citizens prefer a progressive tax structure. In contrast, actual effective tax rates are regressive towards the top of the income distribution. This means that the opinion-policy divergence increases within the 1 per cent, reaching a striking 23 percentage points for the highest income evaluated by respondents.

Finally, a follow-up survey (Study 3) shows that the disconnect between actual and preferred tax rates at the top of the income distribution is entirely driven by preferential taxation of capital incomes. In most countries – including Norway – people at the top of the distribution predominantly receive their incomes from capital investments. Such incomes are usually subject to flat, preferential tax rates, whereas labour is taxed at higher and progressive rates. When asked to set tax rates for labour incomes and capital incomes, the average citizen's preference for labour income taxation was largely in line with how labour incomes are taxed. However, the preference for capital income taxation was much more progressive and higher for top incomes compared to how capital incomes were taxed.

This paper contributes to the literature on unequal responsiveness in several ways. First, by studying the particular case of high-income taxation, the paper demonstrates that political inequality is not just a general statistical pattern, as many studies have now shown. It has real-world consequences. To illustrate, the discrepancies between actual tax rates for the top 1 per cent and those preferred by citizens have substantial implications for government revenues. If the top 1 per cent were subject to the tax rates preferred by public opinion, it would generate roughly 7 per cent in extra government revenues compared to the current situation (Appendix Section 1.6). Second, the findings contradict the idea that unequal responsiveness is simply an artefact of differences in educational levels or policy information (Elkjær 2020). Third, the paper demonstrates the importance of studying responsiveness and congruence together. As the paper shows for high-income taxation, unresponsiveness over a long period increases the cumulative incongruence between public opinion and public policy. Political inequality can only be better understood by recognizing the interlinked nature of these phenomena.

High-Income Taxation in the 21st Century: Out of Tune with Public Opinion?

In a 2007 interview, multi-billionaire Warren Buffet said that he paid less in income tax, as a fraction of his income, than his secretary.¹ Perhaps contrary to some people's expectations, this was not due to tax avoidance or elaborate tax planning. Instead, it resulted from how the US tax system is designed. Buffet receives most of his income from dividends and long-term capital gains, which are taxed considerably lower than labour income.

The Buffet story is no anomaly. In their book, the Triumph of Injustice (2019), Emanuel Saez and Gabriel Zucman used an impressive dataset based on tax returns, survey data, and national accounts to show that the structure of the US tax system constitutes, in their words, ‘a giant flat tax that becomes regressive at the top’. Studies using similar kinds of data have found much the same structure in France (Bozio et al. 2018) and Europe in general (Blanchet, Chancel, and Gethin 2022). In Capital in the Twenty First-Century, Thomas Piketty writes that because of increasing cross-national tax competition, ‘in most countries taxes have (or will soon) become regressive at the top of the income hierarchy’ (Piketty 2014, 634).

Observers often point to two reasons for rich countries’ decreasingly progressive income tax systems. The first is the gradual reduction of the top marginal income tax rate (Piketty 2020, 445–56). In the 1960s, most rich countries had top marginal tax rates above 60 per cent (in the US and UK, it was above 80 per cent). These rates have been gradually cut to levels closer to 40 per cent, making statutory rates less progressive (Piketty, Saez, and Stantcheva 2014).²

The second reason is the preferential taxation of capital income. The truly wealthy receive most of their incomes not from work but from returns on capital. Saez and Zucman (2019) describe this as ‘a constant of capitalist societies: as one moves up the income ladder, the capital share of income rises until it reaches 100 per cent at the tip-top’. The implication is that ‘[w]hen governments reduce the tax burden on capital, they almost always reduce taxes for the wealthy’ (Saez and Zucman 2019, 97). Moreover, the tax burden on capital has been reduced substantively in recent years. For example, Genser and Reutter (2007) note that dual income taxation – where labour income is taxed progressively, and capital incomes are taxed at flat, preferential rates – ‘has become an important blueprint for income tax reforms in Europe’. The duality can easily be observed for the OECD as a whole. While the average top marginal labour income tax rate in 2020 is 44.5 per cent, the average top tax rate on long-term capital gains is 19.1 per cent.³ Preferential treatment is also common for other types of capital income, such as dividends and interest on bank deposits (Harding and Marten 2018).

The Median Voter Model

Is the structure of the modern income tax system, particularly with respect to the tax level for the very wealthy, in line with what the public wants? Although flat taxes that become regressive at the top might seem unattractive, it cannot be ruled out that this outcome results from policy makers following the will of the public. If not directly (as in the public pushing for a regressive system), then perhaps as a result of several tax policy changes supported by the public that makes the system flat or regressive – such as the cutting of top marginal rates or preferential treatment of capital income. There are at least three potential reasons for this.

First, from the perspective of the classic median voter theorem (Black 1948; Downs 1957) – which posits that parties will gravitate towards the policy preferences of the median voter – the public should be fully able to shape the tax system according to its preferences. The Meltzer-Richard model, for example (Meltzer and Richard 1981; Romer 1975), which builds on the median voter theorem, implies that the progressiveness of the income tax will be decided by the voter with median productivity (income). Although the multidimensionality of politics means that this model might not work as smoothly in practice as in theory, if citizens care strongly about taxes (typically the case), a fairly close relationship between policy and average public opinion should be seen.

Second, we know that citizens sometimes favour tax policy choices that mostly benefit the very wealthy and make the overall tax system less progressive. Some examples of this from the US include public support for some of the Reagan tax cuts in the 1980s and the Bush tax cuts in the early 2000s (Gilens 2012, Chapters 6 and 7). In a very different case in Sweden, inheritance tax was abolished in 2004, which had substantial popular support (Henrekson and Waldenström 2016). In a later survey, only around 41 per cent favoured reinstating it with a high exemption threshold (Bastani and Waldenström 2021). Furthermore, in a cross-national survey of five western countries, Alesina, Stantcheva, and Teso (2018) found that estate tax was not supported by more than a third of respondents in any of these countries (even though it only affects a small fraction of the population). Many studies have suggested that such preferences could partially stem from misinformation (Bartels 2005; Kuziemko et al. 2015; Slemrod 2006) or policy options that are too narrowly defined (Hacker and Pierson 2005). Nevertheless, whether by misinformation or manipulation, these were still the expressed preferences of the public within the specific context in which these questions were debated.

Third, periods in which the tax system was made less progressive have sometimes coincided with clear shifts in the electorate's voting behaviour. Many of the changes in high-income taxation in North America and Europe happened during the 1980s when right-wing parties were bolstering impressive election results. At the same time, social democrats had started to decline – and still are (Benedetto, Hix, and Mastrorocco 2020). Furthermore, tax policy was not a hidden issue but rather some of the main talking points for politicians such as Ronald Reagan and Margaret Thatcher. Even seemingly controversial policies, such as cutting taxes for the rich, were openly embraced by right-wing parties under the justification that this would produce economic growth, which would ‘trickle down’ on the less well off. Regardless of the validity of such claims, it would not be surprising if substantial numbers of these parties' voters accepted them as fact and supported tax changes that made the system less progressive. Hence, such changes might have reflected the wishes of right-wing voters, if not the average citizen.

The Disproportionate Influence of the Wealthy

However, the median-voter perspective has some important limitations. As Alesina and Giuliano (2011) rightly point out, ‘the main failure of this model relies on the simplistic assumption about the policy equilibrium, namely the one person, one vote rule and the median voter result’. There is a diversity of political science research showing that even in formally democratic societies, wealthy citizens can achieve much more influence than what their ‘one vote’ would entail. For example, they might help their favourite political candidates win by contributing to their election campaigns (Ferguson 1995; Ferguson, Jorgensen, and Chen 2019); finance lobby groups that voice their views (Coen, Katsaitis, and Vannoni 2021); or stop investing in the economy, pressuring governments to adopt policies that ensured continued investment (Przeworski and Wallerstein 1988; Young, Banerjee, and Schwartz 2018). Moreover, an increasing number of studies show that proposed policy changes are much more likely to be adopted if they are favoured by the affluent than if they are favoured by the middle class or the poor (Elsässer, Hense, and Schäfer 2017; Gilens 2012; Gilens and Page 2014; Schakel 2021). Hence, public policies might end up at very different places than they would have if the median voter decided.

One would expect a high-income tax policy to be particularly vulnerable to such distortions. It is uniquely relevant for the wealthy as a policy area because their material interests are in conflict with those of the general public. That is, issues of economic incentives aside, taxes on the highest income earners entail costs exclusively for them and benefits to the rest of the public through increased tax revenue. Previous research has emphasized that it is precisely on these types of issues, where preference gaps are large, that unequal responsiveness will have the clearest effects on public policy (Gilens 2009).

Furthermore, public opinion research shows that attitudes towards taxation tend to be heavily influenced by self-interest, at least when personal costs and benefits are clear (Chong, Citrin, and Conley 2001; Franko, Tolbert, and Witko 2013). This would suggest that if policy influence is mainly concentrated at the top of the income/wealth distribution (which is what some research has found (see, for example, Gilens 2012, 82), then that influence would be used to reduce taxes specifically at the top, possibly resulting in an observed tax rate regression for the top earners. Income taxes, particularly those affecting capital, are also prone to international competition in a world of free-flowing capital. In fact, in a review of the literature, Fuest, Huber, and Mintz (2005) note that ‘[m]uch economic analysis has viewed that capital taxes will disappear if real capital is perfectly mobile at the international level’. While this might be an exaggeration, there is good evidence that there has at least been a partial ‘tax race to the bottom’ (for example, Clausing, Saez, and Zucman 2021; Keen and Konrad 2013; Lierse 2021). Hence, governments might experience pressure to cut taxes at the top, not only from the wealthy in their own country but from multinational companies and investors who are relatively free to move capital to the places where they are taxed the least.

A Least Likely Case: Norway

The case of Norway is examined. Norway arguably constitutes a least likely case to find substantial discrepancies between public opinion and public policy on economic issues. This makes it a useful case because if tax policy deviates much from public opinion here, it seems unlikely that the situation would be much better in other cases. There are several reasons for this. First, Norway is commonly viewed as one of the most well-functioning democracies in the world. For example, in the Democracy Index by the Economist Intelligence Unit, Norway has been ranked the most democratic country in the world every year since 2010 (EIU 2020). Norway also has long had comparatively strong unions that serve as a counterweight to business influence in the policymaking process. In addition, there are limited possibilities for using money to influence election results since political advertisements on television are banned.

Possibly a testimony to a combination of these institutions, a recent study of Norway comparing public opinion and public policy on hundreds of concrete policy proposals over fifty years found that governments appear to have responded to the preferences of both high- and low-income citizens on issues of economic policy (Mathisen 2022). Furthermore, income did not appear to be as strong a predictor of political influence as education. Nevertheless, because of the particular relevance that high-income taxation holds for the wealthy, it might be that they are still able the muster the influence that they do have to shape policy in this area, even in a relatively egalitarian context like Norway.

An aspect that could potentially undermine the argument of Norway as a least likely case in terms of opinion-policy disconnect on tax policy is if Norwegian citizens were unusually supportive of redistribution. However, this does not seem to be the case. Survey evidence has shown that Norwegian public opinion on tax progressivity is in the ‘middle of the pack’ across seventeen advanced democracies (Barnes 2015, 16).

Study 1: The Evolution of the Top Marginal Income Tax Rate

The empirical analysis starts by following in the footsteps of previous studies of policy responsiveness. Hence, the extent to which changes in the top marginal income tax rate have reflected public opinion over time is examined.

Methods

For such an analysis, one needs survey data collected at strategic points, preferably right before major changes occur. Most of the top marginal income tax rate reductions in OECD countries – including Norway – occurred between 1975 and 2000. While survey questions asking specifically about the top marginal rate are surprisingly hard to come by for this period (for any country), the Norwegian National Election Surveys (NNES), luckily, have asked respondents since 1965 about the general tax level on ‘high incomes’. Sample sizes vary from 1,600 to 2,200 between the surveys. The following question was asked, with minor linguistic variations, in 1977, 1981, 1985, 1989, 1993, and 1997: ‘Using one of the answers on this card, how do you view…Lowering taxes on high incomes?’ From 1977 to 1989, the respondents answered on a bad proposal-good proposal 5-point scale, while in 1993 and 1997, they answered on a disagree-agree 5-point scale. For the analysis below, support is calculated as the percentage of respondents who chose one of the options above the neutral midpoint of the scales – don't knows excluded. The question in 1965 is different⁴ but still asks about the tax level on high incomes, making it roughly comparable to the question in later waves.

Results

Figure 1 shows the evolution of Norway's top marginal income tax rate. In addition, it shows the estimated share of Norwegians who support cutting taxes on high incomes in each of the relevant National Election Studies. These are marked with vertical lines. A couple of things are noteworthy. First, there was never a majority support for cutting taxes on high incomes. Not even in 1977, at the start of the supposed ‘turn to the right’, was this the case. Support was just below 40 per cent both in 1965 and in 1977, which was the highest it ever got. From 1981 onward, no more than a third of citizens supported further cuts. Second, despite this fact, the top marginal income tax rate was slashed from around 80 per cent to 40 per cent over the period. During the 1980s alone, it was cut by 20 percentage points; with the tax reform in 1992, it was reduced by a further 10 points. These vast changes in high-income taxation were not what the average citizen wanted. The question then becomes, were these policies more in line with the preferences of certain public sectors?

Fig. 1. The top marginal income tax rate and popular support for cutting taxes on high incomes over time.

Note: Tax rates are from Piketty, Saez, and Stantcheva (2014). Public opinion estimates are from the Norwegian National Election Studies (1965–1997) .

Figure 2 shows over-time estimates of support for cutting high income taxes among six different social groups (see Appendix Section 1.1). As expected, high-income citizens are keener on cutting income taxes than low-income citizens. However, among the affluent, there was only majority support for cutting the rate in the 60s and 70s. During some of the major cuts in the 80s and 90s, only about a third supported it.

Fig. 2. Support among different sectors of the public for cutting taxes on high incomes.

Note: ‘High’ refers to the top 10 per cent on a variable's distribution, while ‘low’ refers to the bottom 10 per cent. Left-wing: voted for the Labor Party, Socialist Left, the Norwegian Communist Party, or the Red Electoral Alliance. Right-wing: voted for the Conservative Party or the Progress Party. Data are from the Norwegian National Election Studies (1965–1997).

Another possibility is that these changes responded to demands from right-wing voters. However, this explanation only seems plausible for the early 80's initial tax cuts. Looking at support by political orientation (Fig. 2), left-wingers have been quite stable in their opposition towards tax cuts for high incomes, and there has been a radical shift among right-wingers. In 1965 and 1977, strong majorities on the right (>70 per cent) favoured tax cuts for the rich. However, in 1985, the majority within this group shifted, and support was down to around 40 per cent. In only eight years (from 1977), right-wing support fell by a whopping 30 percentage points. These results suggest that while right-wingers likely supported the initial high-income tax cuts of 1981 (bringing the top marginal rate down to about 60 per cent), even they did not want the subsequent cuts of the late 80s and early 90s.

Study 2: Preferred Tax Rates vs Actual Tax Rates

Having established in Study 1 that the reductions in the top marginal income tax rate over the past half-century had limited, and usually very low, popular support, Study 2 addresses the natural follow-up question: To what extent is the current tax system in Norway congruent with the preferences of Norwegian citizens when it comes to the tax burden on high earners?

Methods

The analysis requires two types of data: a measure of citizens' preferred tax rates for a set of incomes (including the highest earners) and a measure of the effective tax rates paid at those incomes.

An original survey in the Norwegian Citizen Panel (NCP) Round 19 in November 2020 (Ivarsflaten et al. 2021) was designed to measure citizens' preferred tax rates. The NCP is a nationally representative online panel based on randomized postal recruitment from the national registry. The survey was implemented on a sub-sample of 1,990 respondents. The following vignette was presented to the respondents:

Below, we have listed several imaginary people with different annual incomes. A person's income can come from work, investments, or other sources. For each of the imaginary people, please enter what you think the tax rate on their income should be; that is, how much of their entire income you think they should pay in income tax. Zero per cent would mean you think the person should not pay income tax.

Respondents were then presented with a list of ten people identified as Person A, Person B, etc., with annual incomes ranging from 100,000 NOK (c. 11,000 USD) up to 100 million NOK (c. 11 million USD). Unlike previous tax preference studies, which usually have not asked about incomes higher than $200,000 (Zelenak 2008, 372), incomes up to 0.01 per cent of the income distribution were included to capture tax preference on truly high incomes. Next to each person was a blank space where the respondent could write any tax rate from 0 to 100 (with decimals if they preferred).

Respondents were randomly assigned to either view the list from the lowest earner to the highest or from the highest to the lowest. A comparison of the average tax rates set by the two groups shows that the order effect was quite limited (see Appendix Fig. A4). Additionally, respondents were asked what they believe to be the current effective tax rate for a randomly chosen income (see Appendix Section 1.3 for details). This was done mainly so that tax knowledge could be controlled for when estimating whether actual tax rates are closer to the preferences of certain groups of citizens.

Data from Statistics Norway, the national statistical body in Norway and the main producer of official statistics, was ordered to measure what people pay in taxes.⁵ They maintain registry data based on tax returns for all Norwegian citizens. They provided data from 2018 on taxes paid around the same ten income levels as those evaluated by respondents in the NCP survey. In order to get robust average tax rates, intervals around the target incomes to obtain a sufficient number of observations were constructed. Naturally, the intervals must be fairly large for the highest incomes since few citizens are at those levels. In the data, income means all taxable income, including wages, salaries, and realized capital incomes (interests, dividends, etc.). See Appendix Section 1.2 for details on how the average tax rates were calculated.

Table 1 presents effective tax rates for approximately the same income amounts that respondents in the 2020 survey were asked about. Perhaps the most striking feature of the table is that the effective income tax in 2018 was only progressive up to the $220,000 amount. From there, it was regressive. Taxpayers near the two highest incomes (5.5 and 11 million USD) paid less in tax, as a fraction of their income than people earning around $110,000. Importantly, this regression only applies to incomes at the top of the distribution (within the 1 per cent). Even though the tax rates at the top intuitively seem rather low compared to the rest of the income distribution, if anything, they are probably overestimated since they do not consider unrealized capital gains. They also do not consider tax evasion, which is much more common among top-income earners (Alstadsæter, Johannesen, and Zucman 2019). For these reasons, the discrepancies documented below between preferred and actual income tax rates for high incomes should be viewed as conservative estimates. The real discrepancies are bound to be larger.

Table 1. Effective tax rates for different incomes in 2018

Note: Registry data for Norwegian taxpayers for the year 2018. Statistics Norway compiled the data in January 2021. Income tax is calculated as the sum of all types of income tax paid to municipality, county, and state, as well as National Insurance Scheme members' contributions (details in Appendix). Locations in the income distribution are from the World Inequality Database (https://wid.world/).

Results

Table 2 shows the respondents' average preferred tax rate at different income levels in the Norwegian Citizen Panel. On average, respondents clearly want a progressive tax structure. Average preferred tax rates vary from 5.9 per cent for someone earning $11,000 per year up to 48 per cent for someone earning $11,000,000 per year. Interestingly, respondents even want progressivity within the top 1 per cent of the income distribution (see the 7.2 percentage point difference between the first and last income within the 1 per cent in Table 2). However, as shown in Appendix Fig. A2 (which plots the full distribution of responses to each income amount), respondents agree more about normal incomes than very high ones. The tax rate standard deviation almost doubles – from eight points to sixteen points – from the lowest to the highest amount (Appendix Fig. A3).

Table 2. Actual and preferred income tax rates by size of income

Note: Survey results are from NCP Round 19 (2020). Actual tax rates are based on registry data for Norwegian taxpayers provided by Statistics Norway. Standard errors in parentheses.

How do average preferences square with effective tax rates paid by different income groups? Quite well for incomes within the normal range. As Table 2 shows, for the first six incomes listed ($11,000 up to $220,000), average preferred rates deviate, at the most, 3.2 points from the actual rates. This income range represents approximately 99 per cent of the income distribution (the bottom 99 per cent). However, as one moves up from here, the deviation increases rapidly. On the highest income ($11,000,000), taxpayers paid an income tax rate 23 percentage points lower than the average preferred rate. This extraordinary discrepancy between public opinion and public policy can be explained by the popular preference for progressivity beyond moderately high incomes and the reality of tax rate regression at the top of the income distribution. The comparison of preferred and actual tax rates is plotted in Fig. 3, which shows a ‘crocodile shape’; that is, a close match followed by rapidly increasing divergence at the top of the distribution. The Appendix contains an extended version of this graph (Fig. A11) that plots what respondents believed were the actual rates when asked to guess. This shows that citizens are largely aware of the discrepancy at the top but underestimate its full extent.

Fig. 3. Preferred vs actual tax rates.

Note: Thin lines indicate 95 per cent confidence intervals (too small to be visible for some of the triangle line).

Does the tax system fit better with the preferences of certain groups of respondents? Figure 4 plots the tax preferences of six social groups based on income, education, and political orientation. The figure shows that actual tax rates lie closest to the preference of high-income and right-wing respondents because these groups favour a somewhat less progressive structure than the others. The highest educated respondents, with the most progressive tax rate preferences, are furthest apart from the actual rates. In order to statistically test these differences, the following OLS model for each income amount, on which respondents were asked to provide a tax rate, was estimated:

$$dev_i = \alpha + \beta _xX_i + \beta _cC_i + \varepsilon $$

Fig. 4. Preferred vs actual tax rates for six groups of respondents.

Note: High refers to the top 10 per cent on a variable's distribution, while low refers to the bottom 10 per cent. Left-wing: voted for Labour, Socialist Left, or the Communist Party in the last election. Right-wing: voted for the Conservative Party or the Progress Party.

Where dev _i is the absolute distance between the actual tax rate and the preferred tax rate for respondent i; X is the group variable of interest; and C is a set of control variables. Control variables include (in addition to the variables already mentioned) occupation, age, gender, and region. The model also controls for respondent tax rate knowledge, measured as the absolute distance between their guess for the actual tax rate on a randomly assigned income amount (out of the ten used to measure preferences) and the actual tax rate for that income. The latter is important since citizens partially form their opinions based on current policies (Popkin 1991; Zaller 1992). Therefore, a group of respondents might appear to have preferences closer to the current system merely because they are more aware of how the system looks. This, however, is no longer a viable explanation for the results when tax rate knowledge is controlled for.

The model results (shown in Figs A5–A7 in the Appendix) show that, after controls, actual tax rates are still closer to the preferences of high-income and right-wing respondents and farther apart for high-education respondents for the three highest income amounts. The effects are statistically significant, and effect sizes vary from 2 to 8 percentage points. These findings suggest that the income gradient for opinion-policy congruence on tax policy cannot be explained away by either education or tax-specific knowledge. This goes against the recent argument by Elkjær (2020) that informational asymmetries drive unequal responsiveness.

Despite these differences, however, Fig. 4 suggests widespread agreement across different social groups on the basic structure of the tax system. On average, all six groups agree that tax rates should be no higher than 10 per cent for the lowest income amount and at least 40 per cent for the highest. In total, 86 per cent of respondents set rates that agreed with the former and 76 per cent with the latter. Even if tax preferences by party ID (Appendix Fig. A8) were broken down, the voters of all nine parties (including the Communist Party and the far-right Progress Party), on average, prefer a tax rate structure that goes from somewhere below 10 per cent to somewhere above 40 per cent.

Nevertheless, this is not how the tax system looks. For incomes in the bottom 99 per cent, the system is never far off from any group's preference. On the other hand, taxes in the 1 per cent, while comparatively closer to the preferences of the right-wingers and the respondents with (relatively) high income, would prefer a considerably heavier tax burden on the truly rich.

Study 3: Preferential Taxation of Incomes from Capital

Study 2 showed that at the top of the income distribution, there is a substantial discrepancy between how much earners pay in tax and what public opinion wants them to pay. However, by design, Study 2 asked respondents about income tax without specifying its source. In most countries, including Norway, incomes from capital have increasingly been taxed at preferential rates compared to incomes from labour – a major reason for the tax rate regression at the top when looking at all income combined (Genser and Reutter 2007; Saez and Zucman 2019, 19). In the Norwegian system (as of October 2021), labour income is taxed progressively from 0 to 46.4 per cent (53 per cent if one includes employer's national insurance contributions). Capital income is taxed at a flat 22 per cent rate. Certain capital incomes – capital gains and stock dividends – are taxed at a somewhat higher effective rate of 31.68 per cent.⁶ The question that Study 3 addresses is how this system compares to citizens' preferred tax rates on capital and labour incomes, respectively.

Methods

A follow-up survey of Norwegian citizens through the survey company YouGov in October 2021 was made. The respondents were presented with a similar design to Study 2, where they set tax rates for a list of ten income earners. However, the new survey asked all respondents to first do this exercise for ten earners who exclusively received their income from capital investments (that is, capital gains, dividends, interests, etc.) and then do it again for ten earners who exclusively received their income from labour wages (the order of the labour and capital tasks was randomized (see Appendix Section 1.4 for details). The preferred tax rates were then compared to effective tax rates for labour and capital incomes (see Appendix Section 1.5).

Results

Figure 5 plots actual tax rates on capital and labour income and the rates that citizens prefer, on average, when asked about the two types of income, respectively. Comparing the two panels in Fig. 5, citizens want both labour and capital incomes to be taxed progressively from approximately 10 to 40–45 per cent, but with a slight advantage for capital (about 5 points lower than labour income for most of the amounts). The left-hand panel shows that when it comes to labour income, this is not very far from actual tax rates. If anything, actual tax rates on some of the highest labour incomes are too high for the average citizen. In contrast, the right-hand panel shows that the highest capital incomes are taxed considerably lower than what the citizens want. This is particularly the case for the capital types subject to the 22 per cent rate. The same plot looks quite similar if we zoom in on right-wing voters or high-income citizens (Figs A9 and A10). These results suggest that the large discrepancy between actual (paid) and preferred tax rates for the 1 per cent (see Fig. 3) is entirely explained by capital incomes being taxed lower than what the citizens want – top labour incomes are not taxed too low.

Fig. 5. Preferred and actual tax rates for labour and capital income. Thin lines indicate 95 per cent confidence intervals. Data is from the YouGov survey of Norwegian citizens.

Conclusion

This article provides evidence that even in a fairly egalitarian country like Norway, public policy on an issue of high relevance to the wealthy can become seriously detached from the preferences of ordinary citizens. Had the government followed average public opinion about the tax burden on the highest income earners, the top marginal income tax rate would likely not have been cut the way it was over the past decades, and capital incomes would not receive the preferential treatment they currently enjoy in the tax system. Perhaps, most importantly, incomes at the top would be taxed significantly higher than they are today.

The discrepancies documented above between actual tax rates at the top of the income distribution and those preferred by citizens are striking. Nonetheless, they are likely underestimated. This is because the official tax data employed here ignore an important part of top incomes, namely unrealized capital gains (that is, capital gains not paid out to investors but withheld in companies). Aaberge, Modalsli, and Vestad (2020) find that effective tax rates at the top almost reach single digits when those incomes are included on the income side of the tax equation.⁷

Suppose high-income taxation is so out of tune with public opinion (including the preferences of right-wingers and the modestly affluent); why did policy in this area develop as it did? A plausible answer to that question must take into account who benefits from the policies that are seemingly so unpopular with most of the public. Based on the analyses presented here, they appear to be earners in the top 1 per cent and upwards, often paying lower effective tax rates than the middle class. It is also of high relevance to consider the numerous studies suggesting that influence on government policy is concentrated at the top of the income hierarchy (Bartels 2016; Elsässer, Hense, and Schäfer 2017; Gilens 2012; Gilens and Page 2014; Jacobs and Page 2005; Rigby and Wright 2013; Schakel 2021). Combining these two pieces of information, a parsimonious explanation would be that high-income taxation is out of tune with public opinion because rich citizens use their disproportionate political influence to change the tax system in ways that minimize their tax burden. This explanation aligns with Jeffrey Winters' argument in his book Oligarchy (Winters 2011). Winters' argument would additionally account for the finding that high-income taxation seems at odds with the preferences of even relatively affluent survey respondents. Winters argues that ‘oligarchs are not just offloading tax burdens to those below them in society, but the mass affluent’ who ‘lack the material power resources’ for ‘income defence’ (Winters 2011, 213, 245).

Another explanation is international tax competition. A large literature suggests that capital mobility in the 21st century means that states have to compete against each other to attract investment. One of the ways that they do this is by reducing taxes on capital, possibly contributing to a ‘race to the bottom’ (Bretschger and Hettich 2002; Lierse 2022; Rodrik 2011). However, it would be artificial to treat this explanation as independent from the explanation focusing on the power of the rich. A main mechanism that the rich use to influence policy is precisely to move (or halt) capital investments. Scholars such as Block (1977) and others (for example, Lindblom 1982; Przeworski and Wallerstein 1988) have argued that capital can forcefully shape public policy through the mere knowledge on the part of the legislator that a failure to maintain a business-friendly political environment (for example low taxes) can lead to disinvestment or capital flight – commonly referred to as the structural power of capital. This mechanism becomes all the more relevant in a world of unrestricted free-floating capital (see Winters 1996).

Arguably, in a country like Norway, where direct political financing of candidates plays a much more limited role than in the US, the structural power of domestic and foreign capital owners might be a more plausible mechanism for the power of the rich. Thus, capital mobility and financial globalization might be seen as factors facilitating the power of capital owners to influence tax policy rather than as a wholly different explanation for tax policy. However, curtailing this source of influence will likely be much more difficult for a country like Norway to do unilaterally, as opposed to regulating money in its election campaigns. Limiting the structural power of mobile capital will likely require cross-national cooperation and coordination (Clausing, Saez, and Zucman 2021).

In sum, it is unclear what is the precise mechanism by which the Norwegian income tax system has become skewed in favour of the interests of the rich and detached from public opinion. The answer to that question probably has important inferential implications. The consequences of tax competition might be viewed to a lesser extent than money in politics as an indictment against the Norwegian political system and more as a predictable result of being a small open economy in a world of free-floating capital. Still, this paper's main contribution shows that mass preferences and public policy are far apart on high-income taxation – a policy area of first-order economic importance. This contradicts the median voter theorem and seemingly goes against several studies arguing that policy and public opinion are usually congruent (for example, Alexander Branham, Soroka, and Wlezien 2017; Enns 2015; Lax, Phillips, and Zelizer 2019). The explanation for the different results might be that while these studies usually look at salient issues on the public agenda, the current paper documents the largest incongruences when it comes to taxation of capital incomes – an issue that is rarely given much attention in public discourse.

It might be unrealistic (perhaps even undesirable) to expect that tax policy should precisely mirror public opinion in every aspect. Tax policy can be complicated and might contain many considerations about which citizens have few opinions (Converse 1964). At the same time, it seems uncontroversial to say that if the overall distribution of the tax burden differs dramatically from what citizens believe is fair, then this is a serious problem for any political system with democracy as its ideal. Not only because tax policy is an important policy area but because it is one of the major tools that governments in the 21st century have at their disposal to counteract rising economic and political inequality.