Understanding Tax Policy: How Do People Reason? (2021) PDF

UNDERSTANDING TAX POLICY: HOW DO PEOPLE REASON?∗ STEFANIE STANTCHEVA Most important, I study their understanding of the mechanisms of tax policy and the reasoning that underlies their policy views. In decomposing policy views, I find that support for income and estate taxes is most strongly correlat...

UNDERSTANDING TAX POLICY: HOW DO PEOPLE REASON?∗ STEFANIE STANTCHEVA Most important, I study their understanding of the mechanisms of tax policy and the reasoning that underlies their policy views. In decomposing policy views, I find that support for income and estate taxes is most strongly correlated with social preferences, that is, the perceived benefits of redistribution and concerns around the fairness of inequality and taxation, as well as with broader views of the government. Efficiency concerns play a more minor role. These correlational patterns are confirmed by the experimental approach, which shows people instructional videos that explain the workings and consequences of one of the aspects of tax policy (the Redistribution and the Efficiency treatments) or that bring the two together and focus on the trade-off (the Economist treatment). The Redistribution and Economist treatments significantly increase support for more progressive income or estate taxes, while the Efficiency treatment has no effect. There are large partisan gaps in both the final policy views and at every step of the reasoning about the underlying mechanisms of taxes. Democrats’ and Republicans’ divergences in tax policy views can ultimately be traced back to different normative criteria (social preferences) and views of the government, rather than to different perceptions of the efficiency implications of taxation. JEL Codes: D72, D91, H21, H23, H24, H41. I. INTRODUCTION People hold starkly different views on many policies, but their reasons are not always apparent. Policy views may be grounded ∗ I thank participants at the NBER Summer Institute in Political Economy and Public Economics, the Selten Lecture, the Hayek Seminar, the CEPR Political Economy Webinar, the National Tax Association Annual Meetings, the AEA meetings, and seminar attendees at UC Berkeley, Oxford, Arizona State, Princeton, London School of Economics, and Paris School of Economics for comments and feedback. I am grateful to Alan Auerbach, Tim Besley, Richard Blundell, Ilyana Kuziemko, Emmanuel Saez, Gabriel Zucman, and Larry Katz for invaluable comments. I thank Daniele Goffi, Clément Herman, Pierfrancesco Mei Innoncenti, Leonardo D’Amico, Lionel Jeanrenaud, Julia Paul-Venturine, and Yannic Rehm, and most of all Beatrice Ferrario for exceptional research assistance. In memory of my dear friend, colleague, and mentor, Alberto Alesina.  C The Author(s) 2021. Published by Oxford University Press on behalf of the President and Fellows of Harvard College. All rights reserved. For Permissions, please email: [email protected] The Quarterly Journal of Economics (2021), 2309–2369. https://doi.org/10.1093/qje/qjab033. Advance Access publication on September 3, 2021. 2309 Downloaded from https://academic.oup.com/qje/article/136/4/2309/6363701 by guest on 18 January 2022 I study how people understand, reason, and learn about two major tax policies: income taxation and estate taxation. Using large-scale social economics surveys issued to representative U.S. samples and associated experiments, I elicit respondents’ factual knowledge about tax policy and the income or wealth distributions. 2310 THE QUARTERLY JOURNAL OF ECONOMICS Downloaded from https://academic.oup.com/qje/article/136/4/2309/6363701 by guest on 18 January 2022 in perceptions concerning the economic effects of these policies, as individuals may disagree as to which policy will be most effective at achieving a given goal. Alternatively, it could be that people assess the benefits and costs to them and those they care about in contrasting ways. Another reason may be disagreement about what the goals of a policy should be and divergent views on what is fair or just. Consider income taxes. Different views about the right level of income taxes could come from different perceptions of its behavioral or efficiency effects (“Will people stop working if income taxes increase?”). Views might be shaped by how people perceive its distributional effects (“Who benefits if taxes are cut?”) and from the normative criteria people apply when weighting winners and losers (“How fair is income inequality?”). They may also be driven by people’s perceptions of the trustworthiness and efficiency of government (“Will the government waste a lot of the tax revenue?”) or by concerns over how tax revenue will be spent (“Will revenues finance investments in infrastructure or defense, or be redistributed to low-income households?”). In addition, views might be rooted in misperceptions or misunderstandings of the current tax system. Variation in policy views could be traced to one or more of these primary considerations, with different considerations taking precedence among different demographic groups (e.g., based on political affiliation, income, education, and gender). In this article, I assess what people know about two major U.S. tax policies: income taxation and estate taxation. More important, I explore how they reason. My goal is to examine the mechanisms that people think each policy operates through, which factors matter to them, and how they trade off the different considerations when they think about which type of tax policies to support. Do people focus more on the distributional implications of taxes or on their efficiency costs? What social preferences or criteria do they use? How do the answers to these questions differ by the socioeconomic and political characteristics of respondents? I also study how readily people can learn about economic policies and change their views. If people are provided simple but accurate explanations of the distributional and efficiency consequences of economic policies—not unlike those they might receive in an introductory economics class—do their views potentially change? UNDERSTANDING TAX POLICY 2311 Downloaded from https://academic.oup.com/qje/article/136/4/2309/6363701 by guest on 18 January 2022 There are at least three important benefits to understanding how people’s reasoning determines support for or opposition to policies. First, it has some benefits of more structural approaches. By digging into more primitive factors, we can learn about people’s perceived parameters and considerations and flag where agreements and disagreements lie. As is the case with structural models in other settings, this can eventually be used to generalize (or recognize the limits to generalizations), predict, and analyze counterfactual changes. Second, identifying gaps in the public’s knowledge or incoherent reasoning about the policy enables pinpointing where information is needed and which groups particularly require the information. For citizens—perhaps especially for those from economically disadvantaged and disenfranchised groups— gaining a better understanding of these policies is a first step toward supporting the ones beneficial to them. Third, it is critical to disentangle diverging perceptions of economic facts from different value judgments and fairness criteria when it comes to taxes. Do value judgments and fairness criteria differ across people or groups (e.g., do people have heterogeneous aversions to income or wealth inequality?). Or is it more a matter of divergent perceptions about economic facts (e.g., do people disagree on how much inequality there actually is, even if they are similarly averse to it)? Social preferences (i.e., how citizens weigh the gains and losses of different people in the economy) depend on both of these. Misperceptions and lack of knowledge could, to some extent, be corrected with better information. In turn, better information about how the economy actually works may ultimately shape people’s normative views. To this end, I run two large-scale social economics surveys and experiments on representative samples of the U.S. population. The questions that make up these surveys are designed to elicit respondents’ factual knowledge about tax policy and their understanding of its mechanisms. The detailed set of questions asked allows me to decompose policy views into primary factors such as perceived efficiency costs, distributional implications, social preferences and fairness concerns, knowledge or misperceptions, and views of the government. This sheds light on which aspects seem to matter more—at least in a correlational sense— for respondents’ ultimate policy views. To establish a causal link between reasoning and policies, I experimentally show people instructional videos that explain the workings and consequences of each policy from three different perspectives: Redistribution, 2312 THE QUARTERLY JOURNAL OF ECONOMICS 1. Ironically, there is bipartisan agreement that Laffer effects exist for the middle class, although there is no convincing empirical evidence for this phenomenon. Downloaded from https://academic.oup.com/qje/article/136/4/2309/6363701 by guest on 18 January 2022 Efficiency, and Economist. The Redistribution perspective focuses on the winners and losers of each policy, while the Efficiency perspective zeroes in on the efficiency costs. Finally, the Economist perspective presents issues in light of trade-offs, combining both of the previous perspectives. Each video aims to be pedagogical rather than just informational. It seeks to increase knowledge and understanding by explaining mechanisms and reasoning, rather than just providing quantitative facts or information, or expressing normative views on how policies should be. The two main findings are as follows. First, there are very large partisan gaps, not just in the final policy views but at every step of the underlying reasoning about taxes. In fact, there is even a “polarization of reality” (Alesina, Miano, and Stantcheva 2020) whereby Republicans and Democrats do not perceive the current tax system in the same way. Left-leaning respondents tend to think actual taxes are lower and less progressive than do rightleaning ones. On the perceived efficiency effects of income tax changes, respondents generally tend to think that high-income earners are more responsive to income taxes than middle-class households, and that the main channels of response are increased tax evasion, higher likelihood of moving to a different state, and being less entrepreneurial. Yet while Republicans believe that taxes drive larger distortions in behavior, particularly with regard to reduced entrepreneurship and labor supply, Democrats tend to believe that income tax changes will not have as substantial an effect. Regarding the estate tax, respondents across the political spectrum perceive the wealthy to be significantly responsive to changes in it. As the effects of taxes considered move from specific (i.e., the effect on individual taxpayers) to broad (i.e., the overall effect on the economy), partisan disagreement increases sharply. Issues such as whether top income and estate tax cuts induce trickledown effects that benefit everyone or whether they lead to Laffer effects (i.e., increased tax revenues on balance) tend to vary substantially across distinct party lines.1 Some of the most significant contrasts between Democrats and Republicans lie in social fairness considerations about the income and estate tax, the causes of income disparities, the fairness of wealth transmission, and views of the government. Overall, the partisan gap in policy views on UNDERSTANDING TAX POLICY 2313 Downloaded from https://academic.oup.com/qje/article/136/4/2309/6363701 by guest on 18 January 2022 income and estate taxes can be traced back to different normative criteria, social preferences, and views of the government rather than to different perceptions of the efficiency implications of taxation. Second, a Gelbach decomposition of policy views (Gelbach 2016) shows that factors related to social preferences are critically important. One such factor is the perceived benefits of redistribution. Another relates to fairness judgments, such as how fair inequality is perceived to be, whether high earners are entitled to keep a large share of their income, and how fair it is that wealthy parents can pass on wealth to their children. Views of the trustworthiness and scope of government are also crucial drivers of support for taxes. (Notably, these are also the factors that best explain the partisan gap.) Although the belief that higher taxes will hurt the economy is predictive of policy views, it is quantitatively less important than social preferences and views of government. In a nutshell, people who think that inequality in income or wealth is a serious issue or who trust the government would still be more likely to support more progressive tax policy, even if they thought it would have detrimental effects on the economy. The experimental results are in line with the correlational patterns from the decomposition. Both the Redistribution and Economist treatments increase support for progressive taxation, even though the latter also shows the economic costs from taxation. The Efficiency treatment has no effect. Taken together, these findings appear to suggest that redistributive concerns tend to dominate efficiency arguments in the case of income and estate taxes. This article is part of a broader agenda that uses social economics surveys to shed light on people’s reasoning that shapes their policy views. Complementary resources include the extensive Online Appendix, which gathers many additional figures and tables and details on the methodology, and the website understandingeconomics.org for several policies (such as health insurance, trade, income taxes, and the estate tax). Surveys are a key way to uncover intangibles and unobservable factors such as perceptions, attitudes, reasoning, and views, which present challenges for our traditional revealed preference methods. Of course, the trustworthiness of the results depends on the quality of the survey. Meaningful results require an attentive and representative sample and good survey design. This will be discussed in the next two sections. 2314 THE QUARTERLY JOURNAL OF ECONOMICS I.A. Related Literature 2. Several papers specifically investigate the divergence in opinions on the economy between the general public and economists (Blendon et al. 1997; Fuchs, Krueger, and Poterba 1998; Sapienza and Zingales 2013). Downloaded from https://academic.oup.com/qje/article/136/4/2309/6363701 by guest on 18 January 2022 People’s perceptions of their own tax rates and the widespread misunderstanding of the distinction between marginal and average tax rates have been studied in De Bartolome (1995), Gideon (2017) and Ballard and Gupta (2018). “Schmeduling,” that is, approximation heuristics along the income tax rate schedule, is investigated by Rees-Jones and Taubinsky (2019). The literature has focused largely on misperceptions and (lack of) factual knowledge, but less so on reasoning about tax policy, that is, the mechanisms that people think will play out for themselves and others. Slemrod (2006) tries to understand why people support regressive reforms and maps this support to broader misunderstandings of the incidence of taxes. Misperceptions about the incidence of taxes are also prevalent in Bartels (2005) who finds that support for the 2001 and 2003 tax cuts, which primarily benefited very wealthy taxpayers, was mainly driven by considerations over a person’s own taxes, despite widespread opposition to increasing wealth inequality. Perceptions of the broader economy have been studied in Blinder and Krueger (2004), who also rank the determinants of views on major policy issues by order of importance. They find that ideology is the most important one, which is in keeping with this article’s finding that political affiliation is the major predictor, followed by knowledge about the economy and then self-interest.2 The article is also related to the literature that studies the effects of experimentally providing information about inequality or the tax system on support for redistribution. Kuziemko et al. (2015) find only moderate effects stemming from providing this increased information in the United States, because information on the degree of inequality and its rate of increase also has the effect of reducing trust in government. However, telling respondents that the estate tax is—contrary to widespread belief—paid by only one in 1,000 households increases support for it starkly, echoing the results of Sides (2011). Fisman et al. (2020) study the joint preferences over income and wealth taxation using online surveys. Lab experiments using distribution games have been used UNDERSTANDING TAX POLICY 2315 II. CONCEPTUAL FRAMEWORK I illustrate how we can model respondents’ preferred policies by considering the top tax rate and linear bequest rate that they would like to set, were they the social planner. The setting is intentionally kept as simple as possible while still allowing me to frame and interpret the survey design and empirical findings. I discuss some extensions and alternative assumptions below.3 3. The derivation for the top income tax rate formula follows Piketty, Saez, and Stantcheva (2014); the derivation of the estate tax formula follows Piketty and Saez (2013). One simple extension added here is the parameter γ capturing views of the government. Downloaded from https://academic.oup.com/qje/article/136/4/2309/6363701 by guest on 18 January 2022 to disentangle the importance of various factors in shaping support for redistribution (for a recent review, see Cappelen, Falch, and Tungodden 2020). I will highlight the role of trust in government in shaping support for more progressive taxation, which is consistent with the results in Kuziemko et al. (2015), who find that experimentally reducing trust in government reduces support for redistribution as a whole. Di Tella, Dubra, and Lagomarsino (2016) also show that more trust in the government, particularly in business elites, causes a decline in desired taxes on the top 1%. In line with my results, the recent study by Almås, Cappelen, and Tungodden (2020) finds that when comparing attitudes towards redistribution between the United States and Scandinavia, it is views on fairness, not efficiency, that differ. The article is organized as follows. Section II develops the conceptual framework that informs the design of the building blocks of the surveys. Section III describes the data collection, the survey structure, and the sample. Section IV presents respondents’ knowledge about taxes and the income and wealth distributions. Section V digs into respondents’ reasoning about the behavioral responses and distortionary effects associated with tax policy, their distributional effects, and their fairness implications. Section VI first decomposes policy views and the partisan gap into the various factors that shape them and then describes the experimental effects. All appendix material can be found in the Online Appendix. 2316 THE QUARTERLY JOURNAL OF ECONOMICS II.A. Respondent’s Perceived Economic Model (1) ui (c, η, y) = c − hi (y) − ki (η). Let us focus on top earners who make income above z̄ and assume they have a mass of 1. The governmentcan set a linear tax rate τ in the top tax bracket. Let z(1 − τ ) := i:zi z̄ zi di be the av erage income of top-bracket taxpayers and π (1 − τ ) := i:zi z̄ πi di their average rent, which are both functions of the top net-oflog(z) be the elasticity of earnings to the nettax rate. Let e = d dlog(1−τ ) log(π) the elasticity of the rent. Define of-tax rate and eπ = ddlog(1−τ ) z a = z−z̄ to be the Pareto parameter of the top tail of the distribution. The average rent in the economy must come at the expense or benefit of some agents. For simplicity, we assume that all agents bear the average rent uniformly.4 Thus, the government can fully tax or rebate back the average rent or surplus to everyone with a lump-sum tax or transfer (the demogrant). Respondents may also perceive the government as being partially inefficient and wasteful; this is captured by the share γ of revenue that is dissipated instead of being used on productive spending and transfers. 4. This assumption can be relaxed, see Piketty, Saez, and Stantcheva (2014). Downloaded from https://academic.oup.com/qje/article/136/4/2309/6363701 by guest on 18 January 2022 Imagine that each respondent j has their own specific model of the economy in mind, where all parameters have respondentspecific values that correspond to a respondent’s perceptions of them. To reduce notational clutter, I do not explicitly index each parameter by j. Let i index agents in the economy, as perceived by respondent j. Each person i exerts effort to produce output yi and is paid zi = ηi · yi. Pay can differ from marginal product and the gap between the two is π i := (ηi − 1)yi. If ηi > 1, pay is above marginal product, and agent i earns rents, for instance, through monopoly power in their business; if ηi < 1, pay is below marginal product, and agent i creates a positive spillover on others, for example, if they are “job creators” whose economic activity benefits even those with lower incomes through increased employment and career opportunities. Agents face increasing and convex costs of producing output and increasing their pay relative to their output, hi (y) and ki (η). Their utility payoff is UNDERSTANDING TAX POLICY 2317 II.B. Respondent’s Objective (2) gi = g(ci , Ti , wi , X−i , Xi ) where the weight on agent i is a function of their consumption ci , their total tax paid Ti , their effort w i , other personal characteristics captured by vector Xi (e.g., age or family status), and possibly the characteristics of others in the economy, captured by X−i. Some of the social preferences highlighted in the literature are as follows. Utilitarian or welfarist preferences feature weights that are decreasing in disposable income ci , either due to diminishing marginal utility of income (concave utility), or to social aversion to inequality (i.e., the social welfare function is a concave transformation of individual utilities), or both. The libertarian criterion places more weight on people who pay higher taxes, according to the view that people are ultimately entitled to their income (and, pushed to the extreme, that “taxation is theft”) and that those that pay higher taxes are more deserving of tax cuts or transfers. Meritocratic or equality of opportunity criteria place weight on a person’s merit or effort they have had to exert and penalize those that have benefited from “luck” in the form of, say, advantageous circumstances or a good family background. This can be captured by the weight gi increasing in individual effort or merit w i.5 To go from these individual weights to social marginal welfare weights for a given income level, respondents need to average the weights across all individuals earning that income level. This will depend on their perceived composition of people at the 5. For a rigorous microfoundation of this type of weight with equality of opportunity considerations, see Saez and Stantcheva (2016). Downloaded from https://academic.oup.com/qje/article/136/4/2309/6363701 by guest on 18 January 2022 Which objective would respondents want the government to maximize, that is, what would be their social welfare criterion were they making decisions on taxes? A general way to capture the heterogeneous objectives of respondents relies on generalized marginal social welfare weights (Saez and Stantcheva 2016). The weight gi on person i measures the social value (according to respondent j) of transferring $1 to person i. These weights can be used to aggregate the gains and losses from tax changes of different people in the economy. They embody the social preferences of individuals when it comes to taxes and transfers and can depend on their social fairness concerns and many other factors. For instance, we can write: 2318 THE QUARTERLY JOURNAL OF ECONOMICS zi gi . in the economy to be ḡtop = i:zzi z̄ i gi Note that such a formulation is quite general. For instance, a purely self-interest driven objective would imply that the respondent only assigns a positive social marginal weight to themselves, with everyone else receiving a weight of zero. This would lead respondents to prefer the tax rate that most benefits people with the same income level as themselves. II.C. Respondent’s Preferred Top Income Tax Rate According to respondent j, the optimal top income tax rate that the government should set is given by: (3) τ top = 1− 1 ḡtop + a · πz · eπ γ top − ḡγ + a · e , where all parameters are as perceived by the respondent—and may or may not correspond to reality. The optimal top tax rate depends on the following perceived factors. First, the distribution of income and inequality appears in the Pareto parameter a of the top tail. A thicker tail represents higher perceived top income inequality and tends to push the respondent’s desired top tax rate up. The income distribution and inequality also influence the distributive factor ḡ. If the social welfare weights are lower at higher incomes, higher inequality leads to a higher top tax rate. Social preferences will appear in ḡ and reflect on the desired tax rates through any of the channels explained above. For instance, social marginal welfare weights that embody more aversion to inequality (e.g., declining faster in c) will lead to higher preferred tax rates. Social marginal welfare weights that feature aversion to taxes according to libertarian views (e.g., increasing faster in T) will lead to lower preferred tax rates. Economic efficiency concerns are captured by the elasticity e of taxable income to taxes. People may change behaviors in response to income taxes, such as reducing productive activities (e.g., working or saving) or evading taxes. Often discussed is a Downloaded from https://academic.oup.com/qje/article/136/4/2309/6363701 by guest on 18 January 2022 given income level. For instance, if a meritocratic-minded respondent believes that the composition of top earners is mainly tilted toward people who have benefited from luck more than others, they may assign them a lower weight relative to lower-income people. We can now define the income-weighted average marginal social welfare weight on top earners relative to the average weight  UNDERSTANDING TAX POLICY 2319 II.D. Respondent’s Preferred Estate Tax Rate We can also derive the preferred estate (or bequest) tax rate. To simplify the exposition, I consider the tax rate in a steady state of the economy, assuming that bequests enter parents’ utility (i.e., parents care about leaving bequests to their children), there are spillovers, the elasticity of bequests is uncorrelated with the individual welfare weight gi , and labor income is inelastic:6 (4) τB = 1− 1 parents ḡchildren (1+e B)+ R ḡ γ  (1 + e B) · 1 − ḡchildren γ , 1−τ B db is the steady-state elasticity of aggregate where eb = d(1−τ b B) bequests b with respect to the net-of-tax rate 1 − τ B. ḡchildren is the bequest-weighted marginal social welfare weight on heirs, and ḡparents is the bequest-weighted marginal social welfare weight on parents. Similar effects to the ones in the income tax formula appear for the bequest tax. The efficiency cost of bequest taxes (eB ) 6. Many of these can be relaxed but lead to more complex formulas; see Piketty and Saez (2013). Downloaded from https://academic.oup.com/qje/article/136/4/2309/6363701 by guest on 18 January 2022 Laffer effect, whereby an income tax cut could ultimately lead to an increase in tax revenues by stimulating economic activity by so much that the additional revenue collected on that increased activity outweighs the direct loss in revenue. The preferred tax also depends on perceived spillover effects from top tax rates, through either trickle-up or trickle-down effects embodied in πz · eπ. Lack of trust in or perceived inefficiency of government, captured by a smaller γ , will lead to lower desired tax rates because the respondent believes that a larger share of revenues will be wasted or dissipated. To sum up, people’s preferred top income tax rate can differ based on their perceptions of the income distribution and inequality, efficiency costs of taxes, possible spillovers, government efficiency, and their social preferences which embody their social fairness concerns. All else equal, respondents who think taxes will hurt the economy, do not consider inequality to be a serious problem, believe in trickle-down effects, or think the government is not to be trusted with revenues would prefer a lower top tax rate. 2320 THE QUARTERLY JOURNAL OF ECONOMICS II.E. Knowledge about the Current Tax System Knowledge of the current tax system may not necessarily be related to people’s preferred tax rates, but it would naturally shape support for particular tax reforms. Imagine two respondents, A and B, have the same desired level of tax progressivity, but A lacks knowledge about current taxes and wrongly perceives the tax system to be more progressive than B. Respondent A may not support progressive tax reform, even if they ultimately agree with respondent B on how their ideal tax system should look like. Results in Section VI.C are consistent with this. II.F. Additional Remarks This framework gives us a way to think about respondents’ preferred tax policies as a function of the parameters they Downloaded from https://academic.oup.com/qje/article/136/4/2309/6363701 by guest on 18 January 2022 captures behavioral responses to taxes that reduce the bequest tax base, such as saving less or choosing to leave smaller bequests. The distribution of estates appears in the distributional factors ḡchildren and ḡparents. Views of government are captured by γ. The major qualitative difference to the income tax formula lies in the social fairness considerations that are relevant to estate taxation. The thorny fairness issues revolve around two conflicting concerns. Respondents who consider the question from the point of view of the parents may view the ability of the latter to pass on wealth to their children tax-free to be relatively fair, perhaps based on the reasoning that people can purchase any types of goods with their money and spend it how they wish. However, once that respondent considers the point of view of the children, they may feel it is unfair that some children receive much higher wealth from their parents through no fault or merit of their own, perhaps based on an equality of opportunity argument. Respondents who believe more strongly in leveling the playing field for children would have a lower ḡchildren , which leads to a higher desired bequest tax; respondents who instead mostly wish to respect parents’ choices to pass on wealth to their children would have a higher ḡparents , which implies a lower preferred bequest tax. For respondents who have some aversion to wealth inequality (as embodied by the social marginal welfare weights decreasing in the level of bequests), ḡchildren and ḡparents will be lower when perceived wealth or estate inequality is higher, which will lead to a higher desired estate tax. UNDERSTANDING TAX POLICY 2321 II.G. A Roadmap The survey contains questions aimed at eliciting each of these considerations about income and estate taxes. These will be the explanatory variables (on the right-hand side) for policy views (on the left-hand side). The steps to achieve this translate into the design of the survey. I start by asking respondents for detailed socioeconomic background questions and political affiliation. This captures characteristics such as income, family status, or partisanship that can shape perceptions as well as one’s private benefit and self-interest from taxes. Then, I study the factual knowledge of respondents about taxes and the income and wealth distributions. After this, I take a deep dive into uncovering the reasoning on efficiency costs, distributional effects, and fairness concerns before asking about overall policy views. Respondents are also asked about views of the role, competence, and efficiency of the government. In Section VI.C, I estimate the effect of these various components on policy views and decompose the partisan gap in policy views. Finally, to establish causality about how the main factors that enter the tax formula (i.e., ḡ and e) shape policy views on taxes, I provide randomized video courses that will shift these 7. Furthermore, although the intention is to capture the most frequently discussed fairness concerns related to taxation through the survey questionnaire, there may be other fairness concerns that could be explored, too. Downloaded from https://academic.oup.com/qje/article/136/4/2309/6363701 by guest on 18 January 2022 perceive. It is quite general. For instance, respondents may overor underemphasize some of the concerns, such as the elasticity of income to taxes. Some terms may entirely drop out of the formula or dominate it. In addition, the social welfare weights formulation allows for a wide range of objectives that respondents may adopt, including a self-interest-driven one. The public economics literature has derived formulas for many other cases for which the effects discussed here would directly carry over, at the expense of more notation and derivations, for example, a nonlinear income tax (Saez 2001), a nonlinear capital or wealth tax (Saez and Stantcheva 2018) or bequest tax (Piketty and Saez 2013).7 How revenue is spent may matter to respondents and will be asked in the survey and studied below. However, incorporating the many possible types of spending in a tractable way in the formal model is challenging because it requires modeling people’s perceptions about how efficient each type of spending or program is, whom it benefits, and so on. 2322 THE QUARTERLY JOURNAL OF ECONOMICS factors. The videos explain these effects of taxes pedagogically, the way one might do in a very brief introductory economics class. III. SURVEY DESIGN AND DATA The core data come from two surveys, conducted between February and May 2019 on U.S. residents aged 18 to 69. The sample sizes are 2,780 for the income tax survey and 2,360 for the estate tax. The surveys were distributed by the commercial survey company Respondi (https://www.respondi.com/EN/). 1. How Were Participants Enrolled? Online Appendix OA4.2 provides more information on how the sample is contacted, the survey company’s targeting, and the overall response rates. In brief, the commercial survey company has a large pool of vetted survey respondents that come from different panels, who can freely log in and take surveys in exchange for rewards. Surveys are typically consumer surveys, that is, related to consumer products or experiences. Respondents come from a variety of backgrounds and are rewarded through a range of different means, such as cash and points on reward programs. Compensation for completed surveys is implemented by each panel company and varies based on the type of arrangement and preferences of the respondent. On the survey entry page, respondents were only told the length of the survey, but not the topic or the sender. This is important to avoid selection based on the topic. After clicking on the link, respondents were channeled to a consent page (see Online Appendix Figure OA-1) that informed them that they were about to take an academic research survey, destined solely for research purposes and run by nonpartisan researchers. They were asked to respond accurately to the best of their knowledge and were assured that participation was entirely voluntary. Respondents were then guided through some screening questions that ensured that the final sample was nationally representative along gender, age, and income dimensions, as well as through the detailed background socioeconomic questions. This meant that if respondents decided to drop out at some point during the survey—say, upon learning the topic of the survey—all their demographic and background information would be known, allowing me to check for differential attrition by observable characteristics such as Downloaded from https://academic.oup.com/qje/article/136/4/2309/6363701 by guest on 18 January 2022 III.A. Data Collection and Final Sample UNDERSTANDING TAX POLICY 2323 political affiliation. There is no clear differential attrition (see Online Appendix Table OA-11). 3. Final Sample. Table I shows the characteristics of the samples and compares them to U.S. population statistics. The final samples for each of the two surveys are close to representative of the U.S. population along many dimensions. This is true by construction for the targeted dimensions of age, gender, income (arranged into five brackets to mimic the way the quotas were imposed during the surveys). In addition, the sample is also broadly representative on many nontargeted dimensions, such as the share of respondents who are married and the share of employed, unemployed, and self-employed. Importantly, the sample is also representative along the political affiliation dimension and on the voting patterns of the 2016 presidential election. However, in both surveys, respondents were more likely to have completed high school and be college-educated than the general population. In addition, African American and Hispanic populations are underrepresented. To address these imbalances, one can reweight the sample so that it is representative along the employment, education, and race dimensions as well. Reweighting was done for all results as checks. Result were not significantly affected. III.B. The Survey Structure The full questionnaires of the Income Tax Survey and the Estate Tax Survey are in Online Appendix OA-2, with a link that leads to the web interfaces of the surveys. This section provides more details on some of the more important survey blocks.8 8. The surveys contain more questions than are exploited in the main part of the article. They are partially analyzed in the Online Appendix and leave room for future analysis using this data. Open-ended questions that extract the firstorder thinking and considerations of respondents, without priming them, are also Downloaded from https://academic.oup.com/qje/article/136/4/2309/6363701 by guest on 18 January 2022 2. Data Quality. Online Appendix OA-4 describes methods used in the survey to ensure quality responses and to check for data quality. The median time for completion of the survey was 35 minutes. I benchmark this against other common surveys and check for possible survey fatigue in Online Appendix OA-4.3. For the benchmark sample, I drop respondents in the bottom 5% of the survey time distribution. The results are not affected by trimming these outliers (see Online Appendix OA-8.2). 2324 THE QUARTERLY JOURNAL OF ECONOMICS TABLE I SAMPLE CHARACTERISTICS Income tax survey Estate tax survey Male 0.49 0.48 0.46 18–29 years old 30–39 years old 40–49 years old 50–59 years old 60–69 years old 0.24 0.20 0.18 0.19 0.19 0.23 0.20 0.19 0.21 0.18 0.22 0.20 0.19 0.19 0.19 $0–$19,999 $20,000–$39,999 $40,000–$69,999 $70,000–$109,999 $110,000+ 0.13 0.16 0.21 0.20 0.31 0.15 0.19 0.23 0.19 0.24 0.16 0.19 0.24 0.19 0.20 Four-year college degree or more High-school graduate or less 0.34 0.38 0.48 0.19 0.46 0.19 Employed Unemployed Self-employed 0.70 0.03 0.07 0.63 0.07 0.07 0.62 0.06 0.06 Married 0.53 0.55 0.53 White Black/African American Hispanic/Latino Asian/Asian American 0.61 0.12 0.18 0.06 0.76 0.06 0.06 0.07 0.76 0.06 0.07 0.07 Democrat Republican Independent 0.30 0.26 0.42 0.34 0.31 0.33 0.35 0.30 0.33 Voted for Clinton at the 2016 presidential election Voted for Trump at the 2016 presidential election 0.48 0.46 0.44 0.44 0.44 0.44 2,784 2,360 Sample size Notes. This table displays statistics for the overall U.S. population and compares it to the characteristics of the samples of the income tax and estate tax surveys. National statistics on gender, age, income brackets, race, education, marital status, and employment status are from the IPUMS-CPS-ASEC data set for March 2019 (Flood et al. 2020). National statistics on party affiliation for March 2019 are from Gallup (2019). Statistics on the 2016 presidential election results are from Leip (2019). See Online Appendix OA-4.4 for details on how the summary statistics are constructed. included. They ask respondents about their main considerations, perceived goals, and shortcomings about the taxes, as well as the effects they anticipate (e.g., which groups would gain or lose from a tax increase). They are analyzed in Ferrario and Stantcheva (2021). Downloaded from https://academic.oup.com/qje/article/136/4/2309/6363701 by guest on 18 January 2022 U.S. population UNDERSTANDING TAX POLICY 2325 2. Knowledge. These questions are about factual knowledge.11 They refer among others to the level of the top federal or top state taxes, the threshold for the top income tax bracket, the top tax rates in the 1950s, the share of households in the top bracket or who pay the estate tax, the share of households paying no income tax, and the average tax rate for the median household or the top household. Respondents are also asked about the share of total income or wealth that goes to the top 1%, the share of wealth inherited, or the occupational composition of the top 1% of earners. To test whether the results may be driven by respondents’ possible lack of attention or careless answers, 85% of the sample received monetary incentives for accurate answers. Misperceptions are virtually unaffected by monetary incentives, suggesting that respondents are already responding to the best of their knowledge. 3. The Information Treatments: Short “Econ 101” Video Courses. In each survey, a randomly chosen subsample of 9. “On economic policy matters, where do you see yourself on the liberal/conservative spectrum?” With options [Very Liberal, Liberal, Center, Conservative, Very Conservative]. 10. I also dug deeper into their political participation by asking whether they were registered to vote, and why not, if they were not; and whether they regularly vote, or why they do not. 11. They are carefully designed to be intuitive and easy to understand. For instance, I ask for a number “out of 100” rather than for a percentage. I first explain what a “share” is, and how post- and pretax income are related for a given tax rate with the use of a figure and worked-out examples. Downloaded from https://academic.oup.com/qje/article/136/4/2309/6363701 by guest on 18 January 2022 1. Background Socioeconomic Questions. I collected information on respondents’ gender, age, income, highest level of education achieved, field of study in college, sector of occupation, employment status, marital status, number of children, place of residence, main source of news, and political orientation. Because it plays such an important role, I investigated the latter category in three ways. First, I asked respondents to classify themselves in terms of their views on economic policy, along a spectrum ranging from “very conservative” to “very liberal.”9 Second, I asked them what they consider their political affiliation to be (Republican/Democrat/Independent/Other/Nonaffiliated). Third, I asked them for whom they voted in the 2016 presidential elections and, if they did not vote, for whom they would have voted.10 2326 THE QUARTERLY JOURNAL OF ECONOMICS (B) Efficiency video {https://youtu.be/9xd-RHMiIcE} (C) Economist video {https://youtu.be/e3NBmrzEmUQ} FIGURE I Income Tax Treatment Videos respondents was shown one of three versions of an instructional video that provided information about the policy. They are described in more detail in Section VI.D, with screenshots in Figures I and II. 4. Reasoning about Taxes. In this section, respondents are asked to think in more detail about how each tax policy works. What behavioral responses, efficiency effects, and effects on the broader economy will it trigger? What are the distributional consequences for different groups? What fairness concerns do people have? To give a concrete example, in the income tax module, respondents are asked to what extent they believe people will engage in the following behaviors if their taxes were to increase: save less, work less, stop working altogether, evade taxes, and so on. They are also asked which income groups will gain most or least if income taxes on high-income earners are increased and whether taxing away the income of different groups is fair or unfair.12 12. As explained in Online Appendix OA-2, these questions were asked in three different, randomized formulations—a randomization that is not exploited in the current article. When using these questions in regression analysis, we control for indicators for the type of formulation used. Downloaded from https://academic.oup.com/qje/article/136/4/2309/6363701 by guest on 18 January 2022 (A) Redistribution video {https://youtu.be/_vq7ZTjBN3Y} UNDERSTANDING TAX POLICY 2327 (A) Redistribution video (B) Efficiency video {https://www.youtube.com/watch?v =pZ47JuiqoOU} (C) Economist video {https://www.youtube.com/watch?v=qnT 07wO8nVA} FIGURE II Estate Tax Treatment Videos 5. Policy Views. Respondents were asked about their views on the current tax systems (Is it fair? Are they satisfied with it?) and for opinions regarding concrete policy actions, for example, if they supported increasing the tax to fund specific programs. 6. Views of Government. In this section, respondents expressed their views about the role and capacity of the government to deal with the issue at hand. For instance, in the income tax module, they are asked to what extent they think the government has the tools and ability to reduce income inequality. They were also asked about their general attitudes toward government (unrelated to the specific policy) such as whether they tend to trust the government and what the scope of government intervention should be. 7. Final Questions. By entering the survey, respondents were informed that they were automatically enrolled in a lottery to win $1,000. At the end of the survey, respondents were asked Downloaded from https://academic.oup.com/qje/article/136/4/2309/6363701 by guest on 18 January 2022 {https://www.youtube.com/watch?v =Wz5Xr723tJk} 2328 THE QUARTERLY JOURNAL OF ECONOMICS IV. KNOWLEDGE ABOUT INCOME AND ESTATE TAXES This section presents the main findings on knowledge about the tax system and the income and wealth distributions. Table II shows some of the knowledge variables about income taxes (in Panel A) and estate taxes (in Panel B), regressed on the full array of respondent characteristics. For clarity, only the coefficients on the four characteristics that exhibit the largest heterogeneity are shown in the table, namely, the indicators for being Republican, high income, having high self-reported knowledge about tax policy, and having a college degree. The bottom rows in each panel show the average perception as well as the true values of the variables. The complete set of knowledge questions and additional heterogeneity patterns are analyzed in Online Appendix OA-3. IV.A. Misperceptions of the Tax System Respondents can quite accurately identify today’s top tax rate. They believe it is 31% when it is 37% (column (1)). However, they believe on average that the top tax rate in the 1950s was 33%, similar to today’s tax rate, when it was, in fact, much higher at 91% (column (2)). The perception of the median-income household’s (column (4)) and the top-tax bracket household’s (column (5)) average taxes are closer together than in actuality, implying that the level of progressivity of the tax system is misunderstood. In the spirit of schmeduling, respondents smooth out the tax schedule in their mind and overinflate the tax paid by the median household by a factor of two, while slightly underestimating the tax paid by the top bracket household (Rees-Jones and Taubinsky 2019). Respondents also greatly underestimate the top tax bracket’s income threshold (column (3)), placing it at around $188,000 Downloaded from https://academic.oup.com/qje/article/136/4/2309/6363701 by guest on 18 January 2022 whether they were willing to forfeit part of their lottery gain to receive the accurate answers to all the knowledge questions. They had to commit to forfeiting that amount before they knew whether they had won the lottery. The price of information was randomized (with possible values $1, $2, $5, and $10), allowing me to extract a willingness to pay for information. They were also asked whether they feel the survey was biased (left-wing biased or right-wing biased) and could leave open-ended feedback in a text-entry box. All variables are comprehensively defined in Online Appendix OA-1 and more briefly in the table and figure notes. Descriptive statistics: Actual value Average perception Observations College degree 37 31 2,779 3.74∗∗∗ (0.84) 0.23 (0.84) 2.78∗∗∗ (0.76) 0.93 (0.72) Top tax rate today (1) 91 33 2,779 − 2.52∗ (1.38) 1.32 (1.39) 8.29∗∗∗ (1.26) 6.12∗∗∗ (1.19) 600,000 187,914.8 2,651 − 8,632.43 (8,915.36) 59,858.63∗∗∗ (8,946.39) 24,268.44∗∗∗ (8,163.86) 39,112.78∗∗∗ (7,714.02) 13 26.3 2,780 1.46∗ (0.80) − 0.00 (0.80) 2.39∗∗∗ (0.73) − 0.41 (0.69) Share of income paid in taxes by median Top tax rate in the 50s Top tax threshold households (2) (3) (4) 32.7 27.4 2,777 6.15∗∗∗ (0.88) 0.15 (0.89) 3.70∗∗∗ (0.81) 0.78 (0.76) Share of income paid in taxes in top bracket (5) Tax system 0.73 20.3 2,762 3.24∗∗∗ (1.03) − 1.76∗ (1.04) − 0.11 (0.94) − 4.94∗∗∗ (0.89) Share of households in top bracket (6) Income distribution 44 25.3 2,779 5.97∗∗∗ (0.98) 0.08 (0.98) 5.38∗∗∗ (0.89) 0.40 (0.85) 20 44.7 2,780 − 7.72∗∗∗ (1.41) − 2.32 (1.41) 5.53∗∗∗ (1.28) 6.52∗∗∗ (1.22) 0.37 2,783 − 0.08∗∗∗ (0.02) 0.01 (0.02) 0.07∗∗∗ (0.02) 0.04∗∗ (0.02) Share of Share of U.S. households income not paying earned by top income taxes 1% WTP for info (7) (8) (9) Downloaded from https://academic.oup.com/qje/article/136/4/2309/6363701 by guest on 18 January 2022 Self-reported knowledge High income Panel A: Income tax Republican TABLE II KNOWLEDGE, MISPERCEPTIONS, AND WILLINGNESS TO PAY FOR INFORMATION UNDERSTANDING TAX POLICY 2329 40 33 2,350 − 0.54 (1.05) − 0.16 (1.06) 4.03∗∗∗ (0.97) 0.00 (0.92) 77 29 2,335 − 3.51∗∗∗ (1.24) 0.80 (1.25) 6.48∗∗∗ (1.15) 4.33∗∗∗ (1.08) 0.7 364.1 2,357 16.15 (15.53) − 42.81∗∗∗ (15.65) 5.81 (14.34) − 50.69∗∗∗ (13.57) 11,400,000 2,428,139.6 2,357 − 486,504.56∗∗∗ (182,797.77) 1,111,072.07∗∗∗ (184,273.68) 792,758.06∗∗∗ (168,747.59) 818,974.82∗∗∗ (159,750.72) Exemption threshold (4) 55 45.7 2,354 − 4.92∗∗∗ (1.24) 1.94 (1.25) 3.32∗∗∗ (1.14) 1.22 (1.08) Share of estates unrealized capital gains (5) ≈ 50 41.9 2,357 − 2.96∗∗ (1.22) − 1.81 (1.23) 1.74 (1.13) 2.23∗∗ (1.07) Share of wealth inherited (6) 41.8 49.1 695 − 7.13∗∗ (2.79) 1.81 (2.75) − 0.98 (2.50) 8.65∗∗∗ (2.38) Share of wealth owned by top 1% (7) Wealth distribution WTP for info (9) − 0.02 (0.02) 0.02 (0.03) 0.11∗∗∗ (0.02) 0.05∗∗ (0.02) 0.40 2,360 Share of wealth owned by bottom 50% (8) 1.64 (1.01) − 0.35 (1.00) 0.74 (0.91) − 2.82∗∗∗ (0.86) 2.5 12.5 695 Downloaded from https://academic.oup.com/qje/article/136/4/2309/6363701 by guest on 18 January 2022 Notes. The values of the dependent variables in columns (1)–(8) are deviations of the respondents’ answers from the “Actual value” reported in the lower sets of rows under “Descriptive statistics”; a positive sign on the “Average perception” indicates that, on average, respondents overestimate the actual value, and a negative sign means they underestimate it. The dependent variable in column (9) is an indicator variable equal to 1 if the respondent is willing to pay the randomized amount of money ($1, $2, $5, or $10) to receive the correct answers to all the knowledge questions (see Section III). The data sources for the actual values of these variables are described in Online Appendix OA-3, Tables OA-1 and OA-2. Standard errors are in parentheses. ∗ p <.1, ∗∗ p <.05, ∗∗∗ p <.01. Descriptive statistics: Actual value Average perception Observations College degree Self-reported knowledge High income Panel B: Estate tax Republican Estate tax rate today (1) No. households Estate tax out of 1,000 rate in the paying estate tax 50s (2) (3) Tax system TABLE II CONTINUED 2330 THE QUARTERLY JOURNAL OF ECONOMICS UNDERSTANDING TAX POLICY 2331 IV.B. Misperceptions of the Underlying Income and Wealth Distributions Table II, Panel B, column (8) indicates that respondents strongly overestimate the share of income going to the top 1% by 25 percentage points on average. Respondents appear to be more accurate on the wealth distribution at the top (in Panel B, column (7)), where they believe that the wealth share of the top 1% is 49% (as opposed to the 42% estimated in recent work, which is itself subject to considerable on-going debate about whether it is an upper bound). Yet respondents also believe that the bottom 13. As explained below, the share of respondents who feel that the estate tax has important direct effects on their lives is 30%, which is interestingly close to respondents’ perceived share of households who are subject to the estate tax. Downloaded from https://academic.oup.com/qje/article/136/4/2309/6363701 by guest on 18 January 2022 annual income instead of the actual $600,000 annual income. Accordingly, respondents believe that the top tax rate applies to many more households (20%) than it actually does (0.73%). They also tend to mentally smooth the share of households in extreme tax brackets: they underestimate the share of households that do not pay income taxes (believing it is 25% when the reality is 44%, as shown in column (7)) but overestimate the share of households in the top bracket (column (6)). When it comes to the estate tax in Panel B, and, as was the case for the income tax, respondents are relatively accurate about the tax rate that applies above the exemption threshold today (column (1)), but they are unaware of how high the 1950s tax rate was and assume it was close to today’s rate (column (2)). The next two columns show that respondents believe on average that the share of households paying the estate tax is 364 out of 1,000 households (the median perception is 300 out of 1,000). The reality is less than 1 out of 1,000 households. Qualitatively (but not quantitatively) consistent with this stark misperception is that respondents believe that the estate tax exemption threshold is much lower than it actually is: the average perception is $2.4 million (the median is lower than $0.5 million) while the reality is $11.4 million per person. The fact that respondents starkly overestimate the share of people who pay the top tax rate and the estate tax could mean that they may (mistakenly) consider themselves more likely to be directly affected by these policies targeted at the top earners and wealth holders—now or in the future—than is actually the case.13 2332 THE QUARTERLY JOURNAL OF ECONOMICS IV.C. Who Knows More? Higher-income respondents are more aware of variables that affect the top of the distribution, such as the top tax rate, the top tax bracket, the exemption threshold for the estate tax, and the share of households that pay the estate tax. They do not, however, differ in their misperceptions of the other variables. People with higher self-reported knowledge (i.e., who say they know more about tax policy) generally do have smaller misperceptions on most (but not all) margins. Those with a college degree are generally more accurate than those without, except that they tend to overestimate the shares of income and wealth of the top (but are more accurate on the lower share of wealth of the bottom 50%). Republican respondents in general tend to think that taxes are higher and more progressive than Democrats do: they perceive a higher top tax rate, a higher share of income paid by households in the top bracket, a higher share of households in the top bracket, and a higher share not paying any income tax. They also think that the estate tax exemption threshold is lower (however, both Democrats and Republicans overestimate the share of households that pays the estate tax to the same extent). Republicans are even Downloaded from https://academic.oup.com/qje/article/136/4/2309/6363701 by guest on 18 January 2022 50% holds 12% of wealth, which is much higher than the actual 2%. Respondents thus inflate the two extremes of the wealth distribution. Note that respondents’ beliefs about which professions compose the top 1% richest people in the United States do not fit reality for all categories (see Online Appendix Figure OA-6). Respondents tend to think that among the top 1% richest people, there are more entrepreneurs, arts, media and sports personalities, teachers, and scientists than there truly are. This could be because these professions are more often seen in the media. Conversely, respondents tend to underestimate the share of executives/managers and physicians in the top 1%. Respondents are relatively accurate in their perception of the share of estates that are unrealized capital gains that have never been taxed (46% relative to 55% in reality, in Panel B, column (5)). Furthermore, they believe that 42% of wealth is inherited (column (6) of Panel B). Unlike for the variables about the tax system, there is considerable uncertainty around this number, with current estimates by economists ranging from 34% to 45% (Kopczuk and Lupton 2007) to 56%–64% (Alvaredo, Garbinti, and Piketty 2017). UNDERSTANDING TAX POLICY 2333 IV.D. Willingness to Pay for Correct Information The last column in each panel reports an indicator for whether the respondent is willing to pay a randomized price to see the accurate answers to all the knowledge questions (see Section III). On average, around 40% of respondents are willing to pay to learn more about the income and estate tax and the income and wealth distributions. To interpret findings on the demand for information here, note that it is partially a private good, since respondents are directly affected by tax policy. Online Appendix Figure OA-11 shows that 53% of all respondents feel that the income tax has important direct effects on their own lives; 30% think so for the estate tax. In addition, respondents may have strong social preferences and thus care about being aware of the current tax system or the distributions of income and wealth. But information could also be viewed in part as a public good if better informed voters are able to make better policy choices for the economy as a whole. Republican respondents are less likely than Democrats to be willing to pay for accurate information on income taxes, even conditional on the (randomized) price for information and their own income. That is not the case for estate taxes. Those with more self-reported knowledge and those who have a college degree are generally more willing to pay for information for the income and 14. There is an interesting contrast between perceptions of the present and past here. Republicans are more likely to overestimate today’s top tax rate, while also underestimating the past tax rate, thus perhaps casting the past in a more favorable light (from the point of view of supporters of lower taxes). This lack of awareness that the United States used to be a higher tax country can contribute to the rhetoric of high taxes being against “American ideals.” 15. This may also be consistent with the motivated reasoning in Bénabou and Tirole (2006, 703), whereby individual beliefs are “(consciously or not) shaped as much by their own functional goals and psychological needs as by actual data: to a certain extent, people believe what they want to believe.” Downloaded from https://academic.oup.com/qje/article/136/4/2309/6363701 by guest on 18 January 2022 less likely than Democrats to be aware of the high top tax rates or estate taxes in the 1950s.14 They believe that a lower share of income goes to the U.S. top 1% and are hence more accurate than Democrats on this issue. In addition, they think that the share of wealth that is inherited and the share of wealth owned by the top 1% are lower. These results are in line with a “polarization of reality” (Alesina, Miano, and Stantcheva 2020)—that is, polarization even in the perception of facts.15 2334 THE QUARTERLY JOURNAL OF ECONOMICS V. REASONING ABOUT TAXES: EFFICIENCY, DISTRIBUTION, AND SOCIAL PREFERENCES This section presents the reasoning of respondents on the efficiency effects and distributional effects of taxes, as well as their social preferences related to taxes. All variables are explained in the main text and table or figure notes, and in more detail in Online Appendix OA-1. V.A. Perceived Behavioral Responses and Distortionary Effects of Taxes 1. Income Taxes. Table III shows the perceived behavioral responses to income taxation. The regressions systematically include controls for respondents’ gender, age, race, income bracket, having children, education, having an economics-related major, employment status, self-reported policy knowledge, self-reported social class, political affiliation, and indicator variables for all treatments. The first panel shows the coefficients on some of the key covariates, namely, political affiliation (Republican), age groups, and income bracket. Panel B shows the main effects of the video treatments, where the branches are denoted by Redistribution, Efficiency, or Economist. The bottom panel provides descriptive statistics. The control group designates respondents who did not receive any video course.17 In Online Appendix OA-6, I test for differential effects by gender or political affiliation. 16. Note in Online Appendix Figure OA-11 and Table OA-14 that those who self-report more knowledge also feel more directly affected by each policy in their own lives, which is perhaps why they know more in the first place. Republicans are more likely to say that the estate tax has important direct effects on their lives. There is no heterogeneity by education. 17. In addition, the control group is only respondents who saw the neutral formulation of the questions when the question was asked differently, see the questionnaire in Online Appendix OA-2. Downloaded from https://academic.oup.com/qje/article/136/4/2309/6363701 by guest on 18 January 2022 the estate tax. Although this sounds paradoxical at a first glance, it is consistent with findings in earlier work on other issues such as immigration (Alesina, Miano, and Stantcheva 2019, 2020). It also offers a possible channel for the perpetuation of misinformation, if those people who know less are also the ones least willing to pay for new information. Beyond these, there are no significant heterogeneities in willingness to pay for information across respondents.16 TABLE III PERCEIVED BEHAVIORAL RESPONSES TO INCOME TAXATION Evade taxes High earners (1) Middle class (2) Stop working Spouse stop working Move state Be less entrepreneurial Middle class (4) High earners (5) Middle class (6) High earners (7) Middle class (8) High earners (9) Middle class (10) High earners (11) 0.14∗∗∗ (0.02) 0.00 (0.02) − 0.06∗∗ (0.03) − 0.13∗∗∗ (0.03) − 0.02 (0.03) − 0.04 (0.02) 0.14∗∗∗ (0.02) 0.01 (0.02) − 0.06∗∗ (0.03) − 0.11∗∗∗ (0.03) 0.00 (0.02) − 0.03 (0.02) 0.09∗∗∗ (0.02) − 0.01 (0.02) − 0.06∗∗ (0.03) − 0.12∗∗∗ (0.03) − 0.02 (0.02) − 0.01 (0.02) 0.11∗∗∗ (0.02) − 0.00 (0.02) − 0.04∗ (0.03) − 0.13∗∗∗ (0.03) − 0.01 (0.02) − 0.02 (0.02) 0.13∗∗∗ (0.02) 0.01 (0.02) − 0.03 (0.03) − 0.08∗∗∗ (0.03) − 0.03 (0.03) − 0.03 (0.02) 0.13∗∗∗ (0.02) 0.01 (0.02) − 0.05∗ (0.03) − 0.11∗∗∗ (0.03) − 0.02 (0.02) − 0.04∗ (0.02) 0.10∗∗∗ (0.02) − 0.05∗∗∗ (0.02) − 0.05∗∗ (0.02) − 0.10∗∗∗ (0.02) 0.02 (0.02) − 0.02 (0.02) 0.17∗∗∗ (0.02) 0.02 (0.02) − 0.04∗ (0.02) − 0.11∗∗∗ (0.03) 0.00 (0.02) − 0.00 (0.02) 0.18∗∗∗ (0.02) − 0.01 (0.02) − 0.03 (0.03) − 0.06∗∗ (0.03) − 0.03 (0.03) − 0.04∗ (0.02) 0.19∗∗∗ (0.02) 0.00 (0.02) − 0.04∗ (0.03) − 0.12∗∗∗ (0.03) − 0.02 (0.03) − 0.02 (0.02) 0.04 (0.03) 0.12∗∗∗ (0.03) 0.11∗∗∗ (0.02) − 0.03 (0.03) 0.14∗∗∗ (0.03) 0.15∗∗∗ (0.03) 0.08∗∗ (0.03) 0.19∗∗∗ (0.03) 0.18∗∗∗ (0.03) Panel B: Video treatment effects Redistribution 0.00 0.04 − 0.02 0.06∗ (0.03) (0.03) (0.03) (0.03) Efficiency 0.08∗∗∗ 0.07∗∗ 0.16∗∗∗ 0.27∗∗∗ (0.03) (0.03) (0.03) (0.03) Economist 0.06∗∗∗ 0.05∗ 0.17∗∗∗ 0.28∗∗∗ (0.02) (0.03) (0.03) (0.03) − 0.01 (0.03) 0.08∗∗ (0.03) 0.04 (0.02) 0.06∗∗ (0.03) 0.17∗∗∗ (0.03) 0.17∗∗∗ (0.02) − 0.04 0.07∗∗ (0.03) (0.03) 0.14∗∗∗ 0.20∗∗∗ (0.03) (0.03) 0.12∗∗∗ 0.22∗∗∗ (0.03) (0.03) − 0.02 (0.03) 0.04 (0.03) 0.04∗ (0.02) Middle class (12) 2335 High earners (3) UNDERSTANDING TAX POLICY Panel A: Personal characteristics Republican − 0.04∗ 0.08∗∗∗ (0.02) (0.02) Female − 0.05∗∗∗ − 0.09∗∗∗ (0.02) (0.02) Age 30–49 − 0.01 − 0.02 (0.02) (0.03) Age 50–69 − 0.02 − 0.02 (0.02) (0.03) Middle income 0.01 − 0.02 (0.02) (0.02) High income 0.02 − 0.00 (0.02) (0.02) Work less Downloaded from https://academic.oup.com/qje/article/136/4/2309/6363701 by guest on 18 January 2022 2336 Evade taxes Work less Stop working Spouse stop working Move state Be less entrepreneurial High Middle High Middle High Middle High Middle High Middle High Middle earners class earners class earners class earners class earners class earners class (1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (11) (12) Panel C: Descriptive statistics Control mean 0.80 Male control mean 0.84 Democrat control mean 0.84 Observations 2,782 0.60 0.66 0.53 2,782 0.48 0.50 0.45 2,783 0.39 0.40 0.34 2,781 0.33 0.33 0.33 2,781 0.28 0.31 0.25 2,781 0.43 0.42 0.41 2,783 0.32 0.32 0.29 2,781 0.78 0.80 0.75 2,783 0.64 0.63 0.59 2,782 0.50 0.52 0.41 2,782 0.45 0.46 0.39 2,782 Note. The dependent variables are indicator variables equal to 1 if the extent to which a respondent thinks that an increase in the federal personal income tax would encourage middle-class or high-income individuals toward the behaviors listed ranges from a moderate amount to a great deal. All variables are defined in more detail in Online Appendix OA-1. Regressions in all panels include controls for sex, age, race, income class, having children, education, having an economics-related major, employment status, self-reported policy knowledge, self-reported social class, political affiliation, and indicator variables for all treatments. Only some of these coefficients are reported due to space constraints. Panel A shows the coefficients on age 30–49 and 50–69, middle and high income, and being Republican. Omitted categories are age 18–29, low income, and being Democrat. Panel B reports the treatment effects of the video courses relative to the control group that saw no video. Panel C reports the mean of the dependent variables for respondents who saw the generic question formulation and no video (“Control mean”), and separately for male respondents (“Male control mean”) and Democrats (“Democrat control mean”). Standard errors are in parentheses. ∗ p <.1, ∗∗ p <.05, ∗∗∗ p <.01. THE QUARTERLY JOURNAL OF ECONOMICS TABLE III CONTINUED Downloaded from https://academic.oup.com/qje/article/136/4/2309/6363701 by guest on 18 January 2022 UNDERSTANDING TAX POLICY 2337 Downloaded from https://academic.oup.com/qje/article/136/4/2309/6363701 by guest on 18 January 2022 The dependent variables in Table III are indicator variables equal to 1 if respondents say that an increase in income taxes will change the listed behavior by a “moderate amount” to a “great deal.” What do we actually know about these effects of taxes in reality? The empirical evidence is not perfect, and economists disagree on the magnitudes and importance of these margins of adjustments to taxes. The goal is thus not to pinpoint quantitatively people’s mistakes, but to see how they reason about various groups’ tax adjustments and use information on their perceived prevalence and ranking of margins of adjustment that can be compared to the empirical evidence to date. Overall, respondents think that the margins along which people respond most strongly to taxes are evasion, moving states, and entrepreneurship. It is interesting that these are relatively less-studied responses in the empirical literature and have been, for the latter two at least, only more recently incorporated into the core tax theory models. Work on mobility responses to taxation (Kleven, Landais, and Saez 2013; Kleven et al. 2014, 2020, among others) shows them to be significant, but not typically large for the general population. They are, however, more substantial for high-income earners and for high-skilled professions involving only little location-specific human capital (Moretti and Wilson 2014, 2017; Akcigit, Baslandze, and Stantcheva 2016). On entrepreneurship, there is evidence that tax progressivity matters (Cullen and Gordon 2006, 2007), as well as that outcomes from entrepreneurship (e.g., innovation) can be significantly affected by taxes (Akcigit and Stantcheva 2020; Akcigit et al. forthcoming). Labor supply responses on the intensive or extensive margins are typically the core ones in the optimal tax literature. Intensivemargin responses include working fewer hours or exerting less work effort. Extensive-margin responses are, for example, switching out of the labor force. The empirical literature has shown that intensive-margin elasticities are typically small, but that extensive-margin elasticities at lower income levels are higher (for good summaries see Eissa and Liebman 1996; Chetty, Friedman, and Saez 2013; Hoynes 2019; Hoynes and Rothstein 2019). Respondents do consider working less to be a consequence of higher taxes, but a less prevalent one: 48% of respondents believe that high earners will work less and 39% believe middle-class earners will work less. Their perceived likelihood of people earning higher incomes or the middle class stopping work altogether are even lower. Finally, around 43% believe that high-income people are 2338 THE QUARTERLY JOURNAL OF ECONOMICS Downloaded from https://academic.oup.com/qje/article/136/4/2309/6363701 by guest on 18 January 2022 likely to have their spouse stop working in response to taxes and 32% believe so about a middle-class person. Interestingly, the empirical evidence shows that secondary earners (usually women) tend to be more elastic to taxes than primary earners (usually men), especially along the extensive margin (see Blau and Kahn 2007, 2017; Gelber and Mitchell 2012; Gelber 2014). In addition, respondents generally perceive behavioral responses to income taxes as stronger for higher-income earners than for middle-class earners, especially when it comes to the margins of evasion, having one’s spouse stop working, or moving states. Thus, 80% of respondents from the control group believe that high-income earners are likely to evade more taxes if their taxes were increased, 43% believe high earners will have their spouse stop working, and 78% think high-income earners are likely to move states in search of lower taxes. The corresponding perceptions for middle-class taxpayers are 60%, 32%, and 64%, respectively. Regarding earnings responses at the top, the evidence suggests that they are indeed a mix of effects that include avoidance and evasion (Saez, Slemrod, and Giertz 2012; Piketty, Saez, and Stantcheva 2014). Table IV, columns (1)–(3) show perceptions on the broader potential efficiency costs of income taxes. In column 1, the dependent variable is equal to one if respondents believe that taxes on higher incomes would hurt the economy. Only around 31% of respondents think they would. Columns 2 and 3 consider the share of respondents who think there is a Laffer effect from taxes on people with high incomes and the middle class respectively, i.e., whether reducing tax rates on these groups can, in fact, boost tax revenues and decrease the fiscal deficit. Quite a few respondents believe such Laffer effects exist, mostly when it comes to middleclass taxes (65% of respondents think a Laffer effect can occur for the middle class; 48% think so for high-income earners). There is no heterogeneity by income in these perceived efficiency effects. However, conditional on income and political affiliation, older respondents think that people respond less strongly to taxes. The share of Republicans who perceive strong behavioral responses to taxes along each dimension is consistently 30% to 50% higher than that of Democrats. The one exception is the perceived evasion of high-income earners, which is slightly weaker among Republicans. Fifty-five percent and 46% of Republicans, respectively, believe that high-income and middle-class earners 2339 UNDERSTANDING TAX POLICY TABLE IV EFFICIENCY COSTS OF INCOME AND ESTATE TAXES Income tax Estate tax Laffer effect (5) Panel A: Personal characteristics Republican 0.35∗∗∗ (0.02) Female − 0.04∗∗ (0.02) Age 30–49 − 0.03 (0.02) Age 50–69 0.01 (0.03) Middle income 0.02 (0.02) High income 0.04∗ (0.02) 0.18∗∗∗ (0.02) 0.06∗∗∗ (0.02) 0.01 (0.03) 0.02 (0.03) − 0.03 (0.03) − 0.03 (0.03) 0.02 (0.02) 0.05∗∗∗ (0.02) 0.00 (0.03) 0.04 (0.03) − 0.00 (0.03) − 0.01 (0.02) 0.15∗∗∗ (0.02) − 0.03 (0.02) 0.04 (0.03) 0.03 (0.03) − 0.05∗ (0.03) − 0.07∗∗∗ (0.03) 0.16∗∗∗ (0.03) 0.05∗∗ (0.02) − 0.04 (0.03) 0.04 (0.03) − 0.00 (0.03) − 0.04 (0.03) Panel B: Video treatment effects Redistribution − 0.01 (0.03) Efficiency 0.14∗∗∗ (0.03) Economist 0.06∗∗∗ (0.02) 0.00 (0.03) 0.03 (0.03) − 0.03 (0.03) − 0.05 (0.03) 0.01 (0.03) 0.00 (0.03) − 0.01 (0.04) 0.05 (0.04) 0.07∗∗ (0.03) 0.00 (0.04) 0.05 (0.04) − 0.00 (0.03) 0.48 0.46 0.39 2,780 0.65 0.63 0.61 2,781 Panel C: Descriptive statistics Control mean Male control mean Democrat control mean Observations 0.31 0.35 0.15 2,782 0.28 0.31 0.23 2,358 0.46 0.43 0.33 2,356 Note. The dependent variables are indicator variables equal to 1 in the following cases: ↑ Taxes on high incomes hurt economy: the respondent believes that increasing income taxes on high-income households would hurt economic activity; Laffer effect high incomes/middle class: the respondent believes that tax cuts on high-income households or on the middle class would decrease the deficit in the long run; ↑ Estate tax hurt economy: the respondent believes that increasing the federal estate tax on wealthy households would hurt economic activity; Laffer effect: the respondent believes that cuts to the estate tax for wealthy households would decrease the deficit in the long run. See the notes to Table III. Standard errors are in parentheses. ∗ p <.1, ∗∗ p <.05, ∗∗∗ p <.01. will work less in response to increased taxes, compared to 45% and 34% of Democrats. Another large difference is on the entrepreneurship margin: 63% of Republicans believe there will be less entrepreneurship among high incomes (as compared with 41% of Democrats) and 54% believe the middle class will be less entrepreneurial (relative to 39% of Democrats). In Table IV, the partisan gap grows even larger. Many more Republicans (52%) than Democrats (15%) perceive negative effects on the economy Downloaded from https://academic.oup.com/qje/article/136/4/2309/6363701 by guest on 18 January 2022 ↑ Taxes on Laffer ↑ Estate tax high incomes effect high Laffer effect hurt hurt economy incomes middle class economy (1) (2) (3) (4) 2340 THE QUARTERLY JOURNAL OF ECONOMICS 2. Estate Tax. Table V shows the perceptions of behavioral responses to the estate tax. Respondents are asked about the effects on wealthy individuals, as well as on currently young people. Questions about the latter group serve to elicit people’s perceptions on anticipation effects, as the currently young have time to plan their labor supply, savings, and other decisions in response to 18. The latter could be perceived as particularly strong for the middle class, given evidence for lower marginal propensities to consume for high-income households. Downloaded from https://academic.oup.com/qje/article/136/4/2309/6363701 by guest on 18 January 2022 from taxing high-income earners. Accordingly, Republicans also think there are more powerful Laffer effects for high-income earners. Yet the two political groups are not significantly different when it comes to Laffer effects for the middle class: 61% of Democrats and 70% of Republicans believe that tax cuts on the middle class will pay for themselves. There is thus bipartisan consensus, somewhat ironically, on a phenomenon whose existence has not yet been convincingly established in the literature. Taken together, these results suggest that respondents do believe there are some behavioral changes induced by taxes, but overall, less than half of all respondents perceive any of these responses to be prevalent, except for evasion and cross-state mobility. They correctly consider some margins to be weaker (e.g., the intensive-margin labor supply), but also overestimate the prevalence of others (e.g., moving across states). Respondents believe that higher-income households are more elastic than middle-class ones, and a sizable share believes that cutting taxes on high incomes will “pay for itself.” However, this does not translate into a generalized belief that taxes on high incomes will hurt the economy. This may be because respondents perceive rich people as being a smaller group, whose changes in behavior in the face of tax increases—even if strong and even if they can lead to lower revenue—are not enough to shape economic performance overall. Consistent with this, the share of respondents who think that middle-class tax cuts will pay for themselves is higher than the share who thinks so about high-income tax cuts, despite people expressing the belief that behavioral changes of each particular middle class are less pronounced than those of top earners. An additional reason for this could be that respondents anticipate other channels through which middle-class tax cuts can boost revenues, such as demand-side fiscal stimulus multiplier effects.18 0.01 (0.02) − 0.02 (0.02) − 0.01 (0.03) − 0.00 (0.03) − 0.01 (0.02) 0.00 (0.02) − 0.04 (0.03) − 0.00 (0.03) − 0.03 (0.03) 0.78 0.74 0.76 2,356 Panel A: Personal Characteristics Republican − 0.01 (0.02) Female − 0.03∗ (0.02) Age 30–49 0.00 (0.02) Age 50–69 − 0.02 (0.02) Middle income − 0.01 (0.02) High income 0.00 (0.02) Panel B: Video treatment effects Redistribution 0.05∗ (0.03) Efficiency 0.04∗ (0.03) Economist 0.03 (0.02) Panel C: Descriptive statistics Control mean 0.88 Male control mean 0.88 Democrat control mean 0.89 Observations 2,357 0.50 0.52 0.51 2,356 0.06 (0.04) 0.26∗∗∗ (0.04) 0.27∗∗∗ (0.03) 0.06∗∗ (0.02) − 0.03∗ (0.02) − 0.03 (0.03) − 0.11∗∗∗ (0.03) − 0.03 (0.03) − 0.03 (0.03) Wealthy (3) 0.53 0.51 0.53 2,356 0.01 (0.04) 0.04 (0.04) 0.07∗∗ (0.03) 0.08∗∗∗ (0.03) − 0.02 (0.02) − 0.05 (0.03) − 0.14∗∗∗ (0.03) − 0.06∗∗ (0.03) − 0.06∗∗ (0.03) Young (4) Work less 0.39 0.44 0.43 2,357 0.08∗∗ (0.04) 0.08∗∗ (0.04) 0.13∗∗∗ (0.03) 0.00 (0.03) − 0.07∗∗∗ (0.02) − 0.03 (0.03) − 0.07∗∗ (0.03) − 0.06∗∗ (0.03) − 0.05∗ (0.03) Wealthy (5) 0.37 0.39 0.34 2,355 0.02 (0.04) 0.03 (0.04) 0.07∗∗ (0.03) 0.03 (0.03) − 0.04∗∗ (0.02) − 0.04 (0.03) − 0.14∗∗∗ (0.03) − 0.04∗ (0.03) − 0.01 (0.03) Young (6) Stop working 0.09∗∗∗ (0.03) − 0.03 (0.02) − 0.03 (0.03) − 0.11∗∗∗ (0.03) − 0.06∗∗ (0.03) − 0.04∗ (0.03) Young (8) 0.57 0.58 0.57 2,355 0.46 0.47 0.41 2,355 Young (10) 0.83 0.85 0.80 2,356 0.73 0.74 0.68 2,357 − 0.01 (0.04) − 0.04 (0.04) − 0.02 (0.03) 0.04∗∗ 0.07∗∗∗ (0.02) (0.02) − 0.07∗∗∗ − 0.00 (0.02) (0.02) − 0.02 − 0.05∗ (0.02) (0.03) − 0.10∗∗∗ − 0.09∗∗∗ (0.03) (0.03) − 0.01 0.00 (0.02) (0.03) 0.04 0.03 (0.02) (0.03) Wealthy (9) Move state 0.01 − 0.01 − 0.01 (0.04) (0.04) (0.03) 0.06∗ 0.03 − 0.05∗ (0.04) (0.04) (0.03) 0.10∗∗∗ 0.07∗∗ − 0.02 (0.03) (0.03) (0.03) 0.03 (0.03) − 0.06∗∗∗ (0.02) − 0.00 (0.03) − 0.09∗∗∗ (0.03) − 0.04 (0.03) − 0.04 (0.03) Wealthy (7) Spouse stop working 0.50 0.53 0.48 2,356 0.11∗∗∗ (0.04) 0.24∗∗∗ (0.04) 0.23∗∗∗ (0.03) 0.06∗∗ (0.03) − 0.03 (0.02) − 0.04 (0.03) − 0.15∗∗∗ (0.03) − 0.01 (0.03) − 0.01 (0.03) Wealthy (11) 0.52 0.51 0.41 2,356 0.03 (0.04) 0.09∗∗ (0.04) 0.07∗∗ (0.03) 0.14∗∗∗ (0.03) − 0.01 (0.02) − 0.09∗∗∗ (0.03) − 0.15∗∗∗ (0.03) 0.00 (0.03) − 0.02 (0.03) Young (12) Be less entrepreneurial 0.07∗∗∗ (0.03) − 0.01 (0.02) − 0.07∗∗ (0.03) − 0.07∗∗ (0.03) − 0.01 (0.03) 0.01 (0.03) 0.05∗ (0.02) − 0.02 (0.02) − 0.01 (0.03) − 0.05 (0.03) − 0.01 (0.03) − 0.00 (0.03) 0.59 0.60 0.59 2,356 0.61 0.59 0.56 2,356 0.02 − 0.03 (0.04) (0.04) 0.22∗∗∗ 0.04 (0.03) (0.04) ∗∗∗ 0.20 0.06∗ (0.03) (0.03) Young (14) Save less Wealthy (13) Downloaded from https://academic.oup.com/qje/article/136/4/2309/6363701 by guest on 18 January 2022 Note. The dependent variables are indicator variables equal to 1 if the extent to which a respondent thinks that an increase in the federal estate tax would encourage the very wealthy individuals or the young (and not necessarily high-income) people toward the behaviors listed ranges from “a moderate amount” to a “great deal.” See the notes to Table III. Standard errors are in parentheses. ∗ p <.1, ∗∗ p <.05, ∗∗∗ p <.01. Young (2) Wealthy (1) Evade taxes TABLE V PERCEIVED BEHAVIORAL RESPONSES TO THE ESTATE TAX UNDERSTANDING TAX POLICY 2341 2342 THE QUARTERLY JOURNAL OF ECONOMICS V.B. Perceived Distributional Effects of Taxes The top panel of Figure III summarizes the perceived distributional gains and losses from either cutting the income tax on 19. The reasoning could be as for the income tax: although respondents believe that the wealthy respond to the estate tax and that their responses may be strong enough to generate Laffer effects, they do not think that the wealthy are sufficiently numerous or that their activity change is sufficiently influential to hurt overall economic performance. Downloaded from https://academic.oup.com/qje/article/136/4/2309/6363701 by guest on 18 January 2022 the estate tax. The strongest perceived responses to an estate tax increase overall are, again, evasion and moving states, followed closely by saving less, being less entrepreneurial, and having one’s spouse stop working. Respondents consider the currently wealthy on average more likely than the young to evade (a striking 88% of respondents believe so for the wealthy and 78% believe so for the young), more likely to have their spouse stop working (57% versus 46%), and much more likely to move state (83% versus 73%). On the other hand, they perceive the currently young as only slightly more likely to work less, be less entrepreneurial, or save less. Consistent with the findings on the income tax, Table IV, column (4) highlights that only 28% of respondents think that increasing the estate tax on wealthy households would hurt economic activity. Column (5) shows that 46% of respondents believe that decreasing the estate tax could improve tax revenues. As for the income tax, these findings suggest that while many respondents think that wealthy households react to the estate tax and around half of all respondents think that decreasing the estate tax on wealthy households could generate more revenues, less than one-third believe that increasing it could hurt the economy overall.19 In the third panel of Table V, Republicans and Democrats are aligned on their perceived (large) behavioral responses for the wealthy. Republicans, however, perceive youth responses to be on average stronger along the margins of working less, having their spouse stop working, and being less entrepreneurial. In Table IV, we can see that Republicans are more likely to believe that a higher estate tax would hurt the economy (37% compared to 23% of Democrats) and that there are Laffer effects from decreasing the estate tax (56% of Republicans versus 33% of Democrats). UNDERSTANDING TAX POLICY 2343 (A) Share of respondents that believe these groups would win from the following income tax changes: FIGURE III Distributional Effects of Income and Estate Taxes The figure shows respondents’ perceived distributional effects of tax reforms. In Panel A, the top part “Tax cut for high incomes,” shows respondents’ perceived effects from a tax cut on high incomes. The first row depicts the share of all respondents who think that the groups indicated in the legend (e.g., “Poor households,” “Upper Class”) would win if taxes on high incomes were cut. The following two rows show the shares among Democrats and Republicans who think these groups would benefit from a high-income tax cut. The middle set of rows “Overall tax increase and more government revenue,” shows the shares of respondents who think that each group of people would win if overall taxes were increased and more revenue was generated; again the shares are shown for respondents overall, as well as among Democrats and Republicans separately. The bottom part, “Trickle down,” shows the share of Republicans and the share of Democrats who think that lowering taxes on wealthy people and corporations would ultimately do more to reduce the income differences between poor and rich families than raising them. Panel B shows the share of respondents that think the groups indicated in the legend would gain if the estate tax were cut. Only respondents who saw no video treatment are included. Online Appendix Tables OA-21 and OA-23 show detailed regression results for these variables. Downloaded from https://academic.oup.com/qje/article/136/4/2309/6363701 by guest on 18 January 2022 (B) Share of respondents that believe these groups would win if the estate tax were cut: 2344 THE QUARTERLY JOURNAL OF ECONOMICS V.C. Social Preferences and Social Fairness Concerns This section explores some of the factors that determine respondents’ social preferences as captured by the social marginal welfare weights in Section II and dives into their social fairness concerns related to income and estate taxation. Perceptions of inequality, which could also affect social preferences for taxation, were already shown in Table II. 20. Online Appendix Tables OA-21 and OA-23 follow the usual more detailed format. Downloaded from https://academic.oup.com/qje/article/136/4/2309/6363701 by guest on 18 January 2022 high earners (top part) or increasing overall taxes (middle part).20 The top two parts of the figure represent the share of respondents who think that each of the following groups will mostly gain from these tax changes: poor, working-class, middle-class, uppermiddle-class, and upper-class households. On average, 32% of respondents believe that lower-class households will benefit from a tax cut on high earners, and 82% believe so about the upper class. Sixty-five percent believe that the lower class will benefit from overall tax increases and this share decreases monotonically to 41% for the upper class. The final row considers whether respondents believe in trickle-down economics, that is, whether lowering income taxes will on balance do more to reduce income inequality than increasing them. Overall, only 32% of respondents believe in the existence of trickle-down effects, consistent with the share who believes that lower-class households will gain if taxes on high earners are reduced. The bottom panel of Figure III shows the perceived distributional gains and losses from cutting the estate tax. Seventy-five percent of respondents believe the upper class would benefit. This share decreases to 42% for the lower class. There are some partisan differences in the perceived distributional effects of tax changes. Republicans are more likely than Democrats to think that all groups below the upper-middle class will benefit from tax cuts on high earners but agree with Democrats that the upper-middle and upper classes will realize the largest gains. Republicans are less likely than Democrats to believe that anyone would gain from an overall tax increase that raises extra revenues to be used by the government. Finally, Republicans are much more ardent believers in trickle-down effects from income taxes: 60% of Republicans compared to only 10% of Democrats believe in them. UNDERSTANDING TAX POLICY 2345 TABLE VI SOCIAL PREFERENCES AND FAIRNESS CONSIDERATIONS ABOUT THE INCOME TAX Panel A: Personal characteristics Republican − 0.42∗∗∗ (0.02) Female 0.04∗∗ (0.02) Age 30–49 0.01 (0.02) Age 50–69 0.00 (0.02) Middle income − 0.03 (0.02) High income − 0.04∗∗ (0.02) − 0.38∗∗∗ (0.02) − 0.00 (0.02) 0.05∗ (0.03) 0.01 (0.03) − 0.06∗∗ (0.02) − 0.06∗∗ (0.02) Panel B: Video treatment effects Redistribution 0.05 (0.03) Efficiency 0.03 (0.03) Economist 0.02 (0.02) 0.10∗∗∗ (0.03) 0.02 (0.03) 0.06∗∗ (0.03) Panel C: Descriptive statistics Control mean Male control mean Democrat control mean Observations 0.48 0.49 0.69 2,781 0.70 0.68 0.92 2,781 People rich due to luck (3) High income entitled to keep their income (4) − 0.34∗∗∗ (0.02) 0.04∗∗ (0.02) 0.02 (0.03) 0.04 (0.03) − 0.03 (0.02) − 0.09∗∗∗ (0.02) 0.36∗∗∗ (0.02) − 0.02 (0.02) − 0.02 (0.02) − 0.05∗∗ (0.03) 0.02 (0.02) 0.05∗∗ (0.02) − 0.01 (0.03) 0.03 (0.03) 0.05∗ (0.02) − 0.01 (0.03) 0.01 (0.03) 0.00 (0.02) 0.60 0.59 0.78 2,780 0.30 0.32 0.10 2,780 Note. The dependent variables are indicator variables equal to 1 in the following cases: Wealth distribution unfair: the respondent thinks that money and wealth in the United States should be more evenly distributed; Inequality serious issue: the respondent believes that income inequality is a serious or very serious issue; People rich due to luck: the respondent believes that a person is rich because they had more advantages than others, rather than because they worked harder than others; High income entitled to keep their income: the respondent believes that high-income individuals are entitled to keep a very large share of their income and should not have to pay high taxes, even if that means less government revenue is available to help low-income families make ends meet. See the notes to Table III. Standard errors are in parentheses. ∗ p <.1, ∗∗ p <.05, ∗∗∗ p <.01. 1. Income Tax. Table VI shows that 70% of the respondents think that “money and wealth in the U.S. should be more evenly distributed,” and almost half of all respondents think that income inequality is a serious or very serious issue. Column (3) indicates that 60% of respondents believe that someone is rich mainly due to luck rather than effort. Consistent with these considerations, just Downloaded from https://academic.oup.com/qje/article/136/4/2309/6363701 by guest on 18 January 2022 Wealth Inequality distribution serious unfair issue (1) (2) 2346 THE QUARTERLY JOURNAL OF ECONOMICS 2. Estate Tax. Recall from Section II that normative considerations on the estate tax are complex and depend on whether one takes the point of view of the parents or that of the children. How do respondents view these thorny normative issues? Table VII, columns (1) and (2) show that 64% of the respondents believe that wealth should be more evenly distributed, and 46% deem wealth inequality to be a serious problem. In column (3), we can see that, very consistent with the income tax results, 62% of respondents believe that someone is wealthy mainly due to luck rather than due to effort. The table also shows more nuanced views on estate taxation. A first group of questions prompts respondents to take the point of view of parents (columns (4) and (5)). Sixty-one percent of respondents believe it is unfair to tax the estate of wealthy parents who worked hard to accumulate their wealth, and 47% of respondents believe it is unfair to do so if wealthy parents themselves had inherited their wealth in the first place. A second group of Downloaded from https://academic.oup.com/qje/article/136/4/2309/6363701 by guest on 18 January 2022 30% of respondents agree with the statement that high-income individuals are entitled to keep a very large share of their income and should not have to pay high taxes even if that means less government revenues. Partisan gaps are much larger on these social fairness issues than on efficiency or distributional considerations. Ninety-two percent of Democrats and 42% of Republicans believe that wealth and money should be more evenly distributed in the United States. Sixty-nine percent of Democrats and 25% of Republicans perceive inequality to be a serious or very serious issue. Seventy-eight percent of Democrats as opposed to 41% of Republicans believe that a person is rich because they had more advantages than others. Fifty-five percent of Republicans and 10% of Democrats believe that high-income earners are entitled to keep their income. There appears to be some level of self-interest, with respondents that are in the higher-income group less likely to believe that the wealth distribution is unfair, that inequality is a serious issue, or that people are rich due to luck. They are also more likely to believe that people earning high incomes are entitled to keep their income. Nevertheless, the gap between high- and lower-income respondents is not that large. Thinking back to the discussion in Section II, respondents clearly also have other concerns than pure self-interest when it comes to social fairness considerations about taxes. TABLE VII − 0.45∗∗∗ (0.02) − 0.01 (0.02) 0.00 (0.03) − 0.05 (0.03) − 0.01 (0.03) 0.00 (0.03) − 0.26∗∗∗ (0.02) 0.07∗∗∗ (0.02) 0.02 (0.03) − 0.00 (0.03) − 0.01 (0.03) − 0.07∗∗∗ (0.03) Wealthy heirs (5) 0.25∗∗∗ 0.25∗∗∗ (0.02) (0.03) 0.04∗ 0.03 (0.02) (0.02) − 0.01 − 0.01 (0.03) (0.03) 0.01 0.04 (0.03) (0.03) 0.01 − 0.02 (0.03) (0.03) 0.02 0.01 (0.03) (0.03) Hard workers (4) 0.20∗∗∗ (0.02) − 0.09∗∗∗ (0.02) 0.02 (0.03) 0.01 (0.03) 0.03 (0.03) 0.06∗∗ (0.03) Access better amenities (6) 0.24∗∗∗ (0.03) − 0.03 (0.02) 0.07∗∗ (0.03) 0.11∗∗∗ (0.03) 0.01 (0.03) 0.04 (0.03) Inherit more (7) Fair that children from wealthy families: Unfair tax estates of: Trade-off: 0.27∗∗∗ (0.03) − 0.01 (0.02) 0.02 (0.03) 0.03 (0.03) − 0.00 (0.03) 0.04 (0.03) Parents should pass on wealth even if unequal for children (8) Downloaded from https://academic.oup.com/qje/article/136/4/2309/6363701 by guest on 18 January 2022 Panel A: Personal characteristics Republican − 0.39∗∗∗ (0.02) Female 0.06∗∗∗ (0.02) Age 30–49 0.01 (0.03) Age 50–69 − 0.01 (0.03) Middle income 0.01 (0.02) High income − 0.02 (0.02) Person Wealth dis- Inequality wealthy serious tribution due to luck issue unfair (1) (2) (3) Children’s side: Parents’ side: SOCIAL PREFERENCES AND FAIRNESS CONSIDERATIONS ABOUT THE ESTATE TAX UNDERSTANDING TAX POLICY 2347 0.01 (0.04) 0.07∗ (0.04) − 0.00 (0.03) 0.62 0.58 0.73 2,358 0.02 (0.04) − 0.02 (0.04) 0.01 (0.03) 0.46 0.50 0.64 2,358 0.61 0.60 0.51 2,357 0.01 (0.04) 0.03 (0.04) − 0.02 (0.03) Hard workers (4) 0.47 0.45 0.38 2,358 − 0.05 (0.04) 0.03 (0.04) − 0.03 (0.03) Wealthy heirs (5) 0.32 0.36 0.19 2,357 0.03 (0.04) 0.01 (0.04) 0.03 (0.03) Access better amenities (6) 0.53 0.53 0.36 2,357 − 0.09∗∗ (0.04) − 0.03 (0.04) − 0.05 (0.03) Inherit more (7) Fair that children from wealthy families: Unfair tax estates of: Trade-off: 0.58 0.55 0.49 2,356 − 0.06∗ (0.04) 0.02 (0.04) − 0.08∗∗ (0.03) Parents should pass on wealth even if unequal for children (8) Downloaded from https://academic.oup.com/qje/article/136/4/2309/6363701 by guest on 18 January 2022 Note. The dependent variables are indicator variables equal to 1 in the following cases: Wealth distribution unfair: the respondent believes that money and wealth in this country should be more evenly distributed among a larger percentage of the population; Inequality serious issue: the respondent believes that wealth inequality in the United States is a serious or very serious problem; Person wealthy due to luck: the respondent thinks that a person is wealthy because they had more advantages than others, rather than because they worked harder than others; Unfair tax estates of hard workers: the respondent believes that it is somewhat unfair, unfair, or very unfair to tax the estate of wealthy people who have worked hard and saved a lot in order to pass on wealth to their children; Unfair tax estates of wealthy heirs: the respondent believes that it is somewhat unfair or very unfair to tax the estate of people who are wealthy because they have inherited a lot from their parents; Fair that children from wealthy families access better amenities: the respondent believes that it is somewhat fair or very fair that children born in very wealthy families have access to better amenities; Fair that children from wealthy families inherit more: the respondent believes that it is somewhat fair or very fair that children born in very wealthy families inherit much more than children born in less wealthy families; Parents should pass on wealth even if unequal for children: the respondent believes that wealthy parents should be able to pass on all of her wealth to her children; as a result, some children will start their own life with much larger wealth just by virtue of being born in a richer family. See the notes to Table III. Standard errors are in parentheses. ∗ p <.1, ∗∗ p <.05, ∗∗∗ p <.01. Panel B: Video treatment effects Redistribution 0.04 (0.03) Efficiency − 0.06∗ (0.03) Economist 0.02 (0.03) Panel C: Descriptive statistics Control mean 0.64 Male control mean 0.62 Democrat control mean 0.85 Observations 2,358 Person Wealth dis- Inequality wealthy serious tribution due to luck issue unfair (1) (2) (3) Children’s side: Parents’ side: TABLE VII CONTINUED 2348 THE QUARTERLY JOURNAL OF ECONOMICS UNDERSTANDING TAX POLICY 2349 21. The timing of the actual wealth transmission depends on the age of parents and children. The way the questions are framed is vague on the actual timing but does unambiguously imply that parental wealth shapes opportunities for children from early on, even before wealth is formally transmitted. The wording chosen implies that at least part of parental wealth is “available” to their children in some form or another throughout their life. This can take the form of direct spending on the children early on, large gifts, and actual wealth transmission later