Do Americans want to tax wealth? Evidence from online surveys

https://doi.org/10.1016/j.jpubeco.2020.104207Get rights and content

Abstract

A vast theoretical literature in public finance has studied the desirability of capital taxation. This discussion largely ignores the political feasibility of taxing wealth. We provide, to our knowledge, the first investigation of individuals' preferences over jointly taxing income and wealth. We provide subjects with a set of hypothetical individuals' incomes and wealth and elicit subjects' preferred (absolute) tax bill for each individual. Our method allows us to unobtrusively map both income earned and accumulated wealth into desired tax levels. Our regression results yield roughly linear desired tax rates on income of about 14%. Respondents' suggested tax rates indicate positive desired wealth taxation. When we distinguish between sources of wealth we find that, in line with recent theoretical arguments, subjects' implied tax rate on wealth is 3% when the source of wealth is inheritance, far higher than the 0.8% rate when wealth is from savings. Textual analysis of respondents' justifications for their tax rates imply limited concern for the elasticity of tax bases with respect to net-of-tax rates.

Introduction

Wealth taxes are levied on the stock of private assets such as real estate, cash holdings, and financial assets (e.g., stocks and bonds). While the idea of taxing assets is not a new one, the rise of wealth-to-income ratios around the world (Piketty, 2014) and increased wealth inequality in the United States (Saez and Zucman (2016); Smith et al., 2019) has led to increased discussion of wealth taxation among both academics and policymakers. Recent academic research, in particular, has focused on wealth tax experiences in other countries, using identifying variation in wealth tax rates and bases to quantify the behavioral responses to wealth taxes in countries such as Denmark (Jakobsen et al., 2018), Sweden (Seim, 2017), Colombia (Londoño-Vélez and Ávila, 2018), and Switzerland (Brülhart et al., 2017). In a widely debated proposal, Saez and Zucman (2019) summarize the recent evidence in advocating for the creation of a progressive wealth tax in the U.S., which, as they observe, has been proposed by prominent candidates for the 2020 Democratic presidential nomination.1

The questions of incidence and implementation of a wealth tax are distinct from questions of political feasibility, i.e., whether the practical political economy of tax setting would allow for wealth taxation. Putting aside legal impediments, behavioral responses, and the practical challenges of implementation, there is the separate issue of whether a wealth tax is even desired by the electorate. And if it is, what are the wealth tax parameters that a responsive legislator would aim to translate into policy? Do citizens understand the difference between taxing stocks and taxing flows? And if they do, do they consider stocks of wealth acquired from saving as normatively different from those acquired via inheritance?2

We provide, to our knowledge, the first investigation into individuals' preferences toward jointly taxing (net) wealth and income, via surveys on Amazon's Mechanical Turk (MTurk) and by adding questions to the (more representative) Understanding America Study administered by the University of Southern California.3 Each subject in our study is confronted with scenarios describing a hypothetical individual's income and wealth. For each scenario, the subject then chooses a tax bill for the individual to pay. By asking for absolute levels of taxation in response to a hypothetical individual's (multi-dimensional, in our case two-dimensional) financial situation, we believe our approach is less likely to lead subjects to use misplaced heuristics (for example, to choose current levels or to confuse marginal and average rates). We also argue that, by asking for desired absolute tax levels rather than rates on income and wealth per se, our methodology is unobtrusive—the implied T(income, wealth) function, which we trace out from individual responses, may be a complicated nonlinear function that would be much more costly to elicit, for example, by asking for separate tax rates on a large set of income and wealth brackets (at the same time). Finally, past work has shown that this methodology produces lower and less progressive tax schedules than asking for preferences as rates, which means it is a demanding test of the claim that Americans prefer a non-zero tax on wealth.4

Our empirical findings are as follows: First, subjects' tax recommendations over income versus wealth are, roughly speaking, “sensible.” Respondents seem to understand intuitively the difference between a stock and a flow and choose implied wealth tax rates that are typically an order of magnitude smaller than those on income. Second, their chosen tax bills imply a linear tax rate on income of approximately 13–15%, in line with past work, another sign that our respondents appear to be roughly representative in their views and to have taken the task seriously.

Third, and of greater interest, subjects' choices imply positive rates of wealth taxation. When we restrict the relationship of the tax bill and wealth to be linear, the implied average tax rate on wealth is about 1.2% in our baseline estimate. In follow-up sessions, we tell subjects the source of the hypothetical wealth. In one treatment they are told it is from saving past income, while in another treatment they are told it is from a bequest from a deceased relative. Preferred taxes on wealth from savings are 0.8%, versus over 3% on wealth from inheritance.

Finally, we examine the reasoning provided by our subjects when asked to justify (in their own words) their chosen tax levels, and in particular how their explanations compare to concerns raised in the optimal tax literature. In Mirrlees (1971), the optimal income tax rate is inversely proportional to the elasticity of the tax base with respect to the net-of-tax rate. Similarly, the classic contributions of Atkinson and Stiglitz (1976), Judd (1985), and Chamley (1986) argue that the tax on capital income should be zero because of the costs resulting from behavioral responses of taxed wealth holders.5 Our subjects do not, however, express concerns for such behavioral responses (e.g., that higher labor income taxes would discourage work, or that higher taxes on wealth would reduce savings or induce capital flight). Simplicity of the tax schedule (e.g., a flat tax) is attractive to many. Also, “double taxation” is often noted as an objection to taxing wealth, with respondents saying it was “already taxed” at the time it was earned. These considerations are quite removed from the tradeoffs that economists weigh in the classic optimal tax framework.

We view our main contribution as two-fold. First, to the best of our knowledge we are among the first to directly elicit preferences for wealth taxation from prospective voters.6 While there are no immediate payoff consequences for survey respondents, the sensible estimates we obtain on income taxation suggest that subjects exert effort in providing responses. While objections to wealth taxation on theoretical grounds or owing to legal or logistical impediments (which we discuss in the conclusion) may still stand, our findings indicate that there appears to be support among the electorate for such policies.

The credibility of our estimates on desired wealth taxation is bolstered by our methodology, which we view as our second contribution. Since we elicit subjects' preferred tax rates through their (absolute) tax choices over a number of hypothetical income/wealth pairs we avoid, for example, leading subjects to gravitate toward responses that reflect current tax rules. With sufficient data, this methodology could be extended across many tax-relevant characteristics (for example, consumption, real estate holdings, and age) to elicit the full tax schedule preferred by respondents. The disadvantage, as we have noted above, is that respondents are typically unaccustomed to thinking in terms of absolute tax bills, and based on past work our methodology will tend to give lower and less progressive rates than when choices are framed as percentages. While bias in any direction is not ideal, we note that this bias pushes against finding our key result: that Americans prefer a positive tax on wealth.

The remainder of the paper proceeds as follows. Section 2 outlines our experimental design. Section 3 describes our data collection procedures and provides summary statistics on our resulting sample of subjects. Section 4 describes results from the baseline experiment, in which we do not specify to subjects the source of the wealth values they are asked to consider, and then Section 5 shares results from the surveys that compare responses for wealth accumulated via saving past income versus wealth gained via a bequest. Section C uses our results, past estimates of relevant elasticities, and recent models of optimal capital taxation to calculate the implied social welfare weights our subjects' place on individuals with varying levels of wealth. Section 6 concludes and offers suggestions for future work.

Section snippets

Experimental design

We developed our survey experiment with two main goals in mind. First, we wanted to be as unobtrusive as possible, allowing subjects to consider both income and wealth levels when choosing their desired tax but not asking them explicitly how much they wanted to tax income versus wealth. We worried that asking for specific rates on income and wealth would prime them, perhaps toward submitting the current tax rate on income or, more worrisome in our context, presuming that there should be a

Data

For the most part, we recruited and compensated our subjects through Amazon's Mechanical Turk (MTurk) market place, but redirected them to surveys that we built with Qualtrics' online survey software. The exception is a round of data collection (N=306) using the Understanding America Study (UAS) run by the Center for Economic and Social Research at the University of Southern California. This platform is more representative of the U.S. adult population, but also substantially more expensive. The

Baseline results when the source of wealth is not specified

We begin with an analysis of preferred tax schedules using the surveys in which the source of the hypothetical individuals' wealth was unspecified. These results will serve as a baseline for examining how preferred tax rates are affected by the source of wealth.

Results when the source of wealth is specified

We now turn to data from surveys in which we specify the source of the hypothetical individual's wealth. We first analyze data collected in 2015. At the end of this section we turn to analyses of data collected in 2018 and 2019, when we tested variants of our original surveys, and replicated the survey experiment on a different, more representative survey vendor.

Before describing our 2015 “inheritance versus savings” results, we note that these data pass the same “quality checks” as the 2014

Discussion and conclusion

A recent literature documents the increasing importance of wealth and wealth inequality. Saez and Zucman (2016) find that 20% of American wealth is held by the top 0.1% of owners, a share that has doubled in the last forty years.24 Piketty (2014) documents a secular increase in

Declaration of competing interest

None of the authors have any conflicts to report.

References (41)

  • M. Brülhart et al.

    The elasticity of taxable wealth: evidence from Switzerland

  • C. Chamley

    Optimal taxation of capital income in general equilibrium with infinite lives

    Econometrica

    (1986)
  • J. Charité et al.

    Reference Points and Redistributive Preferences: Experimental Evidence. Tech. Rep

    (2015)
  • T.L. Cherry et al.

    Hardnose the dictator

    Am. Econ. Rev.

    (2002)
  • D. Dalton et al.

    Gender differences in ethics research: the importance of controlling for the social desirability response bias

    J. Bus. Ethics

    (2011)
  • S. Dennis et al.

    Virtual Private Servers and the Limitations of Ip-based Screening Procedures: Lessons From the Mturk Quality Crisis of 2018

    (2018)
  • P. Diamond et al.

    The case for a progressive tax: from basic research to policy recommendations

    J. Econ. Perspect.

    (2011)
  • R.L. Einhorn

    American Taxation, American Slavery

    (2008)
  • E. Farhi et al.

    Progressive estate taxation

    Q. J. Econ.

    (2010)
  • E. Fehr et al.

    Fairness and retaliation: the economics of reciprocity

    J. Econ. Perspect.

    (2000)
  • Cited by (36)

    • The positionality of goods and the positional concern's origin

      2024, Journal of Behavioral and Experimental Economics
    • Why do (some) ordinary Americans support tax cuts for the rich? Evidence from a randomised survey experiment

      2023, European Journal of Political Economy
      Citation Excerpt :

      Online and laboratory experiments have also been used to explore how redistributive preferences are affected by perceptions of fairness (Almås et al., 2020; Durante et al., 2014) and individuals’ position in the income distribution relative to important reference groups (e.g. the bottom ranking income group) (Fisman et al., 2020; Kuziemko et al., 2014). Lastly, there is a small but growing literature utilising survey experiments to look at preferences for wealth taxation (Bastani and Waldenström, 2021; Fisman et al., 2020). Our paper contributes to the existing literature in two main ways.

    • How morality and efficiency shape public support for minimum wages

      2023, Journal of Economic Behavior and Organization
      Citation Excerpt :

      In particular, we design and implement a survey-based randomized experiment with 2219 Americans to study how morality, employment consequences, wage levels, and equity considerations shape respondents’ attitudes toward minimum wages.3 Our work is inspired by, and contributes to, the literature in which surveys and experiments are used to elicit social preferences with respect to income distribution, redistribution, taxation, and kidney shortages (e.g., Kuziemko et al., 2015, Elias et al., 2019, Fisman et al., 2020, Stantcheva, 2020, and Fisman et al., 2021). Methodologically, we draw particularly strong inspiration from Elias et al. (2019) who use a similar experimental approach to estimate the willingness of the public to allow payments to kidney donors.

    View all citing articles on Scopus

    We thank Wojciech Kopczuk, Christopher Roth, Thomas Piketty, Emmanuel Saez, Stefanie Stantcheva, Matt Weinzierl, Divya Singh, Jon Zytnick, and participants at the American Economic Association 2016 meetings (and in particular discussant Charlotte Cavaillé) and the Journées Louis-André Gérard-Varet conference and seminar participants at Yale and Penn for helpful comments and conversations. Dana Scott provided invaluable research assistance.

    View full text