CHAPTER 1: WHY MIGHT GOVERNMENTS TAX THE RICH?
1.See for example “How to Fix Our Appalling Tax Code” by Dave Camp, Wall Street Journal, February 25, 2014.
2.See Dworkin (1977), p. 370 as well as Anderson (1999).
3.Theoretically, this insight can be used to justify many different types of tax systems, including those where all pay the same rate. Historically, however, ability to pay has most often been used to justify why the rich should pay a higher rate of tax than the rest of the population.
4.For contributions on the U.S. case see in particular Bank, Stark, and Thorndike (2008) and Brownlee (2004), as well as Alstott and Novick (2006) and Kennedy (1980). For the British case, see Daunton (2002, 1996) and Cronin (1991) in particular. See also Feldman and Slemrod (2009).
5.The twenty countries are Australia, Austria, Belgium, Canada, Denmark, Finland, France, Germany, Ireland, Italy, Japan, Korea, the Netherlands, New Zealand, Norway, Spain, Sweden, Switzerland, the United Kingdom, and the United States.
6.Although we think our conclusions are general, this does not mean that fairness debates do not work differently across different types of countries. Even in the sample, we will point out some important differences, for example between democratic and non-democratic countries, that are predicted by our arguments. We would expect further differences in the importance and dynamics of fairness considerations in tax debates by characteristics such as the ability of states to deliver whatever the public goods are for which the revenue is being raised. We will address this possibility for the countries in the sample in what follows, but these issues may generate even greater differences for the politics of taxation in countries outside the sample, especially in the developing world.
7.For several recent contributions to this literature, see Alesina and Glaeser (2004), Beramendi (2012), Beramendi and Anderson (2008), Hacker (2008), Hall and Soskice (2001), Hays (2009), Huber and Stephens (2001), Iversen and Soskice (2006, 2009), Kenworthy and Pontusson (2005), Lieberman (2003), Mares (2003), Moene and Wallerstein (2001), Pontusson (2005), Przeworski and Wallerstein (1988), Rehm (2011), Rodden (2010), Roemer (1998, 1999), Rueda (2007), Steinmo (1993), Swank (2002), and Timmons (2005).
8.It would be misleading to claim that we have a complete picture of the incidence of taxation on different social groups in medieval and early modern Europe. However, the consensus among historians of the period is that European wars, especially after 1300, were mainly fought with money raised by indirect taxes on common consumption goods. These would have fallen less heavily on the rich than on other social groups. Molho (1996), p. 104 makes this claim for Florence, and Blockmans (1987) provides compelling evidence for cities in the Low Countries. O’Brien and Hunt (1993) discuss how England relied increasingly on indirect taxation after becoming engaged in continental wars beginning in 1688. The autonomous city of Siena provided one exception to the general European pattern, as Bowsky (1969) shows that it relied extensively on direct taxation. The French monarchy also relied heavily on direct taxes, and in particular the taille. However, the overall incidence of taxation in France was extremely low by modern standards. In the year 1607, total taxes amounted to 31,437,671 livres tournois (Chaunu 1977). Using a population estimate of 18.5 million people (http://www.ggdc.net/maddison/maddison-project/data.htm), this would amount to 1.7 livres tournois per capita. This would be equivalent to about 3.4 days of wages for a building laborer in Paris at the time. (Data reported by the Global Price and Income History Group, original collection by Robert Allen, http://gpih.ucdavis.edu/Datafilelist.htm.)
9.Data on tax revenues as a percentage of gross domestic product are from Mauro et al. (2013).
10.This idea, introduced by Peacock and Wiseman (1961), has recently been explored extensively by Beetsma, Cukierman, and Giuliodori (2013).
11.Guicciardini (1994, p. 49).
12.In concluding his review of the history of progressive taxation in 1908, Edwin Seligman observed the following. “From the above review it is evident that the tendency toward progressive taxation is almost everywhere on the increase. Whether we deplore it or not, democracy is asserting itself more vigorously and it is precisely in the most democratic countries like Australia and Switzerland that the movement in favor of progressive taxation is the strongest.” Seligman (1908), p.124.
13.Shultz (1926, p. 282).
14.See in particular the contributions by Acemoglu and Robinson (2006, 2000) and by Boix (2003). Some have also proposed a partisan variant of the hypothesis; instead of a hypothetical median voter choosing policy, parties of the left (representing lower income groups) and parties of the right (representing upper-income groups) vie for power with different proposals. It then matters who ends up holding political power. See Przeworski and Sprague (1986) and Roemer (1997) for a formalization. The original political economy model of suffrage expansion and redistribution was produced by Allan Meltzer and Scott Richard in 1981, building on earlier insights by Romer (1975) and Roberts (1977). Meltzer and Richard considered an economy populated by individuals having different incomes who must choose the rate for a proportional tax on income that is used to fund a lump-sum redistributive transfer to each citizen. One way to think of this is as the simplest possible type of progressive tax policy—though the tax rate is the same for all individuals, the rich can end up paying a greater proportion of their income once the lump-sum transfer is taken into account. Under this scenario the preferences of voters are naturally ranked by income; the richer a voter, the lower the tax rate they prefer. Meltzer and Richard demonstrated that if one starts with a society in which the franchise is limited to the rich, then as the suffrage is expanded this would lead to higher taxation and more redistribution from rich to poor. Strictly speaking, the Meltzer-Richard model does not give us a general result about democracy and taxation of the rich. To get that we would ideally need to move to a model with a more complex (and more realistic) tax system where voters are choosing an income tax schedule, something like we see in the United States today where the federal income tax has multiple brackets. Or, to make the theoretical prediction truly general, we should allow voters to choose from a set of any possible tax schedules. The problem is that once we do this, we cannot necessarily rank preferences over taxation by income. Take a society with three groups: the rich, the poor, and the middle class. A middle-class person might prefer a tax system that soaks the rich, or might prefer one that soaks the poor. It is also possible to have a scenario where the poor and rich vote for an outcome that soaks the middle class. There have been various attempts to find a way out of this conundrum, but no definitive solution has yet been found. See Roemer (1999), Carbonell-Nicolau and Ok (2007), Carbonell-Nicolau (2009), and Iversen and Soskice (2006) on this point. In the end, most scholars have a strong intuition that democracies with universal suffrage should tax the rich more heavily, even if there is no general theoretical model that shows this will necessarily be the case.
15.See Aidt and Jensen (2009).
16.See the online appendix at http://press.princeton.edu/titles/10674.html. This archive also has all materials needed to replicate the analyses in this book.
17.Many scholars, particularly in political science, have cited the Meltzer and Richard (1981) study when making this argument, but this is a rather fragile foundation. First, as we have noted, the Meltzer-Richard model involves the choice of a single tax rate on all, combined with a redistributive transfer. So it doesn’t necessarily carry over to a context where separate tax rates are being chosen for separate groups. Second, even in the Meltzer-Richard context, concluding that greater inequality results in more redistribution depends on several additional assumptions that may or may not hold. It’s theoretically possible in this model to have a shift to a more unequal society be characterized by lower taxes on the rich.
18.See Nelson (2004) on this point.
19.Guicciardini (∼1520 [1867]).
20.Jean-Jacques Rousseau, Discourse on Political Economy, edited by Victor Gourevitch (1977, p. 19).
21.Emmanuel Farhi and Ivan Werning have constructed a theoretical model that suggests exactly how this might take place. If levels of wealth taxation are chosen sequentially by democratic vote, even wealthy voters may agree to be taxed today so as to avoid much more substantial redistribution in the future. See Farhi and Werning (2009). A further reason that governments might respond to greater inequality with higher taxes on the rich is that voters are simply inequality averse. See Lü and Scheve (2014) for one version of this argument. Although there is certainly laboratory and survey-experimental evidence consistent with the idea that people do not like too much inequality, it is not at all clear whether this motivation is sufficiently strong to lead to high taxes on the rich. The patterns of inequality and tax policy discussed in the text suggest that this motivation alone is not enough.
22.See Gilens (2012) and Bartels (2008).
23.See in particular Bonica, McCarty, Poole, and Rosenthal (2013).
24.For a good example of this view see Graetz and Shapiro (2005). Fairfield (2015) also presents evidence along these lines by studying the success of economic elites in preventing tax reform across issues and countries in Latin America.
25.See Bonica et al. (2013) for an insightful review of campaign finance, lobbying expenditures, taxation, and inequality in the United States today. See Birnbaum and Murray’s Showdown at Gucci Gulch (1988) for a discussion of lobbying in connection with the tax reform act of 1986.
26.See Acemoglu and Robinson (2008).
27.For a classic contribution that does attempt to answer this question, see Slemrod (2002). For more recent evidence see Giertz, Saez, and Slemrod (2012).
28.See Alt (1979) and Fisman, Jakiela, and Kariv (2014).
29.See, for example, the statement by Michael Boskin (2000) that Francis Edgeworth ignored the possibility that high taxes on the rich could give them an incentive to supply less labor.
30.Edgeworth (1897), p. 553.
31.Our emphasis on war, compensatory arguments, and taxation of the rich is closely related to arguments that highlight the link between warfare and the provision of political or civil rights. Mass mobilization for war has led to important changes in the status of women and of ethnic or racial minorities, based on the simple notion that if a group lacking full rights participated in the war effort, then they ought to be granted the same rights as everyone else. For a particularly stark example of this, see the study by Philip Klinkner and Rogers Smith (1999) on the status of African Americans during war and peace. The same argument has been made for the establishment of democracy more generally. See Ferejohn and Rosenbluth (forthcoming). As we discuss in chapter 7, in ancient Athens it was argued that if the lower classes rowed the boats to power the Athenian navy, then they should be granted the same political rights as everyone else. While we emphasize how mass warfare created opportunities for new fairness-based arguments, we do not pretend that the only effect of warfare has been via this fairness channel. When a population is mobilized to fight, this can mean that citizens find it easier to use violence and to mobilize after a war’s end. A recent and persuasive statement of this phenomenon is provided by Jha and Wilkinson (2012). This phenomenon has been important in a wide variety of historical contexts, but it seems unlikely that it is driving the effect of mass mobilization wars on tax policy. We show variation in the effect of war on taxing the rich across countries—e.g., a stronger effect in democracies for which the norm of treating citizens as equals is stronger—that cannot easily be explained by the fact that wars arm citizens. Another main effect of mass warfare independent of fairness is an improvement in state capacity. Throughout history warfare has arguably been associated with improvements in the capacity of states to tax. This is an argument most closely associated with Charles Tilly that has been studied more recently by Tim Besley and Torsten Persson. See Tilly (1975, 1990) and Besley and Persson (2009). However, this improvement has meant an increased ability to tax all social classes and not just the rich. Finally, fighting a war provides a qualitatively different public good than does providing defense in normal times, and this likely impacts support for taxation. While we will present evidence consistent with this view, this argument suggests broader support for taxes on all income groups and not just the rich.
32.See Levi (1997). This is also related to Levi’s claim in Of Rule and Revenue (1988, p. 53) that “rulers can increase compliance by demonstrating that the tax system is fair.”
33.http://www.dailykos.com/story/2013/05/14/1208899/-Whatever-Happened-to-the-Conscription-of-Wealth
34.This is of course a persistent theme in his work, best illustrated by Piketty (2014) and Piketty (2001).
CHAPTER 2: TREATING CITIZENS AS EQUALS
1.Hall and Rabushka (1981, p. 185).
2.See Regent (2014) for a recent historiographical discussion. Guicciardini’s text was also referred to by Friedrich Hayek in his Constitution of Liberty (1960 [2011]), p. 430. In addition to presenting Guicciardini’s arguments against (but not for) progressive taxation, Hayek also followed earlier work by Guicciardini suggesting that the decima scalata became a tool of the Medici to oppress other members of the Florentine elite (see Guicciardini, “Del Reggimento di Firenze”).
3.Translated by Bruce Edelstein (Guicciardini ∼1520 [1867], p. 355). There also exists a prior English translation of Guicciardini’s text (∼1520 [1959]), but we judged it to be less faithful to the original Italian.
4.“On this matter I say that, as the poor man suffers from a tax, so does a rich man, and so it disorders his necessary expenses, as for the poor; for ordinary expenses are not the same for all, but rather they differ according to the different ranks of the citizens; and so it is necessary for the rich to have high expenses in order to maintain their rank, as a small expense is necessary for the poor, and for me, being of average wealth, an average expense is necessary.” Translated by Maria Carreri.
5.This “equal treatment” account would later be cited by Friedrich Hayek in his Constitution of Liberty. See Hayek (1960 [2011]), p. 430.
6.Matteo Palmieri (1429 [1982], pp. 141–142). Translated by Bruce Edelstein. For further discussion of this passage and of Palmieri’s opinions on taxation, see Ricca-Salerno (1881, pp. 32–33) and Isenmann (1995).
7.For general discussions of the luxury debate, see Hont (2006) and Shovlin (2006). See also the closely related work of Pocock (1985) on “virtue” and “commerce.” Eich (2013) sees debates about luxury taxes as presaging later debates about progressive taxation.
8.Jean-Jacques Rousseau, “Discourse on Political Economy,” edited by Victor Gourevitch (1997), pp. 32–33.
9.For the context of Saint-Lambert’s piece, see Hont (2006) and Shovlin (2006).
10.The full text is: “The subjects of every state ought to contribute towards the support of the government, as nearly as possible, in proportion to their respective abilities; that is, in proportion to the revenue which they respectively enjoy under the protection of the state. The expence of government to the individuals of a great nation is like the expence of management to the joint tenants of a great estate, who are all obliged to contribute in proportion to their respective interests in the estate. In the observation or neglect of this maxim consists what is called the equality or inequality of taxation.” The Wealth of Nations, Book V, Chapter 2, paragraph 25.
11.On this point see Hont (2006), Ignatieff and Hont (1983), and Eich (2013).
12.The complete passage is as follows: “The necessaries of life occasion the great expence of the poor. They find it difficult to get food, and the greater part of their little revenue is spent in getting it. The luxuries and vanities of life occasion the principal expence of the rich, and a magnificent house embellishes and sets off to the best advantage all the other luxuries and vanities which they possess. A tax upon house-rents, therefore, would in general fall heaviest upon the rich; and in this sort of inequality there would not, perhaps, be any thing very unreasonable. It is not very unreasonable that the rich should contribute to the public expence, not only in proportion to their revenue, but something more than in that proportion.” Wealth of Nations, Book V, Chapter 2, paragraph 71.
13.John Stuart Mill, The Principles of Representative Government, Book V, Chapter 2, paragraph 7. “For what reason ought equality to be the rule in matters of taxation? For the reason that it ought to be so in all affairs of government. As a government ought to make no distinction of persons or classes in the strength of their claims on it, whatever sacrifices it requires from them should be made to bear as nearly as possible with the same pressure upon all, which, it must be observed, is the mode by which least sacrifice is occasioned on the whole. If anyone bears less than his fair share of the burthen, some other person must suffer more than his share, and the alleviation to the one is not, cæteris paribus, so great a good to him, as the increased pressure upon the other is an evil. Equality of taxation, therefore, as a maxim of politics, means equality of sacrifice. It means apportioning the contribution of each person towards the expenses of government so that he shall feel neither more nor less inconvenience from his share of the payment than every other person experiences from his. This standard, like other standards of perfection, cannot be completely realized; but the first object in every practical discussion should be to know what perfection is.”
14.The full passage is as follows: “Setting out, then, from the maxim that equal sacrifices ought to be demanded from all, we have next to inquire whether this is in fact done by making each contribute the same percentage on his pecuniary means. Many persons maintain the negative, saying that a tenth part taken from a small income is a heavier burthen than the same fraction deducted from one much larger: and on this is grounded the very popular scheme of what is called a graduated property tax, viz. an income tax in which the percentage rises with the amount of the income.” John Stuart Mill, The Principles of Representative Government, Book V, Chapter 2, paragraph 10, 3rd edition 1852.
15.Ibid., paragraph 11, third edition 1852. For Bentham’s contribution see Jeremy Bentham (1794, p. 388). As Seligman (1908) observes, Bentham was not the ultimate originator of the idea of exempting a subsistence minimum from taxation, and he provides several earlier sources.
16.John Stuart Mill, The Principles of Representative Government, Book V, Chapter 2, paragraph 12, 3rd edition 1852.
17.This was the case of the window tax. See Mill’s letter to the Examiner, January 13, 1833, entitled “Necessity of Revising the Present System of Taxation.”
18.John Stuart Mill, The Principles of Representative Government, Book V, Chapter 2, 3rd edition 1852.
19.See Edgeworth (1897) and Pigou (1928).
20.Matthew Weinzierl (2014) has recently presented survey evidence suggesting that citizens prefer to apply a fairness standard rather than a welfarist one when choosing a tax scheme. See also Young (1990, 1995) and Roemer (1996) on the relationship between the concerns of contemporary economists and modern approaches to fairness in resource allocation.
21.See Lionel Robbins (1932, p. 125).
22.See Simons (1938).
23.Pigou (1928, p. 156).
24.Thomas Piketty (1995). Bénabou and Tirole (2006). See Fong (2001) for empirical support as well as Alesina and Angeletos (2005) for a comparison of attitudes between the United States and Europe.
25.See Murphy and Nagel (2002).
26.Another way to think about the implications of Murphy and Nagel’s argument is that deciding on a normatively appealing tax system requires a compelling theory of distributive justice. Here it is worth noting that a great deal of contemporary work on distributive justice since Rawls (1971) has focused on egalitarian accounts, and a central question is what should be equalized in a just society (see, e.g., Sen, 1980; Dworkin, 1981a, 1981b; Cohen, 1989; and Roemer, 1998). Murphy and Nagel’s argument in part suggests that it is not possible to talk about fairness or equality in a tax system without answering the larger distributive justice questions of what should be equal in a just society.
27.See Bowsky (1969, 1981). The quote is located on page 191 of Bowsky (1981). The emphasis is ours.
28.House of Commons, Select Committee on Income and Property Tax, testimony by John Stuart Mill, May 20, 1852. House of Commons Papers vol. 9, p. 299. See also the discussion in Principles of Political Economy V 3.15.
29.Several of these contributions are reviewed by Seligman (1908, pp. 145–147), and for a more detailed review see Gaston Gros (1907), pp. 556–577.
30.John Rawls (1971, pp. 246–247).
31.Speech by citizen Pierre-Gaspard Chaumette, procureur general de la commune March 9, 1793. Archives Parlementaires p. 6. Author’s translation. This speech is also cited by Seligman (1908, p. 32) using an incorrect abbreviation of the text that derives from Gomel (1902, pp. 389–391). See Gross (1993, 1996) for a broader discussion about progressive taxation during this period.
32.See Seligman (1908, p. 145) and Walker (1883, pp. 453–455).
33.On this point see the review by Seligman (1908, pp. 142–145) as well as Walker (1883, pp. 453–455).
34.Except of course the case of a progressive tax on consumption.
35.See Camerer and Fehr (2004) on ultimatum game results.
36.See Camerer and Fehr (2004).
37.See Engl (2011).
38.See Fisman, Jakiela, and Kariv (2014). Their results show that one-third of their subjects can be characterized as purely fair-minded, 15 percent can be characterized as purely self-interested, and that the remaining 52 percent were intermediate between these two classifications.
39.See Henrich et al. (2010).
40.This is demonstrated in the context of the dictator game by Oxoby and Spraggon (2008) and in the case of the ultimatum game by Carr and Melizzo (2013). See also Cappelen et al. (2007) for related results.
41.See Reuben and Riedl (2013).
42.See, e.g., Gilens (1999), Luttmer (2001), Fong (2001), Alesina and Angeletos (2005), Alesina and Glaeser (2004), Durante, Putterman, and van der Weele (2014), and Cavaille and Trump (2015), among many others. It should be added that there is debate in this literature about the importance of self-interest in determining economic policy opinions. Our view is that there is considerable evidence that self-interest, other-regarding preferences, and fairness concerns all inform preferences with substantial heterogeneity across policies and individuals in the weight given to these considerations.
43.The survey was carried out online by YouGov, which employs matched sampling to approximate a random sample of the adult population.
44.The possible increases for those making between $25,000 and $200,000 were 1.1, 1.25, or 1.5 percentage points. The possible increases for those making more than $200,000 were 2, 3, and 4 percentage points.
45.Further details about the survey are available in the online appendix.
46.Examples include “Everyone should pay the same” and “Rich and poor should pay equally.” Note that this category includes cases where people simply say taxes should be equal, as well as cases that explicitly say the government should treat people the same.
47.This “reason” essentially repeats the preference of the respondent for the progressive choice without providing a justification. Examples include “The rich should pay more” or “Lower taxes on the poor.”
48.Of the 500 respondents, 73 did not answer this question at all or provided an incoherent answer. Further, 16 respondents gave multiple reasons and it was not possible to code a primary argument. Our description of the pattern of arguments is therefore for the 411 respondents for which we could classify a primary argument.
CHAPTER 3: THE INCOME TAX OVER TWO CENTURIES
1.Seligman (1911, pp. 42–53). See Conti (1984) for a more detailed description of the Italian case and Clamageran (1868) for France.
2.The Comparative Income Taxation Database (Genovese, Scheve, and Stasavage 2014) also includes the amount from which the top marginal rate applies, the law governing the tax, and the source of this information (original legislation or secondary literature). The data and codebook are available at http://data.stanford.edu/citd. The countries included in the sample are Australia, Austria, Belgium, Canada, Denmark, Finland, France, Germany, Ireland, Italy, Japan, Korea, the Netherlands, New Zealand, Norway, Spain, Sweden, Switzerland, the United Kingdom, and the United States.
3.This observation does not have much substantive importance. As we discuss in further detail below, we also collected data on local income taxes. Once these taxes are included, Switzerland looks more like the rest of the sample.
4.The data on local income tax rates are not comprehensive for an entire country. Rather, we collected data on local rates assuming the taxpayer lived in the largest city. We then determined if there were income taxes levied at the municipal or provincial level and what these rates were.
5.The tax rates for 1925, 1950, 1975, 2000, and 2010 are based on the nominal GDP per capita. We constructed the full schedule graphs for countries in the Comparative Income Taxation Database for which we were able to obtain rates and thresholds for all of the brackets. We concentrate on national income tax schedules at six 25-year intervals: 1875, 1900, 1925, 1950, 1975, and 2000, to which we add the full schedules from the last year in the database (2010). We report not the raw schedules but the schedules in terms of multiples of GDP per capita. Since we do not have reliable information on which thresholds of GDP apply to which percentiles of the population for each country-year in the database, we relied on Piketty (2001), who reports the income thresholds for individuals in France in the 10th, 50th, 90th (and, for some years, 95th and 99th) percentiles. Piketty’s thresholds are the values we use to generate multiples of GDP per capita for the other countries. We calculated France’s multiple per percentile per year by dividing the threshold by GDP per capita, and then averaged the multiples for all the available years (see Piketty 2001, pp. 671 and 672, col. 8–10 of table D12 and col. 4–6 of table D13; the years in these tables range from 1950 to 1998). We find that the average multiples for France by income group are: 0.48 for 10th percentile; 0.83 for 50th percentile; 1.66 for 90th percentile; 2.12 for 95th percentile; and 3.87 for 99th percentile. We round up to 0.5, 1, 1.5, 2, 4, and use these values to find the income thresholds that correspond to the tax rates in the full schedules in other countries. We also include a high multiple of 100 to capture the very top rates in all country series. The remainder of this note discusses the sources of schedules for each of the six countries under consideration, and the decisions that characterized the calculations of the tax rates in terms of multiples of GDP per capita. France, 1925–2010. The schedules for France start in 1925, because the national series start on July 15, 1914. The sources of the full schedules are the following: for 1925, we use the legislative text of August 1926, also reported in Piketty (2001), table 4.2, at p. 263; for 1950, we follow Piketty (2001), table 4.5, at p. 296; for 1975, we follow Piketty (2001), table 4.5, at p. 297; for 2000 and 2010, we use the OECD Income Tax Database (2014).
The GDP per capita for France up to 1998 is the nominal national GDP divided by total population based on Mitchell’s International Historical Statistics. GDP is in millions of old francs up until 1949, then between 1958 and 1960 the French franc was revalued, with 100 existing francs making one nouveau franc. GDP is in billions of new francs from 1950 until 1998. From 1999 onward, nominal GDP per capita is from Eurostat (2014) and is in euros.
The 1925 tax rates are structured around the 25 “tranches de revenue” in the law. Piketty (2001, p. 263) reports thresholds for these rates aggregated for the years 1919–1935, so the multiples of GDP per capita (which we obtain for the year 1926) are indicative. The 1950 tax rates are based on a rescaled version of nominal GDP per capita because, while Mitchell’s GDP values are in millions of francs in 1950, the thresholds in Piketty are still reported in the “old” francs, so we scaled the GDP multiplying it by 100 to concord with the thresholds. The tax rates of 1975, 2000 and 2010 are based on the multiples of the nominal GDP per capita.
Germany, 1925–2010. The schedules for Germany start in 1925, because the national series starts on March 29, 1920. The sources of the full schedules are the following: for 1925, we rely on Dell (2008), table at p. 119; for 1950, we use the brackets from the first postwar income tax law of 1958, and rely on the thresholds in the Bundesministerium der Finanzen (BMF) 2013 review of historical rates, which shows tax benchmarks based on the tax formula of the fiscal law (https://www.bmf-steuerrechner.de/uebersicht_ekst/); for 1975, we use the thresholds of the BMF review; for 2000 and 2010, we use the OECD Income Tax Database (2014) and the BMF 2013 review.
The GDP per capita for Germany up to 1990 is the nominal national GDP divided by total population based on Mitchell’s International Historical Statistics, where GDP is in millions of marks (Deutsche Mark, DM). From 1991 onward, nominal GDP per capita is from Eurostat (2014) and is in euros (ex post).
The 1925 tax rates were calculated by transforming the nominal GDP in DM to GDP in Reichsmark, which was the national currency between 1924 and 1958 and is the value the thresholds are reported in. The value of Reichsmark in DM was 1 Deutsche Mark = 1 Reichsmark for first 600 RM, 1 Deutsche Mark = 10 Reichsmark thereafter, so we multiplied GDP by 10 in order to calculate the multiple of GDP per capita used to find the rates (Information on the value of Reichsmark in terms of DM was retrieved from the Deutsche Bundesbank: http://www.bundesbank.de/Redaktion/DE/Standardartikel/Statistiken/kaufkraftvergleiche_historischer_geldbetraege.html?view=render%5BDruckversion%5D; accessed June 24, 2014.) The 1950 (or precisely 1958) and 1975 tax rates are based on a rescaled version of nominal GDP per capita because the rates reported in the BMF review are in euros: we know the exchange rate of DM to euros from up to 1975 from the Federal Reserve Bank of St. Louis; we then found the euro value of 1958 nominal GDP per capita, which we used to find the rates (see ex-post exchange rates of USD, DM, and euros at http://research.stlouisfed.org/fred2/data/EXGEUS.txt; accessed June 24, 2014). The tax rates of 2000 and 2010 are based on the multiples of the nominal GDP per capita, though note that the schedules in this year are formula-based and the Bundesministerium reports selected middle schedules as orientation.
New Zealand, 1900–2010. The schedules for New Zealand start in 1900, because the national series start on October 11, 1892. The sources of the full schedules are the following: for 1900, we rely on the brackets of the Land Tax and Income Tax Act, 1892; for 1925, 1950, and 1975, we rely on McAlister et al. (2012) and their supplementary appendix; for 2000 and 2010, we use the OECD Income Tax Database (2014).
The GDP per capita for New Zealand up to 1999 is calculated from millions of New Zealand dollars of national nominal GDP divided by population in millions. The data come from Statistics New Zealand. For the year 2000 we retrieved the nominal GDP per capita from the IMF World Economic Outlook Database (WEO 2000). For the year 2010, we retrieved the nominal GDP per capita from the statistics of the NZ government (statistics available at https://www.nzte.govt.nz/en/invest/statistics/; accessed June 24, 2014).
The 1900 and 1925 tax rates were calculated based on thresholds that in the 1892 fiscal law are expressed in fractions of New Zealand pounds, i.e., in schillings. We transformed GDP from NZ dollars into NZ shillings relying on the fact that 1 NZ pound = 2 NZ dollars, and that 1 NZ pound = 20 shillings, so we divided GDP per capita in NZ dollars by 2 and multiplied by 20 (information from Reserve Bank of New Zealand at: http://www.rbnz.govt.nz/notes_and_coins/coins/0094086.html; accessed June 24, 2014). The 1950 tax rates are expressed in New Zealand pounds, so we calculated the rates per multiples of GDP by transforming GDP in New Zealand dollars into New Zealand pounds, thus we multiplied the GDP per capita by 2, because £NZ 1 = $2 NZ. The tax rates of 1975, 2000, and 2010 are based on the multiples of the original nominal GDP per capita.
Sweden, 1875–2010. The schedules for Sweden start in 1875, because the national series start in 1862. The sources of the full schedules are the following: for 1875 and 1900, we rely on Du Rietz, Johansson, and Stenkula (2010), working paper version, appendix D, p. 55; for 1925, we rely on Du Rietz et al.’s rates at p. 57, where we use their base amounts (lower due to their account of average tax cap) multiplied by the withdrawal percentage for 1925 (see table D.6); for 1950 and 1975 we use Du Rietz et al.’s tables at pp. 58 and 59, respectively; and for 2000 and 2010, we use the OECD Income Tax Database (2014).
The GDP per capita for Sweden up to 2000 is the nominal national GDP by expenditure at purchaser prices from Rodney Edvinsson, http://www.historia.se/. From 2001 onward, nominal GDP per capita is from Eurostat (2014).
The tax rates of 1875 and 1900 are based on the multiples of Edvinsson’s nominal GDP per capita. The 1925 tax rate relies on Edvinsson’s GDP per capita, adjusted to the fact that after 1919 the SEK was recoined (20 Kr coins became 10 SEK). The tax rates of 1975, 2000, and 2010 are based on the multiples of the nominal GDP per capita.
UK, 1875–2010. The schedules for the United Kingdom start in 1875, because the national series start on January 9, 1799. The sources of the full schedules are the following: for 1875 and 1900, we rely on discussion in Mitchell (1988) and Mallett and George (1929, p. 399); for 1925, we rely on the brackets in the Finance Act 1925, Part II (p. 10); for 1950, we rely on the Finance Act 1948, Part III (change in ordinary rates) and IV (extraordinary rates); for 1975, we use the schedules from the Crown’s National Archives (see HM Revenue and Customs Archives, http://webarchive.nationalarchives.gov.uk/+/http://www.hmrc.gov.uk/stats/tax_structure/00ap_a2c_2.htm; accessed June 24, 2014); for 2000 and 2010, we use the OECD Income Tax Database (2014). The nominal GDP per capita of the UK from 1875 to 2010 is based on nominal national GDP from Williamson’s “Measuring Worth” project (see Measuring Worth project at http://www.measuringworth.com/; accessed June 24, 2014). The GDP is million pounds and is divided by population. These values are consistent with the Bank of England’s data from the report “The UK recession in context: what do three centuries of data tell us?”, the IMF World Economic Outlook Database (WEO 2000), and more recent data from the OECD (2014).
The tax rates of 1875 and 1900 are based on the multiples of the nominal GDP per capita. The tax rates of 1925 and 1950 (or more precisely 1948) are based on multiples of the nominal GDP per capita that we calculated by multiplying GDP by 10 in order to scale it to the decimal value that the pound had between 1925 and 1960. The tax rates of 1975, 2000, and 2010 are based on the original nominal GDP per capita.
United States, 1925–2010. The schedules for the United States start in 1925, because the national series, although starting in 1863, only has permanent schedules from October 3, 1913, onward. The sources of the full schedules are the following: for 1925, 1950, and 1975, we use the IRS official schedules in nominal GDP; for 2000 and 2010, we use the OECD Income Tax Database (2014), which we checked with the IRS schedules.
The nominal GDP per capita of the United States from 1925 to 2010 is from Williamson’s “Measuring Worth” project (Officer and Williamson, 2006). The data are consistent with the IMF World Economic Outlook Database (WEO 2000) and, for recent years, the United Nations Statistics (2014).
6.The multiples we select are based on the income thresholds in France approximating the 10th, 50th, 90th, 95th, and 99th percentiles reported in Piketty (2001). For France, on average a 0.5 multiple corresponds approximately to the 10th percentile, a 0.8 multiple to the 50th percentile, a 1.7 multiple to the 90th percentile, a 2.1 multiple to the 95th percentile, and a 3.9 multiple to the 99th percentile. In figure 3.3, we round to 0.5, 1, 1.5, 2, 4, and use these values to find the income thresholds that correspond to the tax rates in the full schedules in other countries. We additionally include a high multiple of 100 to capture the very top rates for each schedule. For most countries and years, a multiple far less than 100 is needed to reach to the top marginal rate. However, because a large multiple is necessary for Sweden in 1925 and the United States in 1925 and 1950, we use 100 to ensure that we have the top rate in the figure for all countries and years.
7.At this time in New Zealand, if an individual’s income passed a threshold with a higher statutory rate, the higher rate applied to all income earned by that individual rather than only to the income above the threshold, unlike the case of contemporary income tax systems. This meant that individuals with incomes well above the highest threshold faced a lower marginal rate than those with incomes near the threshold.
8.Further inspection of the full schedules suggests that this conclusion comes with some qualifications. For example, there is important variation in the rate of increase of the statutory marginal rates among cases with similar top marginal income taxes. This type of variation can matter for how the income tax system influences the income distribution, and it is not captured by the top statutory rate measure. Our claim is simply that the top rate is a reasonable proxy for the overall progressivity of the system.
9.We constructed the graphs for selected countries in the Comparative Income Tax Database for which we were able to find a long series of effective rates. The statutory top marginal income tax rates are defined and reported in the database. The effective rates constitute the income tax rates after credits and bonuses of individuals in the top 0.01 percent of the national income distribution (0.05 percent for the Netherlands and United Kingdom). We retrieved the effective tax rates from the sources cited below. The countries under consideration are Canada, France, the Netherlands, Sweden, the UK, and the United States. With the exception of Canada and Sweden, we compare national level statutory and effective rates. For Canada and Sweden, we compare rates that include sub-national income taxes.
The graphs compare the statutory and effective series from the years where they are both available. We also report the Pearson correlation coefficient (ρ) as a measure of the strength of the relationship between the two series. The remainder of this note discusses the sources for effective rates for each of the six countries under consideration, and the decisions that characterized the calculations of these rates if they were not available from secondary sources.
For Canada, 1920–1997, we used the effective rate series for the top 0.01 percent income share presented by Saez and Veall in Atkinson and Piketty’s 2009 book, table 6F.1, pp. 301–302.
For France, 1915–1998, we used the effective rate series for the top 0.01 percent income share presented by Piketty (2001), table B-20, pp. 636–637.
For the Netherlands, 1946–1999, we used the effective rate series for the top 0.05 percent income share presented by Salverda and Atkinson in Atkinson and Piketty (2007), table 10.6, pp. 455–456. The figure reports data from 1946, though Salverda and Atkinson report some effective rates for earlier years due to a break in the time series.
For Sweden, 1903–1990, we used the effective rate series shared by Roine and Waldenström and presented in the chapter “Top Incomes in Sweden over the Twentieth Century” in Atkinson and Piketty (2010). The 0.01 percent income series is illustrated at p. 323. The tax rates were computed using tax tables in Söderberg (1996) and the Swedish National Tax Board.
For the UK, 1937–2000, we used Atkinson’s chapter “The Distribution of Top Incomes in the United Kingdom, 1908–2000” (pp. 83–114) in Atkinson and Piketty (2007). We divided the total tax amount of the top 0.05 percent (top available fractile, although for a smaller amount of years Atkinson shows top 0.01 percent incomes) by their taxable income. We collected the pre-tax income share from table 4.1, pp. 93–94, the post-tax income share for the same fractile at table 4.2, pp. 104–105, and then the total household income and total tax deducted at columns 3 and 4, table 4B.1, pp. 126–127. Using these measures we calculated the total pre-tax income of the top 0.05 percent, the total post-tax income of the top 0.05 percent, and then the effective tax rate by dividing the difference between these two figures by total pre-tax income.
For the United States, 1916–1995, we constructed an effective rate series relying on the Piketty and Saez chapter “Income and Wage Inequality in the United States” (pp. 171–173) in Atkinson and Piketty (2007) and each of the annual Internal Revenue Statistics of Income Reports (http://www.irs.gov/uac/SOI-Tax-Stats-Archive). Note that prior to 1944, “income” means net income and after 1944, income means adjusted gross income (AGI). Generally speaking, net income equals adjusted gross income less allowable itemized personal deductions. Both concepts include gross income less such items as (1) allowable trade and business deductions; (2) travel, lodging, and other reimbursed expenses connected with employment; (3) deductions attributable to rents and royalties; (4) deductions for depreciation and depletion allowable to beneficiaries of property held in trust; and (5) allowable losses from sales of property. However, net income subject to normal tax is net income less exemptions such as applicable personal exemptions for taxpayers and their dependents, dividends on stock in domestic corporations (through 1935), interest on some U.S. obligations and the earned income “credit” (for 1934 through 1943). Net income subject to surtax is all net income through 1933 and net income less the personal exemptions for taxpayers and their dependents thereafter.
We constructed the effective tax rates for the top 0.01 percent incomes by dividing total income tax paid by the top 0.01 percent income holders and the total income for the top 0.01 percent of taxpayers. We first consulted Piketty and Saez to retrieve the total number of tax units in the United States in the given year. Next we consulted the basic statistics of each of the IRS reports, and identified the table that lists numbers of returns in each income class as well as total net income/adjusted gross income for each income class and total income tax liable. Starting from the richest returns we added up income and tax liability until we arrived at the number of tax units from Piketty and Saez. Since this number was often reached in the middle of an income class, we prorated the total income and total tax liability within this class. Using the above we calculated effective rates for the top 0.01 percent for every year. The specific tables used in the IRS reports were: 1916 pp. 18–19; 1917 pp. 26–27; 1918 pp. 34–35; 1919 pp. 41–42; 1920 pp. 46–47; 1921, pp. 40–41; 1922 pp. 79–83; 1923 pp. 59–63; 1924 pp. 104–108; 1925 pp. 87–93; 1926 pp. 74–79; 1927 pp. 68–72; 1928 pp. 74–78; 1929 pp. 61–65; 1930 pp. 69–73; 1931 pp. 60–64; 1932 pp. 66–70; 1933 pp. 68–71; 1934 pp. 57–60; 1935 pp. 74–77; 1936 pp. 84–88; 1937 pp. 117–121; 1938 pp. 17; 1939 pp. 8–9; 1940 pp. 78–82; 1941 pp. 76–80; 1942 pp. 92–95; 1943 pp. 124–129; 1944 pp. 65–67; 1945 pp. 69–71; 1946 pp. 63–65; 1947 pp. 65–67; 1948 pp. 65–67; 1949 pp. 75–77; 1950 pp. 35–37; 1951 p. 25; 1952 p. 18; 1953 p. 23; 1954 p. 33; 1955 pp. 6, 18; 1956 p. 20; 1957 p. 20; 1958 p. 24; 1959 p. 23; 1960 p. 32; 1961 pp. 32–33; 1962 pp. 32–33; 1963 pp. 33–34; 1964 p. 10; 1965 p. 8; 1966 p. 6; 1967 p. 7–8; 1968 p. 6; 1969 p. 9; 1970 pp. 7–8; 1971 pp. 6–7; 1972 pp. 6–7; 1973 pp. 4–7; 1974 pp. 6–9; 1975 pp. 5–6; 1976 pp. 7–10; 1977 pp. 7–9; 1978 pp. 6–8; 1979 pp. 9–11; 1980 p. 36; 1981 p. 34; 1982 p. 38; 1983 p. 6; 1984 pp. 10–12; 1985 pp. 12–14; 1986 pp. 16–18; 1987 pp. 20–22; 1988 pp. 18–20; 1989 pp. 18–20; 1990 pp. 16–18; 1991 pp. 19–21; 1992 pp. 29–30; 1993 pp. 29–30; 1994 pp. 30–31; 1995 pp. 27–28.
10.See Piketty and Saez (2007). We should note that the effective tax rates series in their paper differs from the one reported here because they adopt a different concept for gross income, and this is linked to the fact that they are also considering the incidence of corporate taxation.
11.None of our conclusions are substantially affected by focusing on overall universal rather than universal male universal suffrage. As is the case with unitary states, for federal states, such as Germany, our universal male suffrage measure takes account only of suffrage laws established at the national level and applying to the national legislature, provided that such laws exist. We also take account of available information involving restrictions on certain categories of men, such as African Americans in the United States prior to 1965. In cases where a country established universal suffrage before becoming fully independent from another power, we use the date of the state’s independence to code this variable. Unless otherwise noted below we used either Caramani (2000, p. 53) or Mackie and Rose (1974) to code this variable. Dates of establishment of universal suffrage for the countries in the sample are as follows: Australia 1902, Austria 1897, Belgium 1894, Canada 1921, Denmark 1918, Finland 1917, France 1848, Germany 1871, Ireland 1922, Italy 1913, Japan 1925, Netherlands 1918, New Zealand 1879, Norway 1905, South Korea 1948 (source: Croissant, 2002), Spain 1869–1875 (1888), Sweden 1911, Switzerland 1848, United Kingdom 1918, United States of America 1965.
12.The before and after approach, however, requires us to believe that had a country not adopted universal male suffrage or democracy, tax policy would have remained constant. It also requires us to believe that universal male suffrage was adopted for reasons unrelated to other factors influencing the top rate of income taxation. These are strong assumptions, but examining the time series evidence is nonetheless informative.
13.The main statistical results in this book remain unaltered if we instead adopt an earlier date for universal male suffrage in the United States.
14.In the online appendix we address some of the potential limitations of the before and after comparisons by conducting a complete set of regression analyses averaging the data over five-year periods and estimating ordinary least squares regressions with country and time period fixed effects and time varying control variables as well as a wide number of alternative specifications. There is no evidence in these further analyses of a significant positive correlation between universal suffrage and the top rate of income taxation.
15.It is worth noting that there are a few cases that appear to be more consistent with the democracy hypothesis. For example, Austria increased its top marginal income tax rate from 0 to just below 5 percent in the same year (1897) that it adopted universal male suffrage.
16.See the online appendix.
17.See Mares and Queralt (forthcoming) for an analysis of voting in parliament on the income tax in 1842. Rather than emphasizing democracy, they highlight the conflict between landowners, who preferred income taxes to direct taxes on land, and the newly emerging industrial class, who preferred the opposite. This intersectoral conflict is also observed in parliamentary debates about the income tax. Its implication is that even an interpretation that the expansion of the franchise contributed modestly to the low rates of income taxation that we observed during the nineteenth century is too generous to the democracy hypothesis. This is consistent with our more general claim that the expansion of the franchise did not typically drive countries to adopt high income taxes on the wealthy.
18.According to the data in Flora (1983 vol.1, p. 149), in the election of 1910, 62.4 percent of males of eligible age could vote.
19.This definition and data is from Boix and Rosato (2001). The definition is a modification of the definition used by Przeworski et al. (2000) to a context where the suffrage may be restricted. The countries in the sample are coded as having competitive elections for the following years: Australia 1901–2013; Austria 1920–1932, 1946–2013; Belgium 1894–2013; Canada 1867–2013; Denmark 1901–2013; Finland 1917–2013; France 1848–1851, 1870–1939, 1945–2013; Germany 1919–1932, 1946–2013; Ireland 1922–2013; Italy 1946–2013; Japan 1952–2013; Netherlands 1897–2013; New Zealand 1856–2013; Norway 1905–2013; South Korea 1960, 1988–2013; Spain 1931–1936, 1977–2013; Sweden 1911–2013; Switzerland 1848–2013; UK 1885–2013; United States 1800–2013.
20.This null finding for competitive elections also holds in the main fixed effects regression analyses that we report in the online appendix. We also considered other potentially important features of democratic political institutions. For example, we investigate whether it is the introduction of direct elections for the lower house that moves countries to tax high incomes at higher rates. Finally, we also consider the effect of having an unelected upper house, which might tend to veto redistributive policies. Again, for the most part, these analyses do not suggest a systematic impact of these institutions on tax policies. A partial exception to this argument is that there is some mixed evidence that the absence of an unelected upper house leads countries to adopt tax policies with higher rates.
21.As in the case of universal suffrage, a simple pooled comparison across all years does yield the expected partisan differences, with left country-years having an average top rate of 49.6 percent and non-left country-years having an average rate of 23.5 percent. The problem, of course, is that this comparison takes no account of the fact that left governments appeared in the twentieth century rather than the nineteenth, and there are many other factors that differed between the two centuries.
22.Canada, Ireland, South Korea, and the United States are not coded as having labor or socialist left parties that are in government. There are also no transitions to left parties for Switzerland because left parties never control a majority of the seats in the executive.
23.See the online appendix for further evaluation of the partisan hypotheses. Among other things, this analysis allows us to relax some of the assumptions of the evidence discussed here. The results are quite similar with there being some mixed evidence consistent with the partisan hypothesis, but the magnitude of the effect is relatively small and the result is sensitive to econometric specifications.
24.This could occur if high top rates prompt top earners to reclassify earnings as something other than income, if they exert less effort to receive higher salaries, or if they substitute leisure for labor.
25.The data were accessed from http://topincomes.g-mond.parisschoolofeconomics.eu/ and are discussed in detail in Atkinson and Piketty (2007, 2010) as well as other publications in the top incomes project.
26.The correlations between the top 1 percent income share and top rate are: 0.06 for 1925, 0.47 for 1950, −0.36 for 1975, and −0.17 for 2000. Obviously, the negative correlations for 1975 and 2000 are substantially stronger than for the top 0.01 percent income share measure, but the key point is that the correlations are not stable over time for either measure.
27.In other words, we conducted a pooled Granger test including country fixed effects and time effects modeled either with a common time trend or year dummies. See the online appendix for a full description and results.
28.Sokoloff and Zolt (2006) present an historical analysis of the role of inequality in the formation of tax systems. Their study suggests that large variations in inequality across the Americas did not lead to significant differences in national-level tax patterns. They do find, however, that greater inequality was associated with less progressive taxes at the state and local level. See also Zolt (2009). On the effect of tax rates on inequality, see Roine, Vlachos, and Waldenström (2009) for a separate analysis that suggests that tax progressivity reduces top income shares.
29.Note that with the exception of a brief period for France under German occupation, all of the countries in figure 3.8 were democracies for the years considered, and so the lack of a positive correlation between inequality and top income tax rates is not due to a lack of democracy.
30.See, e.g., Atkinson and Piketty (2007, 2010) and Piketty (2014).
31.As discussed below, we generally use 2 percent of the population in the military to indicate whether a country is “mass mobilized” for war. Although the data on population mobilization for the full sample do not start until 1816, France and the United Kingdom slightly surpassed the 2 percent level of mobilization for selected years during the Napoleonic Wars. These countries, as well as others, surpassed this level for selected conflicts prior to the beginning of the nineteenth century. Nonetheless, as we discuss in chapter 7, World War I marks a significant departure in the extent of mass mobilization for war. It is thus a natural starting point for evaluating this argument. One plausible alternative would be to start with the American Civil War. This conflict was marked by high levels of mobilization and the early use of railroads to transport men and supplies. As we have already discussed, it also featured an early adoption of an income tax in the North.
32.In 1848, a deputy to the German Federal Assembly proposed a progressive income tax with a top rate of 33.3 percent. Also in 1848, Pierre-Joseph Proudhon proposed to the French Constituent Assembly that it establish an income tax with a top rate of 50 percent. See Seligman (1911, pp. 235, 279).
33.Economist, March 10, 1883.
34.See the findings of Aidt and Jensen (2009) on this point.
35.For one early discussion of the effect of the increase in taxation on the distribution of incomes and wealth, see Bowley (1930).
36.The United States did increase its top marginal tax rate prior to World War I, from a prior rate of 7 percent to 15 percent in 1916. However, the increase upon entering the war was dramatically larger, moving from 15 percent to 67 percent in 1917.
37.See Perry (1955, p. 162).
38.See Atkinson and Piketty (2007, p. 95).
39.See Piketty (2001, p. 556) and Saez and Veall (2007).
40.In 1911 in Sweden and 1918 in the Netherlands.
41.See, for example, Andre (1975) on labor unrest in Sweden in the 1917–18 period, particularly in the wake of the Russian Revolution.
42.For the Netherlands, van Zanden (1997) emphasizes the lack of movement toward progressive taxation as right and center-right governments in the interwar period maintained a system based primarily on indirect taxation and relatively low top income tax rates. This raises the possibility that progressive income taxes failed to develop early in the Netherlands because the left was not yet in government. But among the war mobilization countries that adopted progressive tax systems, such as Canada, France, and the UK, parties of the left were not in power either.
43.Even U.S. entry into the war does not seem likely to be a result of such a selection mechanism. President Woodrow Wilson won the 1916 election on the slogan “He kept us out of war” and likely would have never entered the war if it were not for Germany’s decision to implement unrestricted submarine warfare.
44.The four mobilization cases are Canada, France, the United Kingdom, and the United States. The four nonmobilization countries are Japan, Netherlands, Spain, and Sweden. These countries were selected based on available data and in order to have a balanced number of mobilization and nonmobilization countries.
45.See Scheve and Stasavage (2010) and the online appendix for information on the calculation of these rates.
46.See the online appendix for an econometric evaluation of the proposed interaction between war mobilization and democracy. The evidence reported there is consistent with the impact of war mobilization being larger in democracies, though this result is somewhat sensitive to the specification. In general, we do not have that many non-democratic countries in the sample, and therefore our tests for this interaction are underpowered.
47.Again, this describes nineteenth-century mobilization for the sample, beginning in 1816. Some countries did surpass the 2 percent threshold in the Napoleonic Wars and before. See chapter 7 for a more complete discussion. Although we cannot include the Napoleonic Wars in our statistical analysis due to data limitations, it is in many ways an intermediate case consistent with our argument. While countries mobilized larger armies than in previous conflicts, the mobilization levels were much lower than in the conflicts of the early twentieth century. There was some movement toward taxing the rich, as exemplified by Britain’s adoption of the first modern income tax, but these conflicts did not radically transform the politics of tax fairness as later mass mobilized conflicts would do.
48.More specifically, it would mean that our analysis relies on a comparison between mobilizers and nonmobilizers whose choice of tax rates was affected by the tax rates chosen by mobilizers. Ideally, we would be comparing mobilizers with non-mobilizers who were making strictly independent choices.
49.We have focused on the role that democracy, partisanship, inequality, and mass warfare have played in the timing of income tax innovations. Although we have discussed these independently, we have examined their influence simultaneously as well. These analyses allow us to control for each factor and we arrive at identical conclusions. See the online appendix for detailed results. See also Velez (2014) for an independent analysis that extends Scheve and Stasavage (2010) and also finds that mass mobilization wars cause a substantial increase in tax progressivity.
50.The other logical possibility is, of course, that the observed correlation between mass mobilization and top rates of taxation is spurious and does not reflect a causal relationship. Considering this possibility requires us to identify those factors that might be correlated with both war participation and tax policy that could account for the correlations that we have documented. Some of these factors, such as democracy, partisanship, and inequality, have already been considered. On the one hand, given the modest, at best, correlations between these factors and top rates of income taxation, it is unlikely that these factors account for the relationship between mass war mobilization and top rates. On the other hand, our analyses reported in the online appendix explicitly control for these factors, and we still find a strong relationship between mass warfare and top rates. We also evaluate the potential importance of other factors that can be measured and that could bias our conclusions. These include electoral institutions, trade openness, economic crises, and policy diffusion. We find somewhat mixed evidence of significant correlations between economic crises, trade openness, policy diffusion, and top rates of income taxation but find robust evidence of a positive partial correlation between majoritarian electoral systems and top rates of income taxation (but little evidence of a significant interaction between inequality and political institutions). Incorporating these measures, however, does not substantially change our estimates of the impact of war mobilization. Still there could be other potential confounders that are much more difficult to measure and assess. Our analysis in the online appendix can control for some of these, such as time constant unobserved heterogeneity at the country level, through the use of country fixed effects or common shocks like the Cold War through the use of period fixed effects. We also consider other methods to rule out some types of confounders and strengthen our causal interpretation. That said, a causal interpretation of the correlation that we document requires some assumptions, specifically those associated with difference-in-differences estimates.
51.One might also consider two further alternative explanations. First, progressive taxation may have simply been in vogue during the early twentieth century, and when faced with a wartime expenditure shock, governments resorted to it. In other words, it was a bit like a policy that was in search of an application. This argument is plausible, but it fails to fit the evidence. Governments that did not mobilize for World War I also faced substantial expenditures even if at a lower level than those that mobilized for war. There is no reason why, if it was in vogue, progressive taxation would not have been applied in these countries either. Certainly, this is what proponents of progressive taxation in these countries argued. A second possibility is that war may alter beliefs about the incentive effects of high taxes on income and wealth or that it may make policymakers discount these future effects, providing states with greater flexibility to tax the rich. This mechanism would not suggest that the effect of mass mobilization wars should be larger in democratic countries. Nor would it anticipate that governments would raise top tax rates after a war’s end, and not only during the conflict itself.
52.For a model suggesting this alternative result see Shayo (2009).
53.These surveys have previously been considered by Campbell (2009).
54.See American Enterprise Institute, “Public Opinion on Taxes: 1937 to Today,” 2012.
55.Another aspect of the necessity argument is that mass warfare disrupted trade and tariff revenue to such an extent that taxing high incomes and wealth was needed. To some extent our discussion in chapter 5 covers this possibility. That said, it is important to note that it was not necessarily those countries for which trade was most disrupted that rates increased the most (e.g., the United States). Further, this mechanism would not predict that the effect was largest in democracies as discussed above and documented further in the online appendix. It is also important to consider the fact that war mobilization seems to have continued to influence policymaking after the war. This is inconsistent with the disruption of trade being the primary mechanism. Finally, the war mobilization correlation remains when we control for trade openness.
CHAPTER 4: TAXING INHERITANCE
1.See Book 55, chapters 24 and 25, Harvard Loeb Classical Library.
2.See Johnston (1965), p. 153. Similar views were expressed in the United States when it first established a stamp tax on estates in 1798. See Stabile (1998), p. 126.
3.Piketty and Saez (2013).
4.See Congressional Budget Office, 2009.
5.Full details can be found in Plagge, Scheve, and Stasavage (2011) and in our 2012 American Political Science Review article entitled “Democracy, War, and Wealth: Lessons from Two Centuries of Inheritance Taxation.”
6.See Poterba (2001).
7.See Kopczuk and Saez (2004).
8.See Seligman (1925). The precise quote is on page 25.
9.These results are presented in Piketty (2001, p. 237). They derive from Caillaux (1910, pp. 208–209). See a more complete discussion in chapter 6.
10.The data from the census bureau show that in 1860 the United States had a total national wealth of $16.2 billion, and $12 billion of this was subject to property tax. Taxes collected amounted to $94.2 million, or 0.8 percent of the assessed value of wealth subject to ad valorem (property) taxation or 0.6 percent of total national wealth. In 1912, the United States had a total national wealth of $187 billion and $69.5 billion of wealth was subject to property taxes. Taxes collected amounted to $1.35 billion, representing 2.0 percent of wealth subject to ad valorem (property) taxation or 0.7 percent of total national wealth. The change between 1860 and 1912 was that tax rates had gone up, but the fraction of property subject to property taxation had shrunk. These two effects offset each other so that total taxes divided by total national wealth stayed roughly the same. See Bureau of the Census (1913), pp. 24–25, 747, 750.
11.See Du Rietz and Henrekson (2014).
12.Take the case of someone who held a stock of wealth for thirty years before dying, and for simplicity we will assume that they neither earned income on this nor did they draw on it for consumption. Now assume that the top marginal rate of wealth taxation from 1948 (1.8 percent) applied for each of the thirty years. Then assume that at death the wealth holder’s heirs paid the top marginal inheritance rate from 1947 (70 percent) on the entirety of the remaining sum. Finally, consider what marginal inheritance tax rate would have had to be in place in order to leave the heirs with the same sum in the case where there had been no prior wealth tax. The answer is 82.6 percent. So the difference between 70 percent and 82.6 percent is significant, but not dramatic.
13.See Eichengreen (1989) for a review.
14.See Jens Beckert (2004) for an important survey of historical debates about inheritance in the United States, France, and Germany.
15.See Shultz (1926) for an example.
16.New Zealand had the highest marginal inheritance tax rate in 1913, followed by the United Kingdom at 15 percent.
17.The figure is drawn from Flora (1983, p. 339).
18.See Graetz and Shapiro for a discussion of lobbying over the Estate Tax in the United States (2005). See also Batchelder (2008) for a survey of issues surrounding the U.S. Estate Tax.
19.This is a point that has been previously emphasized by Duff (2005).
20.For the details, see Roine and Waldenström (2014). Roine and Waldenström (2014) also present a comprehensive analysis of the determinants of both top income and wealth shares over the long run. See also the earlier study by Ohlsson, Roine, and Waldenström (2007).
21.This analysis around World War I drops Finland from the list of ten countries for which we have wealth inequality data because it was not independent until the middle of the conflict. We include all ten countries for which we have data in the analyses that follow.
22.This fact does not, however, exclude the possibility that the correlation is spurious due to other factors simultaneously influencing wealth inequality and top inheritance tax rates across countries.
23.The figures for the other cases do not change our conclusions. These six cases were chosen to most closely follow, subject to data availability, those presented for the income tax.
24.As an aside, it might seem obvious that higher inheritance taxes will reduce wealth inequality. There is, however, a debate between economists about this. The key issue revolves around the idea of whether individuals are concerned only about themselves or are acting to further the welfare of a family dynasty. If it is the latter, then when faced with high inheritance taxes, parents may simply save more so that they can still transmit a sufficiently large inheritance to their children after taxes are taken into account. Gary Becker and Nigel Tomes first proposed this idea in a paper published in 1979. More recent work by Benhabib, Bisin, and Zhu has suggested that this effect would be more limited. See the details in Benhabib, Bisin, and Zhu (2011). Like Becker and Tomes, they develop a dynastic model with finitely lived agents. However, they assume, unlike Becker and Tomes, that capital income in the model is stochastic. This generates the different result for taxation.
25.See the online appendix.
26.Soward (1919, p. 130).
27.See Scheve and Stasavage (2012).
28.These averages do not include data from Ireland, Finland, Germany, Norway, and Australia since these states had universal male suffrage from independence.
29.Mass mobilization is defined as in the previous chapter: an active participant in the war with at least 2 percent of the population mobilized at some point in the conflict.
30.Figures on inheritance sizes are from Piketty (2001, pp.746–747).
CHAPTER 5: TAXES ON THE RICH IN CONTEXT
1.For the early history of capital levies, see chapter 11 in Soward (1919).
2.See Eichengreen (1989). See also the earlier study by Hicks, Hicks, and Rostas (1941).
3.See Rostas (1940). The Czech levy was the most successful, whereas the Hungarian and Austrian levies were plagued by implementation problems.
4.See Hicks et al. (1941).
5.See Pigou (1918), p. 145.
6.See Brandes (1997) for a history of war profits in America.
7.Gallup Poll number 63, January 1937, based on a sample of 2,400 adults.
8.British Institute of Public Opinion Survey number 52, November 1938, based on a sample of 1,171 adults.
9.See Hicks et al. (1941), pp. 5–6.
10.See Haig (1929) for the most extensive account and Hicks et al. (1941) for further discussion.
11.See Grotard (1996).
12.See Hicks et al. (1941 pp. 164–168).
13.See the analysis in Frydman and Molloy (2012). They show that there was an effect of salary increase limits, but the impact varied significantly from industry to industry. In some industries, exemptions from the rule were granted in order to aid in the prosecution of the war. Leff (1991) provides the political background to these measures.
14.See Atrostic and Nunns (1991) for a review.
15.See Samuel (1919).
16.See Seligman (1925), chapter 22 and p. 692 for the specific figures.
17.See Adler (1951) and Colm and Tarasov (1941). Adler’s study is not ideal in that he used a slightly different methodology from Colm and Tarasov. Also, his highest income category was only $7,500 per year, so this study is not truly informative about top earners.
18.See Aidt and Jensen (2013).
19.See Amenta and Skocpol (1988) for the US case.
20.See Dudley and Witt (2004).
21.See Seligman (1925), chapter 23 for an example of this reasoning.
22.See Toma (1992).
23.See Grady (1968).
24.See also Tooze (2006) for the role of financial repression in Germany’s financing of World War II.
25.Figure 3.12 in chapter 3 plots the top rate of income taxation and the size of government as measured by total revenues divided by GDP and provides a striking picture that variation in taxes on the rich over time are not well accounted for by the size of government.
CHAPTER 6: THE CONSCRIPTION OF WEALTH
1.It should be noted, however, that the top rate of inheritance tax had reached 8 percent in 1895.
2.Seligman (1911, p. 130). Full speech available at http://hansard.millbanksystems.com/commons/1842/mar/11/financial-statement-ways-and-means#S3V0061P0_18420311_HOC_23
3.Ibid.
4.See, for example, the debate on April 29, 1842 between John Roebuck, Peel, and the Chancellor of the Exchequer, Henry Goulburn, regarding the inequality of the tax. Roebuck argues that “He [Peel] had to maintain three propositions: first, that there was an inequality in this mode of taxation; secondly, that the tax, being unequal, it sinned against those great canons of taxation which had been laid down by every man whose opinion was an authority on the subject, and that therefore if it were within their power to take away that inequality, it was their duty to do so; and, thirdly, that there was a mode by which this inequality might at least be reduced (for he would not go further than that), and that therefore a necessity was imposed upon the right hon. Baronet [Peel] to follow out that proposition.” Source: http://hansard.millbanksystems.com/commons/1842/apr/29/income-tax#S3V0062P0_18420429_HOC_20.
5.It should be noted, however, that the top rate of inheritance tax had reached 8 percent in 1895 and 15 percent in 1907, where it remained prior to the beginning of World War I.
6.“Appendix II: The Incidence of Taxation.” Wednesday, January 27, 1909. (p. 103). Mr. Philip Snowden, Chairman.
7.“Appendix II: The Incidence of Taxation.” Wednesday, January 27, 1909. (p. 105). Mr. Philip Snowden, Chairman.
8.“Appendix II: The Incidence of Taxation.” Wednesday, January 27, 1909. (p. 107). Mr. J. Ramsay MacDonald.
9.See the discussion in Daunton (2001) and in particular p. 358.
10.The Times, June 24, 1914, p. 9, col. G.
11.Mr. Wardle, House of Commons Debate, May 18, 1916, Vol. 82 cc1664-782, http://hansard.millbanksystems.com/commons/1916/may/18/finance-bill#S5CV0082P0_19160518_HOC_306.
12.Trades Union Congress resolution, September 1916, cited in Daunton (1996, p. 890). Ramsay MacDonald, who would later become the first Labour Prime Minister, published an informative pamphlet entitled The Conscription of Wealth. It outlines Labour’s view of how the war should be financed and illustrates the compensatory logic.
13.Economist, March 31, 1917, p. 579.
14.Sprague (1917), p. 5. Note also that Sprague played an important role in the U.S. debate about funding the war, lobbying publicly for high income and profits taxes.
15.Economist, April 8, 1916, p. 663.
16.This was in part because the UK Treasury judged that the imposition of a levy of this type would significantly reduce the revenues generated from the recently adopted high top rates of income tax and estate duty. See Daunton (1996) on this subject.
17.http://hansard.millbanksystems.com/.
18.The coding definitions were: (1) Equal treatment—Responses specify a preference for the government treating citizens the same. The orientation of the speech has to be against the income tax or higher rates. (2) Ability to pay—Responses specify that the rich are better able to afford or will be less harmed by a tax increase than the poor. This could include “equal sacrifice” arguments if the interpretation of equal sacrifice is that taxing the rich more is justified because the utility loss will be equal across the rich and the poor. The orientation of the speech had to be for the income tax or higher rates. (3) Compensatory—Responses that suggest a tax policy is justified because of other inequalities, advantages, or sacrifices due to state policy. This could include reference to the burden caused by other taxes levied by the state. It could include statements about the conscription of labor in the war effort. It could also simply include general ways in which the state facilitates the incomes of the wealthy. The orientation of the speech had to be for the income tax or higher rates. All other arguments such as economic efficiency, bureaucratic efficiency, prudence, or other fairness arguments were coded in a residual category. If more than one of the three fairness arguments was made, we coded it according to which argument was predominant. The coding was done independently by three undergraduate research assistants. The results that we discuss below are evident for each of the three sets of codings. For presentation purposes, we combined the codings by assigning each speech to a category if two or three coders agreed on the coding. For the few speeches for which there was no agreement, we used multiple imputation. The results in figure 6.1 and discussed in the text are based on multiple imputation estimates. See online appendix for the further details about our coding procedures and estimation methods.
19.We independently verified the decreased use of ability to pay arguments with a complementary analysis. We searched all Hansard speeches from the first use of the term “equality of sacrifice” to 1930. We then read the speech and determined if it was about government finance. If so, we then asked the question whether the speech could be interpreted as an ability to pay argument. These data suggested first that, as expected, there were more references to equality of sacrifice in wartime. There were 32 references between 1880 and 1913 but 47 references between 1914 and 1918. More to the point, the frequency of references to ability to pay in the discussion of equality of sacrifice dropped dramatically during wartime. In the period between 1880 and 1913, 39 percent of equality of sacrifice references can clearly be interpreted in terms of ability to pay. This is quite a high figure because in a large number of cases the reference to “equality of sacrifice” may well have meant ability to pay in the speaker’s mind, but the discussion was sufficiently generic that we could not tell. Now consider what happened between 1914 and 1918. Only 3 percent of “equality of sacrifice” references can be interpreted in terms of ability to pay. This is a very dramatic drop-off and is consistent with the data in figure 6.1 suggesting that arguments favoring taxing the rich—including the meaning of equality of sacrifice—moved from emphasizing ability to pay to focusing on compensatory considerations.
20.See Perry (1955, p. 69).
21.Ibid., pp. 72, 107.
22.In addition to concern about the popularity of a new tax, Liberals were concerned about the reaction of subnational governments that had adopted direct taxes. Moreover, Liberals and Conservatives both viewed Canada’s low taxes as a policy good for attracting new immigrants.
23.“Sir John A. Macdonald’s Address to the People of Canada, February 7, 1891,” in Carrigan (1968).
24.Carrigan (1968, pp. 402–404).
25.As mentioned above, another perspective from which to understand motivations of both Conservative and Liberal governments not to resort to direct taxes is that both parties often argued that Canada’s low tax environment attracted industrious immigrants. See, e.g., discussion in Perry (1955, pp. 144–145).
26.See ibid., pp. 143–146, for a full discussion.
27.Cited in ibid., p. 151.
28.Ibid.
29.See Perry (1955).
30.The initial war profits tax was really a high profits tax that taxed 25 percent of profits above a 10 percent return on capital for individuals, firms, and partnerships and a 7 percent return for companies (Hicks, Hicks, and Rostas 1942, p. 171).
31.Perry (1955, p. 152).
32.The Times, July 30, 1915, p. 6, col. F.
33.White’s April 24, 1917 speech in House of Commons printed in the Globe, April 25, 1917, p. 4.
34.It is true that the April 1917 budget did make the war profits tax progressive, raising the top rate to 75 percent on profits in excess of 20 percent of capital (Perry 1955, p. 155). This change followed the compensatory logic described for the initial establishment of the tax in 1916 with continued pressure on the government to prevent some citizens from benefiting from the war effort.
35.The Military Service Act was passed on July 6, 1917. Note that conscription was not implemented until a bitter election was fought in December 1917 primarily over the issue of conscription. Borden won a landslide victory running in coalition with many Liberal MPs under the Unionist Party label but against the Liberal Party’s leader Wilfred Laurier.
36.Borden Papers, cited in Robin (1966), p. 63.
37.Robin (1966).
38.Cited in Perry (1955, pp. 155–156).
39.It should be noted that in its effort to assure victory in the election, the government enacted laws disenfranchising conscientious objectors and citizens born in enemy countries. It also passed legislation granting women related to soldiers the right to vote and allowing the votes of soldiers abroad to choose what constituency they voted in.
40.Liberal Party Platform in Carrigan (1968, p. 72).
41.Unionist Platform in ibid., p. 77.
42.The Union Party won the election by a landslide with almost all the seats won by the Laurier Liberals coming from Quebec. The government continued to push tax policy in a more progressive direction for the remainder of the war and the years immediately afterward. Many Conservatives, including White, had hoped that the income tax would be temporary, but in the early 1920s, war debt and pensions made many of the same arguments observed during the war still relevant if not as widely held.
43.There are many excellent accounts of the history of the U.S. income tax. These include Mehrotra (2013), Bank, Stark, and Thorndike (2008), Brownlee (2004), Joseph (2004), Weisman (2002), and Witte (1985). It is important to note that the role of war in the development of the U.S. tax system features strongly in these works. Our work differs in its account of when and why wars have mattered. We show that this relationship is connected to a more general phenomenon involving compensatory arguments, and we identify this effect in political debates about taxation around the world. We also present new comparative empirical evidence over two centuries.
44.The Confederacy also adopted an income tax and various other levies aimed at the wealthy and those viewed as profiting from the war. Many of these taxes were adopted late in the conflict and there were considerable problems in collecting them.
45.The U.S. Constitution states that direct taxes should be apportioned on the basis of population. The ambiguity and debate was whether or not the income tax was a direct tax as understood in the Constitution.
46.See Hill (1894).
47.Ibid., p. 418.
48.Cited in ibid., pp. 438–439.
49.Cited in ibid., p. 439.
50.Bank et al. (2008, p. 41). It is important to point out that conscription in the Civil War allowed the wealthy to buy substitutes, exacerbating the sense of inequality in the war effort. See also Bank (1996) on this point.
51.Cited in Bank et al. (2008, p. 43).
52.Cited in Hill (1894, p. 425). It should be noted though that Morrill was a reluctant supporter of the income tax overall and played an important role in getting it passed.
53.A partial exception to this was the War Revenue Act of 1898, which instituted a federal estate tax with graduated rates. This tax had a low top rate of 2.25 percent and was associated with the Spanish-American War. It lasted until 1902.
54.The compromise legislation did implement a modest corporate income tax, but overall legislation was a victory for Republican conservatives.
55.See Bank (1996).
56.For a detailed account of the Revenue Act of 1916, see Brownlee (1985).
57.Note that the Act introduced an estate tax that differed substantially in details from the inheritance tax implemented during the Spanish-American War. As noted in chapter 4, we use the terms inheritance tax and estate tax interchangeably in this book.
58.For example, William Jennings Bryan wrote in The Commoner that new taxes should be “on large incomes and inheritances, rather than on the incomes of those who will have to do the fighting if there is any fighting to be done.” Cited in Brownlee (1985), p. 185.
59.This is not to deny that there was still significant opposition to these prewar reforms. There was widespread business and conservative Republican opposition. These arguments featured some simple equality arguments and many efficiency arguments about the impact of the new taxes on economic performance.
60.For example, Sprague also published his arguments from the Economic Journal in The New Republic in February 1917. This, of course, was before U.S. entry into the war, but the article was nonetheless extremely influential.
61.National Tax Association (1917, p. 215).
62.Ibid., p. 216.
63.See Bank et al. (2008, pp. 62–76) for further examples.
64.See Otto Kahn (1918).
65.“Resolution adopted at Nebraska State Convention of Non Partisan League.” The Commoner, Lincoln Nebraska, July 1, 1918. See also “Nonpartisan League Platform” in the Montana Nonpartisan, Great Falls Montana, September 2, 1919, and “Labor Firmly Behind U.S. In Its War” in the Grand Forks Herald, September 3, 1917.
66.“Conscription of Wealth” in Washington Standard, Olympia, Washington. August 31, 1917.
67.“Conscripting Wealth,” in the Grand Forks Herald, Grand Forks, North Dakota, August 14, 1918.
68.“Conscription of Wealth From England’s Viewpoint,” in New York Tribune, January 27, 1918.
69.As we saw in chapter 4, the Revolution did lead to the creation of a modern inheritance tax, even if this tax was not implemented until after the Revolution’s end. This tax was set at very low marginal rates throughout the nineteenth century
70.See in particular Asselain (2006) on this point, as well as the comparison between France and the United States, in Morgan and Prasad (2009).
71.The most detailed history of French debates over progressive income taxation can be found in Marion (1931), though as he was a clear opponent of progressivity his text should be read with that in mind. These debates have also been covered more recently by Delalande (2011), and a brief discussion of them can also be found in Seligman (1911). The unpublished dissertation by Owen (1982) provides a detailed account of the politics of income taxation in France in the early twentieth century.
72.See Gross (1996, p. 125).
73.See Gross (1996).
74.These results are presented in Piketty (2001, p. 237). They derive from Caillaux (1910, pp. 208–209).
75.Wolowski (1872, p. 14).
76.Adolphe Thiers (1871 [1896]). This interpretation is emphasized by Seligman (1911).
77.Ibid., p. 32. All translations from the French in this section are by the authors.
78.See Wolowski (1871).
79.Leon Say, cited in Marion (1931, vol. 6, p. 193).
80.For a description of the sequence of proposals, see Delalande (2011), Seligman (1911), and Marion (1931).
81.Journal Officiel. Chambre. Débats July 10, 1907, p. 1828.
82.Journal Officiel. Chambre. Débats July 11, 1907, p. 1872.
83.For a discussion of this period, see Owen (1982, chapter 4).
84.Parti Socialiste: Section Française de l’Internationale Ouvrière, Programme d’Action du Parti Socialiste, 1919, p. 5.
85.Ibid., p. 14.
86.Journal Officiel. Chambre. Débats April 12, 1920, p. 874.
87.This was the case with Just Haristoy (1918).
88.See Tristram (1999).
89.Journal Officiel. Chambre. Débats April 12, 1920, p. 882.
CHAPTER 7: THE ROLE OF WAR TECHNOLOGY
1.This is an argument emphasized by Andreski (1968).
2.The Politics, Book VI, Chapter 7, as translated by Ernest Barker (1946).
3.The Constitution of the Athenians, as translated by Robin Osborne (2004). It should be acknowledged that not all Greek city-states with large navies had constitutions allowing for broad political participation. See Scheidel (2005) on the case of Corinth. See Ober (2015) for a broader discussion of the political economy of classical Greece.
4.See White (1962).
5.See Roland (2003) for a review of the debate and an assessment.
6.See Hui (2005).
7.See Lewis (2000).
8.Hui (2005).
9.This was the date of the creation of the Stockton and Darlington Railway.
10.See in particular van Creveld (1977, 1989) on this question, as well as Pratt (1915) for an earlier version of the story.
11.This date is offered by Pratt (1915, p. 9).
12.Onorato, Scheve, and Stasavage (2014). For the period from 1815 to the present we used data from the Correlates of War Project. For the years between 1600 and 1815 we relied on existing estimates produced by military historians for individual countries. For full sources, see Onorato et al. (2014). The thirteen great powers are defined by Levy (1983) and are Austria-Hungary (1600–1918), China (1949–2000), France (1600–2000), Italy (1861–1943), Japan (1905–1945), Netherlands (1609–1713), the Ottoman Empire (1600–1699), Prussia/Germany/West Germany (1740–2000), Russia/Soviet Union (1721–2000), Spain (1600–1808), Sweden (1617–1721), United Kingdom (1600–2000), and the United States (1898–2000).
13.See Posen (1993) and Darden (2013) on the link between literacy and nationalism.
14.See Bogart (2009).
15.See in particular Posen (1993) and Snyder (2000).
16.This result was obtained by including a dummy variable in our military size and military mobilization regressions that took a value of 1 in all years beginning in 1789. The full result can be seen in table 2 of Onorato et al. (2014).
17.See the discussion in Schnapper (1968).
18.See Levi (1997).
19.This evidence is presented in the appendix to Onorato et al. (2014).
20.See Litton (2000, p. 3).
21.See Gillespie (2006, p. 27).
22.See ibid., p. 110.
23.See Murray and Knox (2001).
24.See Onorato et al. (2014) for the sources.
CHAPTER 8: WHY TAXES ON THE RICH DECLINED
1.“Let Us Face the Future: A Declaration of Labour Policy for the Consideration of the Nation,” Labour Party, 1945.
2.Ibid.
3.British Institute of Public Opinion, survey 123, question 13.
4.Conseil National de la Résistance, “Programme du Conseil National de la Résistance,” March 15, 1944.
5.Journal Officiel, October 19, 1944.
6.See de Vries and Hoeniger (1950).
7.On this subject see in particular the contribution by Hughes (1999).
8.See Meidner (1993).
9.As we suggested in chapter 1, this idea is explored in a recent paper by Roel Beetsma, Alex Cukierman, and Massimo Giuliodori (2013), who build on earlier findings by Peacock and Wiseman (1961). It is also closely related to Paul Pierson’s (1994) arguments about how welfare states create constituencies prepared to defend their programs, frustrating attempts at policy retrenchment.
10.This was in particular the case in a speech of Winston Churchill’s that was broadcast on June 5, 1945.
11.See Hughes (1999) on these debates.
12.“Wealthy” here is defined as those individuals who were placed in the highest socioeconomic status category based on the interviewer’s subjective assessment. Gallup polls in this early period had no direct questions about annual income.
13.American Enterprise Institute, “Public Opinion on Taxes: 1937 to Today,” 2012.
14.See in particular the discussion in Martin (2013).
15.The legislative history of this movement has been covered by Theodore Meyer (1956).
16.USAIPO 0492 Question 9.
17.In USAIPO 0366 in 1946 Gallup did also find a majority in favor of an income tax limitation, but that was to limit the total tax take (i.e., the effective rate) to 50 percent of one’s income rather than 25 percent.
18.“The New Poor: A Salary Ceiling Story of Mr. Smith and his Fellow Bank Directors,” Wall Street Journal, November 11, 1942.
19.1979 Conservative Party General Election Manifesto.
20.“This Is the Road: The Conservative and Unionist Party’s Policy, Conservative Party, 1950.”
21.Republican Party Platform of 1980.
22.Republican Party Platform of 1952.
23.See the online appendix for the analyses. In our analyses discussed in the chapter 3 notes and reported in the online appendix exploring the robustness of our estimates of the effect of mass mobilization on war, we controlled for various measures of economic crises (e.g., debt, banking, and inflation crises). Those estimates reveal some evidence that countries experiencing economic crises on average lower their top rates of income tax. However, this result is sensitive to econometric specification, and there is no evidence that the relationship between crises and reductions is stronger after 1970 than before—in fact the absolute values of the point estimates are larger before than after. Thus, even if there is something to the idea that countries respond to economic shocks by lowering top rates—and our evidence is mixed on this point—it is consistent with our conclusion here that there is not persuasive evidence that fears about economic growth account for postwar declines in top rates of income taxation.
24.See Auerbach, Hines, and Slemrod (2007) for a thorough review of issues related to corporate taxation today.
25.See the theoretical review by Keen and Konrad (2013).
26.Devereux et al. (2008). See Kumar and Quinn (2012) who analyze a broader panel of countries, concluding that capital mobility has had relatively little effect on corporate tax rates. Other studies include Hallerberg and Basinger (1998), Hays (2009), Quinn (1997), and Swank and Steinmo (2002).
27.See the evidence compiled by Steffen Ganghof (2006) on this point.
28.See Hines and Summers (2009).
29.Swank and Steinmo (2002) report a negative relationship between capital control liberalization and effective taxes on labor but argue that this is not consistent with the usual globalization hypothesis because the expectation is that capital control liberalization forces governments to shift taxes from internationally mobile capital to generally immobile labor. They interpret the correlation as indicating that globalization has sensitized policymakers to the efficiency costs of high labor taxes. They also report no relationship between capital control liberalization and effective taxes on capital and conclude that, although globalization seems to have played a role in the setting of statutory corporate taxes, there is little evidence that it has played a primary role in the determination of other tax burdens. Their results, especially the null result for effective capital taxation, resonate with our analysis below.
30.See the review by Sorensen (2010).
31.See Quinn (1997). The underlying data come from the International Monetary Fund’s annual report on exchange restrictions. The Quinn index has a minimum of zero and a maximum of 100, and it is available for the period between 1950 and 2011.
32.See the online appendix.
33.See the theoretical discussion in Piketty and Saez (2013).
34.See Kleven, Landais, and Saez (2010), and Kleven, Landais, Saez, and Schultz (2013).
35.We established this using the test for cross-sectional dependence proposed by Pesaran (2004).
36.Conservative Party General Election Manifesto, 1979.
37.Republican Party Platform, 1980.
38.“The Labour Way Is the Better Way,” Labour Party Manifesto, 1979.
39.Labour Party Manifesto, 1983.
40.Democratic Party Platform, 1980.
41.Democratic Party Platform, 1984.
42.Fairness claims can also explain lower taxes on the rich in recent decades in countries that did not mobilize for war. For example, in Sweden, Henrekson and Waldenstrom (forthcoming) provide an account in which the possibility of the very wealthiest avoiding inheritance taxes combined with equal treatment norms to lead to the end of the inheritance tax. They argue that tax avoidance reduced political support for the tax because there was a perception that the wealthiest citizens avoided the tax while the less wealthy paid the tax. This exemplifies a different way in which violating equal treatment norms can undermine support for taxing the rich.
CHAPTER 9: WHAT FUTURE FOR TAXING THE RICH?
1.For just one example, see “Americans’ Views on Inequality and Workers’ Rights,” New York Times, June 3, 2015.
2.“Also, all tax rates refer to marginal rates—this means that all individuals only pay that rate on the portion of their income that falls into that income category.”
3.Figure 9.1 graphs the weighted data. The raw data medians are qualitatively similar though somewhat higher for the highest income category. Given that even this higher raw median is only 35 percent, this difference does not change our conclusion that there is little support for substantially higher rates than current law.
4.As discussed above, a common alternative survey question employed both in the popular press and in academic work asks individuals whether they support increased taxes on individuals with incomes over some threshold—usually $250,000 per year. In the United States over the last several years, a majority of respondents often answered this question in the affirmative. This fact has been interpreted as evidence that citizens want a much more progressive tax system with substantially higher rates than they are getting. Moreover, some survey researchers might prefer this question because our question requires respondents to express a specific quantitative preference, which is a relatively harder task. This is a problematic position for several reasons. First, respondents to this more common question may not know how high taxes already are on top earners. Second, the question is binary and it is not possible to infer whether respondents support slightly higher or much higher taxes on top earners. At least some survey evidence suggests that only small minorities support increased top rates of more than 5 percentage points. This is consistent with our more general point here that there is little evidence that the public prefers much more progressive income tax policies than currently exist. Third, the distribution of responses to our question is not substantively different if the sample is limited to more numerate or educated respondents. Fourth, questions that refer to individuals above a certain threshold paying higher taxes do not distinguish between higher statutory rates and increased effective rates through the reduction of exemptions and other privileges in the tax system. See Ballard-Rosa, Martin, and Scheve (2015) for the details of an alternative analysis that suggests preferred income tax rates in the U.S. public similar to those we report here. This study implements a conjoint experimental design to identify the salient characteristics of public preferences about alternative income tax plans taking into account revenue needs/constraints and reports results for the top rate that are quite similar to the simple survey question discussed here. Overall, it does not seem that in the United States there is strong support for significantly higher income tax rates on high earners despite substantial increases in inequality over the last three decades.
5.The authors did, however, find that providing information about how few people are subject to the estate tax in the United States today was associated with support for a substantially higher marginal rate of estate taxation. See Kuziemko, Norton, Saez, and Stantcheva (2013).
6.Internal Revenue Service, Statistics of Income.
7.Statistics on kindergarten teacher compensation are drawn from the Bureau of Labor Statistics for 2013, http://www.bls.gov/oes/current/oes252012.htm
8.See Landais, Piketty, and Saez (2011).
9.See Bank (1996).
10.See Landais et al. (2011, pp. 50–52).
11.One of the most detailed sources on this issue is the “Who Pays?” report on state level taxation by the Institute on Taxation and Economic Policy. According to this source, averaging across all fifty states, those individuals in the bottom 20 percent of earners pay about 7 percent of their income in sales and excise taxes, whereas those in the top 1 percent of earners pay less than 1 percent of their income in sales and excise taxes. Institute on Taxation and Economic Policy (2013).
12.We can also ask what equal treatment logic implies for proposals to establish a progressive consumption tax in the United States. The idea behind this tax is that individuals would submit a declaration at the end of each year showing how much they had earned and how these earnings had been devoted to either consumption or savings. They would then be liable for a tax on the portion consumed, with higher rates applying to higher levels of consumption. People with a wide range of political views have advocated such a tax (Frank 2011; Caroll and Viard 2013). The idea of a progressive consumption tax is most often justified on grounds of economic efficiency, encouraging saving and limiting the effect on work incentives from something like a progressive income tax. However, independent of efficiency considerations, we can see how the logic of equal treatment can be applied to justify a progressive consumption tax. Instead of trying to justify a progressive rate structure by referring to economic efficiency or ability to pay, it could be justified by the norm of equal treatment. This could be done if the rates on this tax are set so that individuals from different income categories all spend roughly the same fraction of their income on consumption taxes. Under these circumstances, referring to a “progressive” consumption tax would be a misnomer. The new tax would actually be an “equal treatment consumption tax.”