INTRODUCTION

There is much debate about wealth inequality today both in the United States and elsewhere. Should something be done about it, and if so, what? These contemporary concerns are closely related to two research questions. What is the effect of wealth inequality on the emergence and sustainability of democracy? What is the impact of democratic government on wealth inequality? If one were asked to sum up the received wisdom about these questions, it would be that wealth inequality is bad for democracy, and yet democracies are also likely to implement policies that reduce wealth inequality. The simple reason for this is that those with no or little wealth are more numerous and therefore have more votes than the wealthy. This same pattern has already been noted for democratic politics and income inequality; Acemoglu & Robinson (2000, 2006) and Boix (2003) are the most heavily cited proponents of the idea that democracy involves equalizing policies, although these authors have been challenged on multiple fronts, and some of their own recent empirical work suggests a different conclusion.¹ Because the distribution of wealth is almost invariably more unequal than is the distribution of income, this same argument should apply even more forcefully for wealth. Political scientists have spent a great deal of time analyzing the empirical correlates of income inequality, but far less effort has been made in the area of wealth inequality. There is a good reason for this: Wealth inequality data are generally much harder to come by.

A second reason for focusing on wealth inequality is that, even more than disparities in income, it can have “snowballing” properties. For two individuals holding fortunes of different initial values and earning the same rate of return, the person who is initially wealthier will become even more wealthy relative to his or her counterpart unless some other factor intervenes. If one thinks that wealth inequality is bad for democracy, then this simple fact is particularly unsettling. Economists have long been preoccupied with the question of what factors might intervene to produce a stationary distribution of wealth, as opposed to one that becomes ever-increasingly skewed. This work has been surveyed by Benhabib & Bisin (2016), who draw in part on the model of Benhabib et al. (2011). They find that a stationary wealth distribution can emerge strictly as a result of economic conditions independent of government policy. However, government policy in the form of taxation of capital also determines how unequal this stationary distribution is. The role of government taxation of capital and its influence on the wealth distribution has also been emphasized by Piketty (2014). Other economists have presented theoretical models in which rising wealth inequality will prompt those at the top to agree to have some of their wealth taxed away out of the fear that refusing to do so will lead to outright expropriation (Farhi & Werning 2014). In other words, the democratic process may be self-regulating.

In this review, we consider what political science can say about the relationship between democracy and wealth inequality. We first ask whether democracy and wealth equality tend to go hand in hand, finding relatively little evidence for an automatic link between the two. To show this, in the next two sections we use comparisons between the United States and Latin America, cross-regional data on land inequality, and evidence from top wealth shares over the last 200 years. In a fourth section we then consider three different reasons why democratic politics might lead to reduced wealth inequality in some cases but not others.

The first reason is that citizens without wealth may be divided by other social cleavages, such as religion or ethnicity, and this could inhibit the adoption of wealth-equalizing policies. The literature on the political economy of income inequality has emphasized some of the possible cleavages. These can apply with the same force with regard to wealth inequality. As one illustration of this phenomenon, we show how divisions over religion delayed the adoption of universal and compulsory primary education in a number of European states. Over the long run, this had an impact on the wealth distribution.

The second way in which democratic politics has a contingent effect on wealth inequality has to do with citizen beliefs about the fairness of policies that redistribute wealth. It is often assumed that those with less wealth want redistribution, but support also depends on whether citizens think these policies are fair. Whether citizens think wealth-equalizing policies are fair depends on the particular fairness criterion that they themselves hold. We have provided extensive evidence for this elsewhere (Scheve & Stasavage 2016).

A third factor we consider is the possibility of captured democracy. If the wealthy are able to exert disproportionate influence on government despite the principle of one person–one vote, then they may use this influence to block policies that equalize wealth. It is not hard to think of how this might occur; the real question is whether we can trace the importance of capture across countries and over time. Simply observing that wealth inequality is high and then inferring that there must be capture is not a very convincing empirical strategy. We suggest that although the best empirical assessments of capture have considered the US context, it will be difficult to extend these same methods to other political systems. Even the evidence for the United States may be less convincing than commonly believed, as countries with very different opportunities for capture all often fail to adopt wealth-equalizing policies. Establishing metrics for capture that can be used cross-nationally is an important goal for future research.

EXISTING COMPARATIVE EVIDENCE

Wealth inequality tends to evolve slowly over time. Therefore, if we want to understand its determinants, we should consider long-run evidence. One of the most commonly referenced illustrations of the relationship between wealth inequality and democracy involves the contrast between Latin America and the United States over the long run. Initial conditions circa 1500 AD are said to have produced a persistent pattern in Latin America of high inequality combined with either nondemocratic rule or captured democracy. Initial conditions in the United States (or at least the United States north of the Mason-Dixon line) are said to have produced exactly the opposite effect. The inequality evidence supporting this interpretation has been called into question, and in this section we present further new evidence to doubt the received wisdom. In about 1860, levels of wealth inequality in New York, Boston, and Philadelphia matched or exceeded those in Buenos Aires and Rio de Janeiro. In spite of this, the United States continued on its democratic trajectory while Argentina and Brazil did not. We then turn to cross-country evidence on land inequality, the most important form of wealth throughout much of history. Here too we fail to see an automatic link between land inequality and democracy.

Historical Latin America versus the United States

For many scholars, the comparison between Latin America and the United States shows how democracy and wealth equality go hand in hand. Engerman & Sokoloff (1997, 2000) describe how different initial conditions at the time of European conquest helped set Latin America on a route to persistent inequality without democracy, whereas conditions in the United States did just the opposite. Initial conditions in Latin America involved the suitability of land for sugar cultivation by slaves, deposits of minerals that could be worked by forced labor, and large landholdings with abundant local populations able to work them. None of these three conditions prevailed to the north. Initial differences in inequality then fed through into differences in political institutions, as Latin American countries maintained a restricted suffrage whereas Britain's North American colonies, and subsequently the United States, moved more rapidly to establish universal suffrage for white males (see Sokoloff & Zolt 2006). Acemoglu et al. (2002) tell a closely related story involving the persistence of autocracy and inequality, but they focus above all on the initial density of the native population rather than on initial inequality.

The thesis of enduring inequality in Latin America and enduring equality in North America has been subject to challenge by Williamson (2015). He presents data from a variety of sources to argue that there have been notable trends of both increasing and decreasing inequality in Latin America over the past 500 years. Likewise, Arroyo Abad (2013) shows substantial variation in inequality across time and regions in Latin America during the nineteenth century and argues that, far from being persistent, inequality has been sensitive to changes in factor endowments and international trade. Coatsworth (2008) also reviews several forms of evidence suggesting that wealth inequality in Latin America was no higher than in areas to the north in the colonial era.

Some of the most interesting evidence contradicting the idea that Latin America was always more unequal than North America comes from wealth inequality at the urban level. On the basis of probate records, Johnson & Frank (2006) have produced measures of wealth inequality in Buenos Aires and Rio de Janeiro in the 1850s. By their calculations, in Buenos Aires at this time the top 10% of the population held 68% of total private wealth. In Rio de Janeiro the analogous figure was 78%. Comparing their results with then-available estimates for North American cities, they concluded there was no evidence that wealth inequality in Latin America was higher.

Since the publication of Johnson & Frank's article, data have become available in electronic form that allow us to construct more precise estimates of wealth inequality in North American cities. Between 1850 and 1870, the US census asked questions to determine the amount of real property (land) and personal property (wealth other than land) that individuals owned. For three of the largest US cities at the time (New York, Philadelphia, and Boston), we used the census data to construct a measure for 1860 defined as the share of total private wealth held by the top 10% of households. The first extensive study that used census data from this period to look at wealth inequality was by Soltow (1975). Steckel (1994) considered the reliability of the census wealth responses by comparing wealth inequality measures based on these with measures derived from property tax records. He found that top wealth share measures derived from these two sources were very similar.

The results of our investigation suggest that, if anything, the largest cities in the United States were actually more unequal than those in South America (Figure 1). We also asked whether our result changed much if we considered US inequality in terms of adult males rather than households, and it did not. Finally, we considered whether inequality was similar when we restricted our view to all adult males over the age of 50, since Johnson & Frank's estimates are based on estates at death. Again, there was relatively little difference from the statistics reported in Figure 1.

The evidence in Figure 1 strongly suggests that inequality in US cities historically has often been on the same scale as inequality in Latin America. However, there still remains the possibility that Latin America has always had greater wealth inequality than the United States in rural areas, where it would be determined above all by inequality in landholdings. Soltow (1975) established that inequality measures constructed on the basis of US census data between 1850 and 1870 were invariably lower in rural areas than in cities. Frank (2005), in contrast, showed using probate records that in one rural area of southeastern Brazil, wealth inequality was on the same level as that found in Rio de Janeiro, although this is based on only one small area. It is also the case that data from the early twentieth century show significantly lower levels of landholding inequality in the United States than in several South American countries (Frankema 2008). In the next section, we consider the relationship between land inequality and democracy in a broader set of countries.

Land Inequality and Democracy

In many societies throughout history, land has been the most important form of wealth, and it is thought that inequality of landholdings is invariably bad for democracy. Abundant evidence from individual countries shows that this can be the case. To take one important example, Ziblatt (2008, 2009) has shown how inequality of landholding helped thwart and corrupt the development of democracy in Germany. But what about the broader comparative evidence? Do democratic countries enact policies that reduce land inequality more than their authoritarian counterparts do? Likewise, are countries with more unequal landholding less likely to sustain democracy?

The answer to the first question seems to be a clear “no.” Many of the most significant twentieth-century land reforms were implemented by autocratic governments in places such as China, Korea, and Taiwan. In the most comprehensive study to date on land reform in Latin America, Albertus (2015) has found that democratic governments have engaged in far less land redistribution than have their authoritarian counterparts. He argues that whether land reform occurs depends less on whether a democratic or authoritarian regime is in place, and more on whether a split emerges between a country's landholding elite and its ruling elite.

The Latin American context provides a very useful environment to consider land reform and democracy because so many countries in the region have oscillated between democratic and authoritarian rule. Albertus (2015) has constructed a variable that, for each country in the region for each year since 1930, measures the percentage of total land that was redistributed through land reform. Using his data, one can compare the average level of redistribution in democracies and autocracies using the definition of democracy first proposed by Przeworski et al. (2000), the presence of elections in which multiple political parties compete. Based on Albertus's data, in autocracies the average level of redistribution was about 1% of total land per year. In democracies the level of redistribution was only about a third of this (0.38%). The figures for both groups are low because in most years in any country there was no redistribution. However, they still point to a clear distinction between democracies and autocracies, and it is one that goes against the received wisdom.

The next question to ask is whether societies with unequal landholdings are less likely to sustain democracy. To address this, we used cross-national and cross-regional data compiled by Frankema (2008, 2010), drawn from agricultural censuses submitted by individual countries to the United Nations’ Food and Agriculture Organization and, prior to that, to the International Institute of Agriculture. Frankema has constructed a data set of Gini coefficients measuring inequality in landholdings, which covers a broad swath of countries (up to 82, depending on the time period) over a relatively long time span. For some countries in Frankema's data set, it is possible to consider land inequality over the entire twentieth century. In most cases, however, coverage begins around 1950, and so we focus on this more recent period. We adopted a very simple empirical strategy, dividing the data into three periods (1950–1969, 1970–1989, and 1990–2009) and regressing the proportion of years during each two-decade period that a country was a democracy on inequality in landholding at the beginning of that period. As a measure of democracy, we again used Przeworski et al.’s (2000) competitive-elections definition.

Our findings for the 1950–1969 period parallel those found in an early study by Russett (1964); there is a negative and statistically significant correlation between the land Gini and democracy. If we regress democracy on the Gini using ordinary least squares (OLS) and interpret this relationship causally, a one-point increase in the land Gini is predicted to decrease the democracy variable by 0.01; that is, it decreases the percentage of years that are democratic by one percentage point. This would be a sizable effect. We do fail though to see a statistically significant correlation after adding region dummies to the regression. This is because the negative correlation for 1950–1969 is driven by the contrast between Western Europe and North America on the one hand and the rest of the globe on the other.

Our findings for the 1970–1989 and 1990–2009 periods fail to provide any support for the idea that higher land inequality means less democracy. For the 1970–1989 period, the correlation between the land Gini and our democracy variable is actually positive, though not statistically significant, and for the 1990–2009 period the correlation is positive and statistically significant. However, the correlation does not remain statistically significant when we add a set of region dummies to an OLS regression of democracy on inequality.

An arguably more convincing empirical strategy would be a difference-in-differences approach looking only at within-country variation instead of cross-sectional variation. With the caveat that land inequality is extremely persistent over time (the correlation between land inequality circa 1950 and land inequality circa 1990 is 0.90), estimation of a difference-in-differences specification suggested no correlation whatsoever between land inequality and democracy. We also considered whether our null findings for the 1970–1989 and 1990–2009 periods and in the difference-in-differences specification were attributable to the fact that we had adopted a fairly minimalist definition of democracy based on the simple presence of competitive elections. To evaluate this possibility, we also examined the correlation between land inequality and indices of freedom of expression and the free character of elections that have been constructed by the Varieties of Democracy project. Our results paralleled those for our main democracy measure; there was a negative correlation between land inequality and democracy between 1950 and 1969 but not in other specifications. We should emphasize that the above findings are based on a feasible identification strategy rather than a particularly credible one. It is still surprising how little support there is for the idea that higher land inequality makes it more difficult for democracies to emerge and sustain themselves. In the next section, we consider this same question using data on top wealth shares.

EVIDENCE FROM TOP WEALTH SHARES

Several important research projects have recently produced new data or compiled existing data on top wealth shares in the nineteenth and twentieth centuries. Roine & Waldenstrom (2015) review the progress made in measuring historical wealth inequality and compile data for 10 countries. We use these data along with wealth inequality measures for Ireland produced by Turner (2010). The countries that we consider are Australia, Denmark, Finland, France, Ireland, the Netherlands, Norway, Sweden, Switzerland, the United Kingdom, and the United States. We focus our attention on the proportion of wealth owned by the top 1%. As Roine & Waldenstrom emphasize, there are important differences across countries in the methodologies used for constructing this measure. For some countries the data are based on wealth taxes, and the wealth-holding unit is the household, whereas for other countries the data derives from estate taxes, and the wealth-holding unit is the individual. These and other differences mean that comparisons across countries need to be made with great caution and that our primary emphasis is to describe changes over time within countries. Finally, there are considerably fewer data available for the nineteenth than the twentieth century, which affects how we approach the evidence.

Figure 2 presents the variable Top 1% Wealth Share for the 11 countries since the beginning of the nineteenth century. The striking feature of these data is that wealth concentration in almost all countries was high in the nineteenth century and the beginning of the twentieth century and declined substantially until the 1970s and 1980s.

Because we know that all of these countries eventually became full democracies, it is certainly possible, consistent with conventional wisdom, either that democracy was hard to sustain while wealth inequality was high or that democracy, once adopted, caused wealth inequality to drop. But a great deal else was happening during this period, and so the general trend toward democracy and wealth equalization may be spurious. To evaluate these relationships more carefully, we first consider a specific case, France, for which we have both exceptionally good wealth inequality data over the long run and significant variation in political regime type.

Democracy and Top Wealth Shares in France

The experience of France during the nineteenth and early twentieth centuries provides evidence contrary to the ideas that wealth inequality slows democratic development and especially that democracy leads to greater equality. Figure 3 provides a basic description of the pattern of democracy and wealth inequality as measured by top 1% wealth share.

Despite the upheaval of the Revolution, including the expansion of the franchise, the percentage of national wealth held by the top 1% was 46% in 1810. This level of wealth concentration is high by almost any metric and reflects substantial wealth inequality. Following the Restoration of 1814 and under the July Monarchy, France adopted a restrictive suffrage regime known as the régime censitaire. Only individuals who paid a sufficiently large sum of direct taxes were eligible to vote. Top 1% wealth shares remained high and relatively unchanged in this period.

Did wealth inequality slow France's transition to stable democracy in the nineteenth century? It is clear that political actors on the left and right expected that expansion of the franchise would bring greater redistribution. This certainly gave workers a reason to advocate for a republic and elites a reason to resist. Reflecting these expectations, in his 1833 Lettre aux Prolétaires the worker organizer Albert Laponneraye wrote, “Be republican because under the republic, you will no longer have to pay any taxes, and the rich alone will pay them.” Rosanvallon (1992) argues that Laponneraye's comments reflect a more general sentiment at the time; universal suffrage was advocated because of the effects it was expected to produce, rather than for abstract philosophical reasons. France's experience with democracy during the Second Republic was short lived, and this might support the idea that wealth inequality impeded democracy. However, high wealth inequality did not prevent a durable transition to stable democracy after 1870.

A second question is how France's somewhat sudden and dramatic transitions to democracy influenced wealth inequality. The most important evidence in Figure 3 is that wealth inequality did not decrease during the first two decades of the Third Republic and then significantly increased in the subsequent two decades, with the top 1% wealth share reaching 60% in 1910. A 40-year period of democracy and high inequality seems inconsistent with the idea that democracy and wealth equality necessarily go together.

Did a High Top Wealth Share Impede Democracy?

If we want to use the top wealth shares data more systematically to ask whether wealth inequality impedes democratic development, we run into the problem of sparse data for the nineteenth and early twentieth century. Consider what we know about top wealth shares around 1800. Denmark's top 1% wealth share in 1789 was 56%, France's in 1810 was 46%, Sweden's in 1800 was 48%, the United Kingdom's in 1810 was 55%, and the United States’ in 1774 was 28%. If we code democracy as the presence of competitive elections and at least 50% of adult males eligible to vote, only the United States would be considered a democracy at this time. It is tempting to support the received wisdom by noting that the United States also had the lowest top 1% wealth share at this time. However, we would be relying on the US observation alone for this conclusion, and other evidence does not support the received wisdom. Among the remaining four countries, France and the United Kingdom were the first two to transition to stable democracy as defined above (1870 and 1885, respectively) while having the lowest and nearly the highest top 1% wealth shares at the beginning of the nineteenth century. Denmark's and Sweden's transitions were timed closely (1902 and 1911, respectively) although they had very significant differences in initial wealth inequality. More generally, for 1900, Denmark and Sweden are the only two independent countries for which we have top wealth shares data and that had not yet transitioned to democracy, but their top wealth shares were below all the independent democracies for which we have data except the United States. Finally and perhaps most importantly, many of the countries in our sample experienced many decades of democracy and high wealth inequality before top wealth shares declined later in the twentieth century. In summary, these data fail to support the idea that wealth inequality impedes democracy.

Did Democracy Decrease Top Wealth Shares?

Top wealth shares are largely determined by two sets of factors. First, they are influenced by levels of income inequality. The more concentrated are earnings in a society, the higher will be levels of wealth inequality, although it takes many years for changes in income inequality to have a substantial effect on the distribution of wealth. It is well known that wealth distributions are more concentrated than income distributions, and so this mechanism gets us only so far. Second, top wealth shares are influenced by the process of wealth accumulation. This includes differential savings rates across income groups, the return to capital and its variability, and the ability of families to pass on wealth to future generations (Jones 2015, Benhabib & Bisin 2016).

Democracy might matter for top wealth shares if the presence of democratic institutions leads countries to adopt policies that make either the earnings process or the wealth accumulation process more equal in a society. Land redistributions, discussed above, are a classic wealth-equalizing policy, but there are many other possibilities. Consider, for example, the provision of public education. Public education can expand the supply of human capital and reduce the skill premium in the labor market. Thus, if the expansion of the franchise and democratic political competition leads countries to expand public education, it may reduce top wealth shares through its effect on earnings and wealth accumulation. Capital taxation provides another mechanism for how democracy may influence wealth inequality. The taxation of capital income and of inheritance helps to equalize wealth because it reduces the after-tax return on capital and the ability of one generation to transfer its wealth to the next. Another potential set of wealth-equalizing policies are those influencing the real return to capital and the relative return of different types of capital likely to be differentially owned by the wealthy and non-wealthy. Government policy can influence the real return to capital most directly through inflation. High inflation, certainly if it is not fully anticipated, tends to lower real returns to capital. If democracies are more likely to have high inflation, e.g., because politicians weigh unemployment more heavily when facing short-term trade-offs between unemployment and inflation, then democracy may lead to a more equal distribution of wealth through its effect on the real return to capital. Another possibility takes more seriously the types of assets that the poor and middle class are likely to own if they hold wealth at all. The relative performance of housing assets—more likely to make up the bulk of wealth held by the poor and middle classes—versus financial assets may be important, as would any policies that subsidize the returns of these asset classes. Government policies that expand and subsidize housing or heavily regulate finance may significantly reduce top wealth shares.

For an initial evaluation of the impact of democracy on top wealth shares through these and other wealth-equalizing policies, our analysis needs to control for the two most important economic determinants of wealth inequality: the real returns on capital and the riskiness of these returns. The intuitive reason why capital income risk matters is that it magnifies wealth inequality. In a world where returns on investments are stochastic, some wealth holders turn out ex post to have made good investments whereas others turn out to have made bad ones. We constructed the variable r equal to the nominal yield on 10-year government debt minus the inflation rate. This follows the approach used by Acemoglu & Robinson (2015). This is a very imperfect proxy for the true return on capital, so, for example, our measure would be a poor proxy in the current period of extremely low government bond yields. However, by using it, we recover plausible estimates of the effect of r on wealth inequality. The yield and price data are from the GFDatabase version 2.0. We linearly imputed missing data for this series. To measure the variance of capital returns, we calculated the standard deviation of the variable r over five-year periods, constructing the variable var(r). To measure democratic political institutions, we constructed the variable Competitive Elections, which is set equal to one if the legislature is elected in free multi-party elections, if the executive is directly or indirectly elected in popular elections and is responsible either directly to voters or to a legislature elected according to the first condition, and finally if at least 50% of adult males have the right to vote. This definition and data are from Boix et al. (2012). The definition modifies the one used by Przeworski et al. (2000) to a context where the suffrage may be restricted. We also conducted parallel analyses yielding similar results using a dichotomous variable set equal to one if a country has reached universal male suffrage and zero otherwise. To evaluate whether our results were sensitive to adopting minimalist measures of democracy, we estimated these same specifications using indices of freedom of expression and the free character of elections that have been constructed by the Varieties of Democracy project and found results similar to those reported below for competitive elections.

Although our data are annual, we do not expect democracy to have an immediate impact on wealth inequality and do not have strong expectations about how long it will take to have an effect. Consequently, our analysis uses five-year averages. Even with these five-year periods, it is clear that if democracy influences wealth inequality through its effect on income inequality via policies like mass education, it should operate with a substantial lag. We show estimates that lag democracy one period and six periods. A lag of six periods is used with the idea that during these 30 years, citizens would have received higher levels of education, started to earn higher salaries, and had sufficient time to accumulate some personal wealth. Our approach in this analysis requires wealth inequality data over time for each of our cases. We, therefore, restrict our analysis to the twentieth century, for which we have reasonably good data for all 11 countries. We estimate the following econometric model:

1.

where i indexes each country and t indexes the five-year time period; W is the wealth inequality measure Top 1% Wealth Share with missing values imputed by linear interpolation; r is the real return on capital; var(r) is the within-period within-country standard deviation in r; D is our measure of democracy; T is a time trend; α, ρ, and β are parameters to be estimated; η_i are country fixed-effects parameters also to be estimated; θ_t are period fixed-effects parameters; γ_i are country-specific time trends; and ϵ_it is the error term.

The inclusion of a lagged dependent variable in this model is necessary because top wealth shares will adjust only gradually. The country fixed effects allow us to focus on within-country variation over time. This is especially desirable in this setting because of the differences in measurement and methodology in the construction of the Top 1% Wealth Share. The fixed effects also control for time-constant factors that may be correlated with both our policy measures and wealth inequality. Although the presence of a lagged dependent variable and country fixed effects can generate bias, that bias is decreasing as the length of the time series increases and we have long time series, making this bias a minimal concern. The period fixed effects are included to control for common shocks, such as changes in the global economy, technology, and political events. The country-specific time trends are necessary because wealth inequality trended down during the twentieth century, and any trending variable will appear to be correlated with it. This problem is exacerbated by the fact that we have linearly interpolated the Top 1% Wealth Share variable. As suggested above, it is possible that r and var(r) are a consequence of public policies and therefore post treatment in our evaluation of the impact of democracy, so we also estimate specifications omitting both variables. We present the OLS estimates of this model and report robust standard errors.

Our analysis yields informative estimates of the partial correlations between democracy and the top 1% wealth share. The estimates have a causal interpretation under the usual assumptions of a generalized difference-in-differences design. Both reverse causality and time-varying unobservables correlated with democracy and wealth inequality are concerns. Most importantly, we cannot eliminate the possibility that a country's regime choice is in part a function of wealth inequality, and this should be kept in mind in interpreting our estimates.

Table 1 reports the results of our main evaluation of the impact of democracy on wealth inequality. Across all specifications, the coefficient on Competitive Elections is small, positive rather than negative, and statistically insignificant. In unreported results, we find that the lack of a negative and significant estimate for Competitive Elections is evident across a wide variety of alternative specifications. There is no evidence that democracy decreases top 1% wealth shares. It is worth noting that our estimates for both r and var(r) are positive, as predicted, and marginally statistically significant in our baseline specification.

Table 1

Democracy and wealth inequality, 1900–2010

A natural question to ask is whether the null result that we observe is due to policy makers not pursuing wealth-equalizing policies or to the ineffectiveness of policies thought to be wealth equalizing. Scheve & Stasavage (2016) provide evidence that a number of policies, such as capital income and inheritance taxation, are indeed correlated with reduced top wealth shares, but at the same time wealth is not taxed more heavily in democracies. Other recent studies have also concluded that democracy does not cause important wealth-equalizing policies to be adopted; Mares & Queralt (2015) look at the income tax and Aghion et al. (2012) at education. In the next section, we ask why democracies might not implement wealth-equalizing policies.

EXPLAINING WEALTH INEQUALITY TRENDS

It is clear from the different contexts we considered in the previous section that the presence or absence of electoral democracy doesn't do a very good job at explaining patterns of wealth inequality. In this section, we consider three explanations for this: citizens might be divided across cleavages other than wealth, citizens might not agree that wealth-equalizing policies are fair, and finally policy might be captured by the wealthy.

Divided Societies

Political competition is often presented as a contest where policy preferences can be laid out on a single dimension based on personal income. The same can be done with wealth. Lipset & Rokkan (1967) argued that nineteenth-century political competition in European states took place across a number of different cleavages, such as those involving religious divisions, and between center and periphery. But as time wore on, democratic politics in Europe hinged increasingly along a single dimension—the “worker–owner” cleavage. One problem with using this as a general claim about the politics of wealth inequality is that in many countries outside of Europe, the worker–owner cleavage never became dominant. Also, even within Europe other cleavages have continued to shape political competition. Lipset & Rokkan (1967) readily acknowledged this fact.

When a society is divided over cleavages other than wealth, this may inhibit the adoption of wealth-equalizing policies. If a cleavage other than wealth is most salient, then voting intentions will be determined by, above all, party positions on this second cleavage. This holds as long as voters must choose between candidates who offer a bundle of policies (e.g., on taxation and social issues) instead of selecting policies separately. Roemer (1998) has explored the general conditions under which an effect like this can emerge. Shayo (2009) has explored how social psychological theories about identity can also explain this phenomenon.

A second possibility is that even if voters all support a wealth-equalizing policy, they may disagree over how to implement it. Consider, for example, the provision of greater access to education so that individuals can earn better incomes and accumulate wealth. Voters might agree with such a policy, but they might disagree on a state-versus-free-market dimension over how to implement it. Should the state provide subsidies to make access to private schools affordable, or should it instead increase public provision?

Divisions in nineteenth-century Europe over how best to provide primary education show how a second social cleavage can delay the adoption of wealth-equalizing policies. In many European countries at this time there were prominent conflicts over religion, and in particular over whether there should be a single state religion, multiple religions recognized, or a secular model. Some time ago, Bendix & Rokkan (1962) argued that where this conflict was sorted out more quickly, states moved earlier to adopt universal compulsory education. They suggested that this happened earliest in the Scandinavian countries, where support emerged for a Lutheran state church that was given an important role in educational instruction. In countries such as France, to take a counterexample, the establishment of universal instruction was delayed by conflict over the proper role of the Catholic church. It is possible to use existing data to provide a statistical foundation for the original insight of Bendix & Rokkan (1962). On average, countries with a state religion established universal compulsory education some 30 years before those without a state religion, and this relationship is evident even if one controls for whether the country was predominantly Catholic or not. One can also use top wealth shares data to show that the timing of educational expansions eventually fed through into lower levels of wealth inequality.

Fairness

Another reason we might not observe wealth-equalizing policies in a democracy is if voters’ preferences are informed by fairness beliefs. Fairness beliefs may dictate that wealth should be equalized in some contexts but not others. This idea has been most extensively studied with respect to income inequality, but the implications can also carry over to wealth. Piketty (1995) examined how individuals can draw from their personal and family experience to infer (potentially with error) whether income is acquired as a result of luck or effort, and he suggested that this would guide voter preferences. Bénabou & Tirole (2006) extended this to a setting where individuals can manipulate their own beliefs (or those of their children) to maintain effort. Important earlier work in political science by Lane (1959) and Hochschild (1986) considered how individuals form beliefs about whether the distribution of resources in society is fair and how these beliefs influence their policy preferences. There is a large and growing body of evidence from surveys and laboratory and survey experiments suggesting that fairness matters for individual policy preferences (see Ballard-Rosa et al. 2017, Cavaille & Trump 2015, Durante et al. 2014, Fisman et al. 2014, Fong 2001, McCall 2013).

In our study Taxing the Rich (Scheve & Stasavage 2016), we demonstrated how citizens’ fairness perceptions influence a government's decisions on how heavily to tax its wealthiest citizens. As a result of that influence, fairness perceptions, and in particular changing fairness perceptions, have influenced the distribution of wealth. At the heart of our argument is the idea, supported by empirical evidence, that fairness motivates choices but is a contested notion with different individuals applying different standards. In the realm of taxation, one fairness standard is equal treatment, the simple idea that all should pay the same tax rate just as all have one vote in a democracy. An alternative fairness standard is that of ability to pay, the idea that those with more should pay a higher tax rate because they can better afford it. Support for ability to pay points in the direction of wealth-equalizing policies whereas the flat tax rates implied by equal treatment will have a lesser impact on the distribution of wealth. On average, if ability-to-pay arguments carry the day, then we should logically observe that as the wealthy get wealthier compared to everyone else, top rates of inheritance taxation should go up. Across our 11 countries for which top wealth shares data are available, we found no evidence that this was the case.

Although we found in our work that ability-to-pay arguments alone rarely carry the day, there are other fairness arguments that have had a large impact on taxation of the wealthy. We call these “compensatory arguments” because they involve the idea that the wealthy should be taxed at a higher rate than everyone else if this is needed to compensate for other privileges that they have received from the state. This is in the same spirit as the work we cited above; people are more motivated to tax the wealthy not only when they think inequality is high but also when they are convinced it is fundamentally unfair. Compensatory arguments have come in different forms, but over the past 200 years the most powerful arguments of this sort have involved taxing the wealthy in order to equalize war sacrifice. If the wealthy do not have their labor conscripted for the war effort, for example because they are too old, then they should have their wealth conscripted. The conscription of wealth, and the fairness argument underlying it, explains much of the variation in top income and inheritance tax rates over the course of the twentieth century.

Captured Democracy

Aside from the effect of multiple social cleavages and different fairness beliefs, a third reason why democracy and equality might not go hand in hand is if democracy is captured by the wealthy. It is easy to think of how democratic policy can be captured by the rich, but the real question is how we might assess the degree of capture across countries and over time. Without a well-specified way of testing the theory, there is an inevitable tendency to assume that if things have stayed unequal or become unequal in a democracy then capture must be taking place. As Dahl (1958) noted, “the hypothesis has one very great advantage over many alternative explanations: It can be cast in a form that makes it virtually impossible to disprove.” The principal target of Dahl's critique was Mills (1956) and his concept of a “power elite.” The importance of making the study of capture comparative is further evident in the tendency for single-country studies to identify the importance of capture in a particular policy area but fail to address the fact that countries with very different opportunities for capture adopted or failed to adopt the same wealth-equalizing policies. In this section, we consider four different ways in which scholars have thought about captured democracy.

The first and most direct approach to thinking about captured democracy is to suggest that in a democracy power is proportional to wealth. For authors such as Winters & Page (2009), this means that the United States today functions like an oligarchy precisely because wealth is distributed so unequally. Yet there is an immediate problem with this approach; in just about all countries, wealth is always distributed unequally. If we follow the recent study by Saez & Zucman (2016) [though see Kopczuk (2015) on whether their measure is appropriate], in 2012 the top 0.1% of households held 22% of total household wealth. That's an extraordinary figure. However, even in the 1950s the top 0.1% held about 10% of total household wealth. That is still a very unequal distribution of resources in society in absolute terms. So should we also say that the United States was an oligarchy in the 1950s? What this literature has not yet systematically studied is whether differences across countries and over time in the concentration of wealth—and thus, in this framework, the extent of capture—are correlated with the ability of the wealthy to obtain their political objectives.

A second approach to captured democracy is to suggest not that wealth inequality in general generates capture, but instead that wealth inequality in the context of a market economy generates capture. This is the phenomenon that Lindblom (1982) referred to as “the market as prison.” Likewise, Przeworski & Wallerstein (1988) suggest how the state can be “structurally dependent” on owners of capital because without their investments the economy will not flourish. One version of the market-as-prison approach simply invokes this as a static condition in all market economies, but this is not particularly helpful for explaining variation in inequality within the set of countries with market economies. A second, and more useful, way to think about the problem is to look for conditions that would affect the leverage of those owning capital. Financial capital might have more clout than would landholdings, precisely because it is more mobile. Capital mobility also depends on whether there are capital controls in place. However, we actually found no evidence that countries with capital controls were more likely to maintain higher top rates of inheritance taxation and personal income taxation (Scheve & Stasavage 2016).

A third approach to investigating captured democracy is to measure the extent to which the behavior of legislators is most highly correlated with that of their high-income constituents. Using this approach, Bartels (2008) provided initial evidence for a bias of this sort. Gilens (2012) then provided more extensive evidence that pointed in exactly the same direction. This approach provides a useful way of thinking about captured democracy. It does not, however, tell us why capture occurs, and it is possible that both politicians and constituents are reacting to other political or economic stimuli, making the relationship spurious. Nonetheless, the correlation may be a useful indicator of when and where capture is occurring. These findings can then be used to try to explain variation in the degree of capture. Unfortunately, the approach pioneered by Bartels and Gilens is likely to remain most useful for the United States and less so for many other countries. The key reason is that in the United States it is much more common for legislators to vote independently and against their party compared with systems elsewhere. It is this variation at the level of individual legislators that makes this strategy most useful.

A fourth approach is to consider explicitly how those with wealth are able to use their resources to capture policy making. This has been considered in a general theoretical setting by Acemoglu & Robinson (2008). Their key insight is that if autocracy is considered to be a regime where those with wealth govern, then if there is a shift to democracy, members of the elite will have incentives to invest more of their resources in sources of de facto power, which could include lobbying, bribery, resources for violence, or appointment powers. The key questions are when and where members of a rich elite find it easier to transfer their resources into these forms of de facto power. Bonica et al. (2013) have provided an insightful review of the way that expenditures on lobbying and on political campaigns have influenced policy in the United States. The United States is an ideal environment for studying these phenomena because lobbying and private contributions to campaigns are legal but are subject to strict disclosure requirements. The question for the future is what techniques one might use to assess the sources of capture in countries where the role of money in politics is real but substantially less transparent. Promising strategies appear in studies by Eggers & Hainmueller (2009), Truex (2014), and Faccio (2006) that evaluate the incidence and financial value of political connections across different settings.

CONCLUSION

Wealth inequality is a subject of great discussion today, and political scientists ought to be part of this debate. There is little evidence of an automatic link between democracy and wealth inequality. Democracy has many virtues, but it does not necessarily put societies on a path to greater wealth equality. This is not because public policy does not make a difference but rather because democracies do not necessarily implement wealth-equalizing policies. In societies where people are divided along cleavages other than wealth, it is less likely that wealth-equalizing policies will be adopted. The effect of democratic politics on wealth also depends fundamentally on beliefs held by individual voters about fairness. Voters are most likely to support redistribution or policies fostering equal opportunities not just because wealth is unequal but also because this situation arose for reasons deemed unfair. Finally, it is also possible that democratic politics can be captured by those with wealth. An important avenue for future research is to consider when, why, and how this problem of capture is more severe in practice.

disclosure statement

The authors are not aware of any affiliations, memberships, funding, or financial holdings that might be perceived as affecting the objectivity of this review.