Inequality and Preferences for Redistribution: A Comparison Between States and Across Time in the US

Virtually all proposed alternatives to reduce income inequality presume active government involvement. In democratic societies, what governments can do is constrained or enabled to a great extent by public demand and approval. Research shows that public demand for different forms of government-led strategies to reduce inequality, namely, redistributive policies, varies across societies depending, among other things, on the level of inequality itself. But although the nature of the relationship between inequality and demand for redistribution has been under scrutiny for many years, inconclusive or conflicting results abound in the literature and the issue remains largely unresolved. Insofar as constituents’ preferences for redistribution affect the kind and scope of government intervention, an examination of this classic question about inequality and public opinion remains a crucial task for the broad agenda of inequality reduction. Taking a fresh methodological perspective, this paper aims to provide new insights in this direction.      

In this paper I look at the relationship between local level of inequality and preferences for redistribution across states in the United States. In order to do this, I carry out two different procedures for two separate modeling purposes. First I map the distribution of both preferences and income inequality across the US geography. Second, I decompose the variation in those preferences between states and within states across time. In the following paragraphs I provide the motivation for each procedure as well as details about the data and methods that I utilize.

With respect to the first procedure, research shows that while all states have experienced an increase in income inequality since the 1980s, there is also increasing cross-state dispersion in the magnitude of inequality. This is true for both standard-of-living measures of income inequality and measures constructed from administrative tax data (Fisher, Thompson and Smeeding, 2015). Comparisons between states using these measures are common in other research and no further explanation of this first step is needed here.

The available data for preferences for redistribution comes from nationally representative public opinion surveys (mainly, the General Social Survey and the American National Elections Study). As such, it is not representative of smaller units. Appending various opinion polls, I use multilevel regression with post-stratification (MRP) to correct for unequal sampling probabilities and estimate survey response distributions at the level of states. This technique has proven to produce reliable estimates (Gelman and Little, 1997) and has been applied to studies using pre-election polls data (Park et al., 2004) and public opinion about social policy (Lax and Phillips, 2012). At the same time, for validation purposes, I use restricted geocoded data from the General Social Survey to have a benchmark against which to compare the MRP state estimates. This first procedure, thus, yields a distribution of the geographic variation in both inequality and public opinion favorable to redistributive policies.

The second procedure merits some additional motivation. Until recently, researchers had studied the relationship between inequality and preferences for redistribution taking a cross-national, cross-sectional perspective or a longitudinal perspective. The limitation of cross-sectional research is that comparing between countries can be misleading when, as is often the case, countries differ in other respects than inequality. The relationship between inequality and demand for redistribution might be confounded by institutional differences that are related to both the level of inequality, “culture”, or the logic of solidarity adopted by the public. From a methodological point of view, results of studies of this kind are sensitive to the controls used in the analyses and suffer from the problem of having only a few degrees of freedom (and many controls for a small number of observations). On the other hand, studies that take a longitudinal perspective are often better suited to look at the inequality-demand for redistribution relationship overtime but, as they are usually based on aggregate-level data, they can’t test for the implications of this relationship at the individual level. Also, longitudinal studies often do not adequately control for compositional effects from the individual level.

A recent paper published in the Socio-Economic Review (Schmidt-Catran, 2014) proposed the use of a multilevel hybrid model to take advantage of individual- and country-level data structures and allow for the decomposition of country-level effects into their between (cross-sectional) and within (longitudinal) components, while simultaneously controlling for compositional effects from the individual level. More specifically, using data from the European Social Survey, the author proposed to carry out a comparison via the simultaneous estimation of cross-sectional and longitudinal effects. For this paper, I construct an analogous dataset to carry out a similar comparison between states in the US and across time.

Using the geographic distribution of inequality and the modeled distribution of preferences, and applying the hierarchical modeling technique described above, I propose an empirical and analytical exercise that advances our understanding of the relationship between inequality and public opinion in general, but also at a level that constitutes an important site of political action. While in many of the countries that constitute the objects of cross-national research major national policy changes are unlikely, in the US, states are important sources of political change and political action. State administrations adapt, implement and innovate existing policy, and are well-positioned to act to reduce existing inequality. The study of public opinion allows us to test the propositions of rational-choice models that predict the direction of the preferences for redistribution under changing conditions of inequality, but the analysis of this issue in a comparative perspective, specifically, between states and across time sheds new light on the potential of inequality reduction actions and their relation with other aspects that systematically vary across states (crucially race and poverty). In sum, this paper contributes methodologically to the debate about the direction of redistributive preferences under changing circumstances of inequality; and substantively, it contributes to our thinking about the support for actions to reduce inequality in the United States.