Posc Question

PL18CH02-Erikson ARI 2 April 2015 7:12

Income Inequality and Policy Responsiveness Robert S. Erikson Department of Political Science, Columbia University, New York, NY 10027; email: [email protected]

Annu. Rev. Polit. Sci. 2015. 18:11–29

First published online as a Review in Advance on March 23, 2015

The Annual Review of Political Science is online at polisci.annualreviews.org

This article’s doi: 10.1146/annurev-polisci-020614-094706

Copyright c© 2015 by Annual Reviews. All rights reserved

Keywords

public opinion, representation, inequality, voting participation

Abstract

The growing concern about economic inequality leads to a similar concern about political inequality. This article explores the seeming contradiction between the literature pointing to inequality in political representation in the United States and the literature showing that public policy does tend to represent public opinion in general. Low-income voters are much less likely to vote or to be politically knowledgeable than high-income voters, which limits their influence and creates an upper-income bias to effective public opinion. Considerable research suggests that low-income voters’ opinions count for even less than would be implied by their low participation rate, a matter that should continue to be the subject of research. Seemingly contrary to any upper-income bias to policy making, major legislation usually moves policy in the direction favored by low-income voters (e.g., redistribution, government programs). Upper-income voters and interest groups, however, are able to slow the pace of liberal change.

11

A nn

u. R

ev . P

ol it.

S ci

. 2 01

5. 18

:1 1-

29 . D

ow nl

oa de

d fr

om w

w w

.a nn

ua lr

ev ie

w s.

or g

A cc

es s

pr ov

id ed

b y

U ni

ve rs

ity o

f C

al if

or ni

a -

R iv

er si

de o

n 05

/2 8/

19 . F

or p

er so

na l u

se o

nl y.

PL18CH02-Erikson ARI 2 April 2015 7:12

INTRODUCTION

Income inequality has become a central concern in the United States, as evidence mounts that the rich get richer at the expense of the middle class and, especially, the poor. As wealth becomes increasingly concentrated among a few, economic circumstances for ordinary citizens have become stagnant or declined. Economic mobility in the United States has fallen behind other developed countries, in defiance of the American myth of economic opportunity (for reviews, see Corak 2013, Piketty & Saez 2014). This recognition of economic disparity directs our attention to the possibility of a companion inequality of political representation.

Presumably, more people would gain than lose from economic redistribution or increased economic mobility. Yet we see few signs that politicians are seriously addressing income equality or economic mobility. Are polarization and gridlock responsible (Bonica et al. 2013)? Is the public not interested? Economists’ formal models of the electoral process (e.g., Meltzer & Richards 1981) predict that the median voter would vote for the greatest amount of redistribution when the division of pretax wealth is at its most unequal. So what, if anything, has gone “wrong” with American politics?

Social scientists have taken the question of political inequality to heart. A series of edited volumes from the Russell Sage Foundation address the problem directly (Neckerman 2004, Jacobs & Skocpol 2005, Enns & Wlezien 2011a). Leighley & Nagler’s Who Votes Now? (2014) documents the sharp gradient by which voting participation increases with income. Schlozman et al.’s (2012) The Unheavenly Chorus shows that the socioeconomic gradient extends beyond voting to all forms of political participation. Bartels’ (2008) Unequal Democracy consolidates the case that the rich benefit politically at the expense of the poor. Amassing a wealth of evidence, Gilens’ (2012) Affluence and Influence argues that, to the extent public opinion determines policy, it is the views of the wealthy—and perhaps the very wealthy—that really matter. One point is clear: The rich participate more than the poor and enjoy greater influence on political outcomes.

Some critics (Winters & Page 2009, Hacker & Pierson 2010, Gilens & Page 2014) carry the argument further, arguing that the very fabric of democratic responsiveness is threatened. In their eyes, it is not just that the rich have more influence than the poor. They claim that the very rich and their interest group partners have their way at the expense of the median voter and the general public. In their view, the visible politics that reflects the interests of ordinary citizens and preoccupies the attention of most scholars of American politics is a largely irrelevant sideshow. The real decisions (or at least the ones that are important to the very wealthy) occur in a different arena that is largely veiled behind closed doors. Gilens & Page (2014) resurrect the idea that democratic institutions in the United States have failed—it is the elites who govern, they say, not the public. In the language of the Internet headlines that popularize this work, the United States is an oligarchy, not a democracy.

This literature that highlights political inequality must be reconciled with a separate literature that finds a certain degree of policy representation in the United States. This more optimistic literature on representation finds government responsiveness to public opinion in that policy outcomes correlate, sometimes strongly, with public preferences. Responsiveness is most evident when public opinion and policy are each in a broad aggregate—the liberal versus conservative tendency of public opinion on the one hand and the liberal versus conservative content of public policy on the other. The American states that are more liberal in terms of ideological identification are the ones that enact the most liberal policies across a variety of issues (Erikson et al. 1993). At the national level, public opinion as measured by its policy mood (also termed ideological mood) over a variety of issues predicts the ideological tendency of legislation, albeit with a lag (Erikson et al. 2002). When estimated for specific policy issues rather than general ideology, weaker

12 Erikson

A nn

u. R

ev . P

ol it.

S ci

. 2 01

5. 18

:1 1-

29 . D

ow nl

oa de

d fr

om w

w w

.a nn

ua lr

ev ie

w s.

or g

A cc

es s

pr ov

id ed

b y

U ni

ve rs

ity o

f C

al if

or ni

a -

R iv

er si

de o

n 05

/2 8/

19 . F

or p

er so

na l u

se o

nl y.

PL18CH02-Erikson ARI 2 April 2015 7:12

relationships between opinion and policy are found. At the state level, the more the electorate supports a particular policy (for instance, gay marriage), the more likely the policy will be enacted (e.g., Lax & Phillips 2009). And at the national level, spending levels for particular functions appear to react to public support for spending on those functions (e.g., Wlezien 1995a,b). Further, the process appears to be dynamic, with the public’s support for legislation conditional on the degree to which government has resisted or responded in the past (Wlezien 1995b, Erikson et al. 2002, Soroka & Wlezien 2010). For thoughtful and thorough reviews of the growing literature on public opinion and public policy, see Shapiro (2011) and Burstein (2014).

On the one hand, then, public opinion and policy are statistically related in a manner that suggests government responds to public opinion, while on the other hand, the rich get their way politically at the expense of the poor. One way of reconciling two seemingly conflicting ideas is to proclaim one as correct and insist that the other is wrong. (For example, some critics of democratic performance dismiss the evidence of policy representation as likely to be spurious, perhaps as the result of elites manipulating a gullible public.) However, it is difficult to deny either that political influence rises with affluence or that public opinion influences policy.

The present article tries to answer the following question: Given the evidence for policy rep- resentation in the United States, to what extent is this representation equitable across levels of income? It lays out the case for considerable policy representation of public opinion but considers the possibility that policy influence can be strongly conditioned by level of affluence. The article elaborates on some reasons why low-income voters could be underrepresented and reviews studies that estimate the influence of the affluent to dominate that of the poor. The precise degree to which income matters for representation continues to be a matter for future research.

THE DOWNS MODEL AND THE MEDIAN VOTER

A useful starting point is median voter theorem, which is the cornerstone of much theorizing about democratic representation but a target of critics who see little policy representation. The usual formalization is the Downs (1958) model, in which the median voter is decisive. The familiar argument begins with three assumptions: voters who decide based on their ideological proximity to the candidates, one policy dimension, and two opposing candidates solely interested in getting elected. It follows that candidates’ policy positions converge to the preference of the median voter. Public opinion prevails.

Some critics of the current state of US democracy dismiss median voter theory as wrong or irrelevant. According to Hacker & Pierson (2010), median voter theory emphasizes “politics-as- electoral-spectacle” rather than the real story involving the influence of wealthy individuals and interest groups. According to these authors, efforts to validate median voter theory “fail on their face” (Hacker & Pierson 2010, p. 167). Similarly, Gilens & Page (2014) express bewilderment that “despite many criticisms of the median voter theorem as simplistic and empirically inapplicable or wrong, a good many scholars. . . . still cling to the idea that the policy preferences of the median voter tend to drive policy outputs” (Gilens & Page 2014, p. 565).

It is easy to mock the Downs model because, when taken literally, clearly both the assumptions and the key predictions are almost comically inaccurate. Voters are not very ideological and react to other considerations besides policy views. Political candidates care about more than simply winning elections. Their personal ideological agendas keep them from converging toward the median voter position. Further, even with today’s vast polling technology, candidates cannot pretend to know where the median voter stands. And voters are often uncertain about candidate positions. Election outcomes rarely converge toward 50–50 as predicted by the model. Given these empirical realities, can one proclaim that the median voter generally prevails?

www.annualreviews.org • Income Inequality and Policy Responsiveness 13

A nn

u. R

ev . P

ol it.

S ci

. 2 01

5. 18

:1 1-

29 . D

ow nl

oa de

d fr

om w

w w

.a nn

ua lr

ev ie

w s.

or g

A cc

es s

pr ov

id ed

b y

U ni

ve rs

ity o

f C

al if

or ni

a -

R iv

er si

de o

n 05

/2 8/

19 . F

or p

er so

na l u

se o

nl y.

PL18CH02-Erikson ARI 2 April 2015 7:12

Rather than reject the paradigm of median voter theory, formal modelers have expanded the model to incorporate the considerations mentioned above—policy-motivated candidates and sup- porters, nonideological voters, uncertain voters, and uncertain candidates (see Grofman 2004 for a good discussion). The central tenet of the Downs model holds: In general elections, candidates win more votes by moving toward the ideological center than by moving away from it. The de- gree to which this happens depends on the degree to which voters take policy into account and politicians anticipate the electoral consequences of their actions.

People have political influence to the extent that they participate politically and attend to policy issues. Should we care when people vary in their attention to and influence on government? If the task of the public is to offer its collective judgment about the solution to some problem where people all share a common interest, then perhaps an inequality of influence is not such a bad thing. For example, US foreign policy responds more to the views of the business and professional elite than to those of the general public ( Jacobs & Page 2005). Arguably, under many circumstances, this type of unequal influence is preferable to a foreign policy governed by the populist urges of an ill-informed public. Yet we might think differently when elite interests clearly differ from the public’s, such as when people are drafted to fight wars. When the division is between the rich and the poor, economic interests collide. To the extent that political power is distorted by wealth, the pivotal voter is not at the median but somewhere higher on the income scale. Thus we have our central question: How much does the greater political leverage exercised by the rich distort policy outputs, especially as these outputs have differential impact on the rich and the poor?

MODELING POLICY REPRESENTATION

It is helpful to depict the dynamic relationship between public opinion and policy with a stylized version of the thermostatic model originated by Wlezien (1995b) and applied by Erikson et al. (2002). We symbolize the public’s preference for change (more liberal versus more conservative) by M, standing for the public’s mood. Policy is symbolized by P. The public has an ideal policy or set point P∗ so M is the gap between policy and the preferred set point.

M t ≡ Pt − P∗ t .

The usual write-up assumes all opinions count equally so that M represents a mean,

M t =

n∑ i=1

mi

n ,

where the m’s are the equally weighted opinions of the population n. For our purposes here, let us loosen our definition of M to be the weighted mean of individual opinion where individuals’ weights are a function of their varying influence on policy.

M t =

n∑ i=1

wi mi

n and

∑ w = n.

The thermostatic model assumes that public opinion responds instantly to policy change.

M t = M t−1 + �Pt + ut .

Opinion also can shift idiosyncratically to random shocks ut, allowing the set point, which is unobserved, to change. Time might be described in biennial bands corresponding to federal elections.

14 Erikson

A nn

u. R

ev . P

ol it.

S ci

. 2 01

5. 18

:1 1-

29 . D

ow nl

oa de

d fr

om w

w w

.a nn

ua lr

ev ie

w s.

or g

A cc

es s

pr ov

id ed

b y

U ni

ve rs

ity o

f C

al if

or ni

a -

R iv

er si

de o

n 05

/2 8/

19 . F

or p

er so

na l u

se o

nl y.

PL18CH02-Erikson ARI 2 April 2015 7:12

Policy responds to mood, both because politicians directly react to shifts in public opinion and (more importantly) because voters, based on their mood, elect politicians to respond to their interests. Summarizing,

�Pt = α + λM t−1 + et . 1.

Policy changes in response to M at rate λ. Policy also responds to other influences that are independent of public opinion (e). Policy can be biased as represented by the constant term α.

According to Equation 1, public opinion influences policy (as long as λ is positive). But the relative importance of this influence is a matter of degree, depending on all the terms in the equation. As a democratic ideal, λ might equal 1.0, indicating a rapid and proportional response. The constant α would equal 0, so that policy contains no consistent liberal or conservative bias. The variance of e shocks to policy would approach 0, so that other forces besides public opinion were at a minimum. Also, representation would be easiest to maintain if the variance of u shocks to opinion were small so that the public has a stable set point. Finally and importantly, as the model is revised here, the democratic ideal would be that the w weights would all be equal, with everybody’s opinion counting equally.

Empirically, The Macro Polity (Erikson et al. 2002) estimates mood using Stimson’s (1999) algo- rithm to obtain a composite measure of liberalism–conservatism from national polls. It estimates a certain degree of policy responsiveness, which is attributed mainly to the voters’ choices for party control but also to elected officials’ direct response to public opinion. Figure 1 shows the thermostatic nature of mood: Mood generally becomes more liberal under Republican presidents and conservative under Democratic presidents. (The one exception is a growth in conservatism under Republican presidents Nixon and Ford as they pragmatically dealt with a heavily Demo- cratic Congress.) Policy change—the “laws” index, the net sum of major liberal versus conservative legislation—follows mood, but with some lag, which is expected given the standard checks and

N et num

ber of liberal law s

45

50

55

60

65

70

Bi en

ni al

p ol

ic y

m oo

d

1952 1972 1992 2012 Year

Biennial policy mood

Net number of liberal laws

0

5

10

15

20

Figure 1 Policy mood and major policy change, 1952–2012. An extension of findings from Erikson et al. (2002).

www.annualreviews.org • Income Inequality and Policy Responsiveness 15

A nn

u. R

ev . P

ol it.

S ci

. 2 01

5. 18

:1 1-

29 . D

ow nl

oa de

d fr

om w

w w

.a nn

ua lr

ev ie

w s.

or g

A cc

es s

pr ov

id ed

b y

U ni

ve rs

ity o

f C

al if

or ni

a -

R iv

er si

de o

n 05

/2 8/

19 . F

or p

er so

na l u

se o

nl y.

PL18CH02-Erikson ARI 2 April 2015 7:12

Eisenhower 1

Eisenhower 2

Kennedy/Johnson

Johnson

Nixon 1

Nixon/Ford Carter

Reagan 1

Reagan 2

GHW Bush Clinton 1

Clinton 2

GW Bush 1

GW Bush 2

Obama 1

0

10

20

30

40

N et

n um

be r o

f l ib

er al

la w

s, pr

ev io

us p

re si

de nt

ia l t

er m

45 50 55 60 65 70

Mood, lagged 4 years

Figure 2 Net liberalism of new laws by mood lagged four years, by presidential administration, Eisenhower to Obama.

balances.1 The visually clearest evidence of representation is when policy change (laws) is mea- sured for each four-year presidential term and mood is measured for the previous term, four years earlier (see Figure 2).2

As measured by Stimson’s index, mood represents a public opinion in which all voices count equally. It could be, however, that this seemingly egalitarian measure of mood is actually a marker for wealthy (or elite) opinion and it is really this elite that drives policy. This is conceivable because when opinions change across different income groups, they generally move in tandem (Ura & Ellis 2008, Wlezien & Siroka 2011). This similarity across income groups applies only to changes, not levels, of opinion. (The poor, of course, tend to be more liberal than the rich.) Suppose that the poor are more favorable to redistribution and that only the rich influence policy. In that case, policy will persistently be more conservative than it would be if people’s preferences counted equally, even as policy responds to shifts in (wealthy) opinion. As measured by the mood and laws indices, opinion and policy are not calibrated on a common scale. It is possible for the mood–laws relationship to be causal and yet for policy to be consistently more conservative than the median voter wants, as policy gets distorted by the views of the wealthy few (see sidebar, “Elite Control of Mass Opinion?”).

Despite the evidence that public opinion influences policy, we must consider the likelihood that the rich receive more (perhaps much more) representation than the poor. There are two general questions. First, considering normal politics, we can inquire about the degree to which political

1To calculate the laws score for each Congress, the first step is to consult Mayhew’s (1991) list of major legislation, updated from Mayhew’s website: http://davidmayhew.commons.yale.edu/books-with-datasets-2/. The laws index was scored as the net number of liberal minus conservative major laws for a Congress. For further details, see Erikson et al. (2002). 2Consistent with the thermostatic model, Obama’s election in 2008 followed a liberal high-point in terms of mood. Also consistent with the model, mood turned decisively more conservative during Obama’s first term. For a discussion, including rival interpretations, see Brooks & Manza (2013).

16 Erikson

A nn

u. R

ev . P

ol it.

S ci

. 2 01

5. 18

:1 1-

29 . D

ow nl

oa de

d fr

om w

w w

.a nn

ua lr

ev ie

w s.

or g

A cc

es s

pr ov

id ed

b y

U ni

ve rs

ity o

f C

al if

or ni

a -

R iv

er si

de o

n 05

/2 8/

19 . F

or p

er so

na l u

se o

nl y.

PL18CH02-Erikson ARI 2 April 2015 7:12

ELITE CONTROL OF MASS OPINION?

Some critics charge that even if policy seems to track opinion, this could be because elites are manipulating public opinion rather than because opinion is shaping policy. Indeed, it is plausible to argue that public opinion is not autonomous but instead arises from elite ideas (e.g., Zaller 1992, Bartels et al. 2005). If one defines “elites” broadly, it could hardly be otherwise. How could mass opinion change without some form of leadership? With a menu of multiple elite viewpoints from which the mass public can choose, it is difficult to see how elite influence over public opinion inherently diminishes the practice of democracy. The chief concern would be the presence of a hegemonic elite with dominant resources and capability that allows it to get what it wants by routing its preferences through the channel of public opinion. One can imagine a democracy with pliable voters acting against their own economic interests, as the unwitting agents for an economic elite.

influence varies by economic status when it comes to voting behavior and the electorally induced behavior of politicians—the stuff of much of the study of American politics. This approach would be a matter of differentially weighted voters in models like Equation 1. Second, speculating about influences outside the realm of normal politics, we can ask if politicians pay more acute attention to elite interests than to the preferences of the general public. A disproportionate elite influence could stem from special attention to the views of wealthy campaign contributors, but also the fact that politicians often are, or aspire to be, members of the economic elite themselves. In the context of Equation 1, this would be the constant term always pulling policy away from public opinion, presumably in the conservative direction. One can imagine that, by wealth and stealth, elites manipulate politicians into satisfying their special interests while possibly the general public does not even notice. Normal democratic politics, no matter how equitable, would become a sideshow as elites exercise their will.

What follows is largely limited to the first of these concerns—the degree of unequal influence in the context of normal electoral politics. Election outcomes and politicians’ anticipation of them allow for the opinions of some people to influence policy. The focus here is on the degree to which this influence is disproportionally affected by wealth. I first evaluate inequality among the components of normal democratic politics and then consider studies that attempt more globally to estimate the relative sway over policy of the rich and the poor.

The questions asked should be familiar. How much does voting participation vary with in- come? How much does political capability vary as a function of income? Do the poor follow their presumed self-interest and favor redistribution more than the rich? How much do the poor (and the rich) engage in issue voting based on redistributive issues? And ultimately, how strong are politicians’ electoral incentives to heed the views of the poor?3

UNEQUAL VOTER PARTICIPATION BY INCOME

Americans vote at a lesser rate than citizens of most other democracies. And US turnout is lowest among voters of low socioeconomic status, as defined by low income and low education. This pattern is well known (see Leighley & Nagler 2014). The focus here is on estimating the degree to

3Instead of family income, we could use some other measure of socioeconomic standing, such as occupational status or perhaps an amalgam of income and education [as do Schlozman et al. (2012)]. Personal wealth might be the ideal indicator if we had a measure. Though not perfect, personal income appears to be the best indicator of economic well-being (see Leighley & Nagler 2014 for discussion).

www.annualreviews.org • Income Inequality and Policy Responsiveness 17

A nn

u. R

ev . P

ol it.

S ci

. 2 01

5. 18

:1 1-

29 . D

ow nl

oa de

d fr

om w

w w

.a nn

ua lr

ev ie

w s.

or g

A cc

es s

pr ov

id ed

b y

U ni

ve rs

ity o

f C

al if

or ni

a -

R iv

er si

de o

n 05

/2 8/

19 . F

or p

er so

na l u

se o

nl y.

PL18CH02-Erikson ARI 2 April 2015 7:12

100

75

50

25

0 0 20 40

2006

2010

2008

2012

60 80 100

Pe rc

en t v

ot in

g

Income percentile

Figure 3 Voter turnout by income percentile, 2006–2012, based on Current Population Survey data.

which differential voter turnout affects the likely political power of the rich and poor. The working assumption is that citizens in different income categories have political influence in proportion to the frequency with which they vote. For instance, if the very rich vote three times as frequently as the very poor, a starting point for discussion is that the per capita influence of the very rich is three times that of the very poor.

Every other November, following each national election, the US Census’s Current Population Survey (CPS) interviews several thousand citizens and, among other questions, asks its respondents if they (and others in their household) had voted in the recent election. The CPS Voter Participa- tion Survey is widely regarded as the gold standard for estimating voter turnout, in part because it inflates voter participation less than other surveys such as the American National Election Studies (ANES).

Using CPS data, Figure 3 shows turnout rates as a function of income percentile from the four national elections 2006–2012. The figure stretches out the percentile range to include the very rich and the very poor. We are interested in the ratio of turnout among the rich to turnout among the poor. We see two patterns: one for the two presidential elections with their greater turnout and one for the two midterms with their lesser turnout.

Focusing on the two presidential elections, we see that turnout rates vary from approximately 50% among the very poor to more than 80% among the very rich. We could claim a ratio of per capita electoral power of approximately 8:5 between the rich and the poor. For midterms, however, the gradient of the curves is steeper, suggesting a greater variation in electoral influence. Comparing the richest with the poorest groups yields a ratio of 2:1 or more. This increase makes sense: The lower the turnout, the less influence the poor have over the outcome.

One might think that the nonparticipation of the poor is driven mainly by low voting rates among racial minorities, but this is not the case. Although racial minorities are disproportionately found among the poorer percentiles of citizens, the slopes shown in Figure 3 are similar if limited to white voters. Minorities and whites vote at approximately the same rate when income level is held constant. Among the poor, Hispanic-Americans are less likely to vote, whereas African- Americans are more mobilized to vote than are whites.

When evaluating turnout levels by income with displays such as Figure 3, a thorny problem is that surveys, including the CPS, inflate the voter rolls. People overreport having voted, largely

18 Erikson

A nn

u. R

ev . P

ol it.

S ci

. 2 01

5. 18

:1 1-

29 . D

ow nl

oa de

d fr

om w

w w

.a nn

ua lr

ev ie

w s.

or g

A cc

es s

pr ov

id ed

b y

U ni

ve rs

ity o

f C

al if

or ni

a -

R iv

er si

de o

n 05

/2 8/

19 . F

or p

er so

na l u

se o

nl y.

PL18CH02-Erikson ARI 2 April 2015 7:12

due to social desirability. In the two presidential election years, the actual turnout was a bit more than 60%. Among CPS respondents, however, approximately 70% claimed to have voted. In the two midterms, the actual voting rate was slightly more than 40%, but in the CPS midterm surveys, more than 60% claimed to have voted.

For many purposes, overreporting of turnout might not matter if it were roughly random in origin. But here we are interested in ratios of voting rates for different income groups. Suppose, for example, there were virtually no underreporting of voting but a rate of overreporting that was constant for each income group. The result would be a greater net overreport among the poor because of their lesser rate of actual participation. In a hypothetical example, set the true rates as 60% among the rich and 20% among the poor for a ratio of 3:1. Now assume 20% of nonvoters at every income level falsely report voting. The reported rate among the poor zooms from 20% to 36% whereas the observed rate among the rich rises by only 8 percentage points. The hypothetical observed ratio of 68:36 deflates the true 3:1 turnout ratio to be less than 2:1. Could we be underestimating the turnout gradient separating the voting rates of rich and poor?

Here I introduce a new method to estimate the true voting rates by income levels. First, I use citizen characteristics of respondents from the New York Times/CBS News (NYT&CBS) surveys for the year surrounding Election Day for two elections—2006 and 2008. From these pooled surveys, I extract the percentage of respondents in each of the five income categories used in these surveys. Second, I use National Exit Poll (NEP) data for each of the two years to extract the proportion of actual voters in the same income categories.4 Third, I record the proportion of eligible voters who actually vote in the election.5 With this information, I back out estimates of voter turnout percentages within each of the five income categories. The advantage of this plan, of course, is that voter turnout is not measured by self-reports (although income, of course, is self-reported).

The results are shown in Figure 4. The vote participation by income percentile gradient is shown two ways, with CPS data (from the same data as Figure 3 but with the income categories collapsed to be identical to the NYT&CBS income categories) and using the NEP and NYT&CBS data. The graphs also show linear regression lines fitting the NYT&CBS data, allowing possible extrapolation to as far as the first and 100th percentiles.

By forcing the net voting rates to be equal to the national results, we see in Figure 4 that NYT&CBS/NEP data yield decidedly lower voting rates than do CPS data. For midterm year 2006, the two slopes are roughly parallel, as if the rate of overreporting increases somewhat with income level. The net result is a mild increase in the estimated gap between the voting rates of the richest and poorest, from about 2:1 to 2.5:1.

The results for 2008, however, are a bit of a surprise (Figure 4b). The CPS vote inflation for 2008 apparently occurred mainly among the poor. Among the wealthy, the voting rate estimated from the NYT&CBS/NEP data approximates that derived from CPS data. The rich vote in presidential elections and evidently do not overreport. Meanwhile a major gap occurs for the poor, as the NYT&CBS/NEP slope is much steeper than the CPS slope for 2008. In effect Figure 4b suggests that the ratio of 2008 participation comparing the very rich to the very poor is about 2.5:1, similar to the conclusion for 2006 from Figure 4a. To understand this result, consider that low-income voters were surprisingly scarce in the 2008 NEP compared to general surveys.

4Although exit polls are notoriously inaccurate for immediate Election Day prediction, they are presumably accurate when poststratified and properly weighted. 5Turnout among eligible voters is obtained from Michael P. McDonald’s United States Elections Project website, http://www.electproject.org/home.

www.annualreviews.org • Income Inequality and Policy Responsiveness 19

A nn

u. R

ev . P

ol it.

S ci

. 2 01

5. 18

:1 1-

29 . D

ow nl

oa de

d fr

om w

w w

.a nn

ua lr

ev ie

w s.

or g

A cc

es s

pr ov

id ed

b y

U ni

ve rs

ity o

f C

al if

or ni

a -

R iv

er si

de o

n 05

/2 8/

19 . F

or p

er so

na l u

se o

nl y.

PL18CH02-Erikson ARI 2 April 2015 7:12

100

75

50

Pe rc

en t v

ot in

g

Income percentile

2006 2008

25

0 0 25 50 75 100

CPS

NYT&CBS/NEP

100

75

50

Income percentile

25

0 0 25 50 75 100

CPS

NYT&CBS/NEP

a b

Figure 4 Vote participation by income, measured two ways, in (a) 2006 and (b) 2008.

From this exercise, one can conclude that citizen surveys underestimate the true gap in voting between the richest and the poorest income percentiles. To understand the significance of the income gradients shown in Figure 4, consider that with a turnout ratio of 2.5:1, the median voter in terms of income is at the 66th percentile of citizen income.

This exercise has identified one major reason why the rich enjoy more political influence than the poor: A member of the richest 1% is 2.5 times more likely to vote than someone in a homeless shelter. Further, it is easy to imagine that the ratio is 3:1 or greater if one considers all sources of political participation and their possible effectiveness (Schlozman et al. 2012). Democracy may still be at work, but with a severe economic bias. An optimistic interpretation is that the influence gap between rich and poor could be substantially narrowed by simply inducing greater participation by the poor. But closing the influence gap entirely would also require closing the gap in personal resources, a matter to which we turn next.

INFORMATION INEQUALITY AND INCOME INEQUALITY

One advantage the rich have over the poor is greater access to news about politics. The rich are more knowledgeable about politics than the poor, resulting in more effective political influence. How severe is this pattern?

Here I turn to the ANES as the data source. ANES surveys ask respondents about their family income, and the cumulative file contains a record of family income by percentile. The percentiles are five unevenly spaced groupings so that their means are percentiles of 8, 25, 50, 81, and 96. For the presidential election years 1952–2004, there is also a common scale of political information, calibrated so that each year’s distribution has a standard deviation of one. Zaller (2004) has compiled scores on this index of political knowledge for all ANES surveys from 1952 to 2000, and Benjamin Highton has updated the index for 2004. The index is based on whatever information items and interviewer ratings are included in each year’s survey.

With an income scale in percentiles that is normed to be constant across years and a standardized information scale that is normed to have a constant variance across years, we can put the two scales together without the need to control for specific year variables. The ordinary least squares (OLS) regression of information level on income percentile yields a coefficient of 1.31 (1.10 for reported

20 Erikson

A nn

u. R

ev . P

ol it.

S ci

. 2 01

5. 18

:1 1-

29 . D

ow nl

oa de

d fr

om w

w w

.a nn

ua lr

ev ie

w s.

or g

A cc

es s

pr ov

id ed

b y

U ni

ve rs

ity o

f C

al if

or ni

a -

R iv

er si

de o

n 05

/2 8/

19 . F

or p

er so

na l u

se o

nl y.

PL18CH02-Erikson ARI 2 April 2015 7:12

0 25 50 75 100 0

25

50

75

100

Income percentile

Po lit

ic al

k no

w le

dg e

(p er

ce nt

ile )

More than 1 SD below mean

Below mean

Above mean

More than 1 SD above mean

MEANMEANMEAN

Figure 5 Political information by income percentile, 1952–2004. Data are from the American National Election Studies.

voters only), which implies that the difference between the richest and poorest voters is slightly more than one standard deviation on the knowledge scale.

Figure 5 depicts the pattern in detail. Political knowledge scores are divided into four groups: more than one standard deviation below the mean amount of information; below the mean, but not by more than one standard deviation; better informed than the mean, but not by more than one standard deviation; and more than one standard deviation better informed than the mean. The figure shows low information concentrated among low-income voters and high information con- centrated among high-income voters. Almost 50% of respondents in the highest income category score have political knowledge scores at least one standard deviation above the mean; similarly, almost 50% of respondents in the lowest income category score at least one standard deviation below the mean. Meanwhile, high-income respondents have less than a 10% chance of being in the lowest information category, and low-income respondents have less than a 10% chance of being in the highest information category. One question is whether the poor’s information deficit retards their political response to their economic circumstances. This is addressed next.

INFORMATION, INCOME, AND POLITICAL PREFERENCES

We focus on income as a predictor because the rich and poor diverge both in terms of their economic interests and their political resources. To what extent do voters—rich and poor— choose policy positions on redistribution issues consistent with their presumed self-interest based on income? Similarly, do rich and poor channel their redistributive preferences into vote choices consistent with their economic circumstances? And how much do these tendencies depend on the citizen’s degree of political knowledge?

It would be a mistake to assume a major gulf between the opinions of rich and poor. Polls show a general concern about inequality across income groups but also only limited demand for government action (e.g., Page & Jacob 2009, Bartels 2008; but see McCall 2013). Differences across income groups do exist on questions of domestic spending and welfare (Bartels 2008, Ura & Ellis 2008, Gilens 2009; but see Soroka & Wlezien 2008).

The effect of income or social class on political attitudes is confounded by race and culture. Despite claims by Democrats that their party represents the working poor, the white working class (typically defined as non-college-educated whites) is often said to be the base of today’s

www.annualreviews.org • Income Inequality and Policy Responsiveness 21

A nn

u. R

ev . P

ol it.

S ci

. 2 01

5. 18

:1 1-

29 . D

ow nl

oa de

d fr

om w

w w

.a nn

ua lr

ev ie

w s.

or g

A cc

es s

pr ov

id ed

b y

U ni

ve rs

ity o

f C

al if

or ni

a -

R iv

er si

de o

n 05

/2 8/

19 . F

or p

er so

na l u

se o

nl y.

PL18CH02-Erikson ARI 2 April 2015 7:12

0 25 50 75 100 0

25

50

75

100

Income percentile

O pi

ni on

: s oc

ia l w

el fa

re s

pe nd

in g

(p er

ce nt

ile )

Conservative

In between or no opinion

Liberal

Low knowledge

0 25 50 75 100 0

25

50

75

100

Income percentile

Conservative

In between or no opinion

Liberal

High knowledge

Figure 6 Social welfare opinion by income percentile, 1980–2004, by level of political knowledge. Whites only. Data are from the American National Election Studies.

Republican Party. The argument is that the white working class has become inflamed by racial resentment and that their cultural conservatism has been exploited by Republicans. Does that cultural conservatism trump considerations of their economic plight? (For a useful review, see Enns & Wlezien 2011b.) Meanwhile, blacks and other racial minorities tend to support social welfare programs regardless of their income level and vote Democratic, sharing “linked fate” with their poorer brethren (Dawson 1995).

To make the test hard for detecting an income effect, the following analysis is limited to white survey respondents only. As Bartels (2008) has shown, social welfare policy opinions and vote choice vary by income even when the analysis is restricted to whites. Here we ask to what extent this pattern is conditional on political awareness. The data are from the ANES, this time for presidential elections 1980–2004. This choice of years is designed to match the years when Zaller’s (2004) information scale is available along with a constant measure of social welfare policy preference. The measure of social welfare policy is the ANES question about whether the respondent favors more taxes and government services or lower taxes and fewer services. This question is measured on a 1 to 7 scale, with 4 as the midpoint. I divide responses into three groups: social welfare liberals who prefer more spending (1 through 3), social welfare conservatives who prefer lower taxes and spending (5 through 7), and those who say they have no opinion or no interest in the question (a score of 4).6

Figure 6 shows the relationship between opinion and income (whites only), further subdivided by scores on the Zaller information scale. For ANES respondents below the information mean, there is virtually no relationship between income and policy preference, with a large share of respondents not taking sides (scoring 4 or “don’t know,” “no interest”). For the more knowledge- able voters (above the mean), we see a clearer pattern whereby support for spending is greatest among the poor and least among the rich. The more informed one is, the more one supports policy positions consistent with one’s economic standing.

6For all respondents in cumulative ANES surveys, the regression coefficient predicting opinion on the seven-point scale from percentiled income is −0.9 for white voters (less than one point from richest to poorest), shifting slightly to −1.2 when racial minorities are included.

22 Erikson

A nn

u. R

ev . P

ol it.

S ci

. 2 01

5. 18

:1 1-

29 . D

ow nl

oa de

d fr

om w

w w

.a nn

ua lr

ev ie

w s.

or g

A cc

es s

pr ov

id ed

b y

U ni

ve rs

ity o

f C

al if

or ni

a -

R iv

er si

de o

n 05

/2 8/

19 . F

or p

er so

na l u

se o

nl y.

PL18CH02-Erikson ARI 2 April 2015 7:12

0 25 50 75 100 0

25

50

75

100

Income percentile

Pe rc

en t v

ot in

g D

em oc

ra ti

c fo

r P re

si de

nt

Tax/spend liberals

Tax/spend liberals

Tax/spend conservatives

Tax/spend conservatives

Low knowledge

0 25 50 75 100 0

25

50

75

100

Income percentile

High knowledge

Figure 7 Presidential vote by social welfare opinion and income, 1980–2004. Whites only. Data are from the American National Election Studies.

Next, do poor (and rich) voters cast ballots consistent with their position on the spending ques- tion? Figure 7 (again considering only white respondents) shows that the answer is yes, for voters in the top half of the information scale. The vast majority of liberals voted Democratic in presiden- tial elections, 1980–2004, and the vast majority of conservatives voted Republican—regardless of income level. The comparable difference in vote choice between liberals and conservatives among voters below the information mean is visible but considerably smaller, suggesting less issue voting on taxing and spending. In both information groups, the liberal-versus-conservative gap in vote choice rises somewhat with income.

With their greater evidence of issue voting, the more informed voters show much more of an income gradient to their vote choices than the less informed. An OLS regression of Democratic presidential voting on income percentile results in a coefficient (for whites) of −0.18 for voters above the mean in knowledge. This implies almost a 20-point differential in Democratic versus Republican voting between the poorest and richest income percentiles, among the informed. For those below the mean on the knowledge scale, the coefficient is a mere −0.05. Thus, the more informed one is, the more one votes in accord with the presumed interests of one’s income level.

This section illustrates that income makes some difference in terms of attitudes toward re- distribution and voting Democratic versus Republican, even when nonwhite voters are excluded from the analysis. Citizens are more likely to take stands that represent their “class interests” when they are more knowledgeable and to vote accordingly. For white voters (less so for minorities, not shown), these effects are largely conditional, however, on above-average political knowledge. Information is correlated with income, reinforcing the class bias in representation. As with par- ticipation, one can infer a potential reform. An improvement in political information among the poorer voters could change election outcomes and change policy.

ESTIMATING INCOME DIFFERENCES IN POLITICAL INFLUENCE

From the estimates above, the richest Americans enjoy at least 2.5 times the political influence of the poorest Americans. Considering all the resources of voters in different economic strata, the ratio could actually be far greater. The working assumption is a “Downsian” world where elections matter; it is just that some participate more effectively than others. As discussed in the Introduction, however, some contemporary observers (e.g., Winters & Page 2009, Hacker &

www.annualreviews.org • Income Inequality and Policy Responsiveness 23

A nn

u. R

ev . P

ol it.

S ci

. 2 01

5. 18

:1 1-

29 . D

ow nl

oa de

d fr

om w

w w

.a nn

ua lr

ev ie

w s.

or g

A cc

es s

pr ov

id ed

b y

U ni

ve rs

ity o

f C

al if

or ni

a -

R iv

er si

de o

n 05

/2 8/

19 . F

or p

er so

na l u

se o

nl y.

PL18CH02-Erikson ARI 2 April 2015 7:12

Pierson 2010, Gilens & Page 2014) see elections as essentially a sideshow with policy manufactured elsewhere. So how much does political influence vary by economic strata actually? As mentioned at the outset of this article, several studies attempt to estimate the relative degree to which policy represents the preferences of the rich, the poor, and those in between.

Using the 1978–1980–1982 ANES Senate Election Study, Bartels’ (2008) pioneering study modeled the relative contributions of the ideological preferences of states’ high-, medium-, and low-income groups to predicting the roll call votes of US senators. Even with party affiliation of the senator controlled, Bartels’ statistical evidence suggested that senators respond much more to the views of the rich than to voters in the middle of the income spectrum, and based on the coefficients, the views of the poor count negligibly in determining the roll call behavior of a state’s US Senate delegation.

Bhatti & Erikson (2011) corrected a hidden problem of statistical identification in Bartels’ analysis but reported similar estimates of influence by income category using Bartels’ original data from the ANES Senate Election Study. They obtained somewhat erratic results, however, using alternative state-level surveys—the 2000 and 2004 Annenberg studies and the 2004 NEP. Although not with the severe gradient found using the original ANES data, their various tests followed the general pattern implying that the views of affluent voters have more sway with US senators than the views of the poor.

Using the Annenberg surveys, Rigby & Wright (2013) estimate the effects of low-, medium-, and high-income opinion in the states on the positions of Democratic and Republican political candidates. Their study is distinct for using multi-item survey instruments for economic and social policy rather than summary liberal–conservative identification. On both economic and social dimensions, they find the positions of each party’s candidates much more predictable from the views of the states’ rich than from the economic middle, and not at all predictable from the views of the poor.

Why does the behavior of politicians track the views of the rich much more than those of the poor? The explanation might be that candidates are more familiar with the views of their fellow well-to-do citizens than with those of ordinary constituents. Or it might be that politicians choose to ignore the views of the poor on the assumption that their views are politically irrelevant. Al- though the poor dilute their power by their low turnout rate and by their lesser political knowledge (as we have seen), indifference to the views of the poor could have electoral consequences, because the votes of the poor contribute to the determination of winners and losers in elections.

A further test would be to see whether actual policies (rather than roll call votes or candidate positions) neglect the opinions of the poor. Rigby & Wright (2011) apply such a test, predicting the general liberalism–conservatism of state economic and social policies from the liberal versus conservative ideological identification of state electorates, where opinion is partitioned by high, medium, and low income. Again they find that the views of the rich matter more and the views of the poor count little or not at all. Flavin (2012) finds a similar pattern for specific state policies.

What should be made of these studies? Do American electoral institutions do a good job of responding to public opinion except that the voices that matter have a strong upper-class accent? Are the interests of the poor excluded from the factors that affect policy? The literature presents a persistent narrative that political representation may be a luxury available to the wealthy alone. Caution is required, however, when interpreting the statistical evidence. The statistical models are necessarily complex and the evidence from multiple specifications tends to be wobbly, in part because the views within the different income strata often are highly correlated. The difficulty of measuring state opinion increases exponentially when state samples are divided into three separate income groups. And the opinions of the rich are easier to measure; the rich more freely offer opinions in surveys and their opinions are more constrained ideologically. These factors

24 Erikson

A nn

u. R

ev . P

ol it.

S ci

. 2 01

5. 18

:1 1-

29 . D

ow nl

oa de

d fr

om w

w w

.a nn

ua lr

ev ie

w s.

or g

A cc

es s

pr ov

id ed

b y

U ni

ve rs

ity o

f C

al if

or ni

a -

R iv

er si

de o

n 05

/2 8/

19 . F

or p

er so

na l u

se o

nl y.

PL18CH02-Erikson ARI 2 April 2015 7:12

might give the rich a head start in a statistical race among imperfectly measured indicators of segmented public opinion (Stimson 2011, Bhatti & Erikson 2011).7

We turn finally to Gilens’ Affluence and Influence (2012), which was highlighted in the Introduc- tion. Gilens analyzed more than 2,000 national policy proposals that were polled over the years 1980–2002 with a supplement for 1964–1969. For each proposal, Gilens estimated the opinions at the 10th, 50th, and 90th income percentiles and recorded whether the proposed change happened within four years. For each percentile, he modeled the outcome (change or not) as a function of opinion at the specific income percentile. His key finding was that affluent opinion (at the 90th percentile) predicts policy outcomes more than opinions at the median or opinions of the poor (at the 10th percentile). For instance, when the absolute difference in the opinions of the 10th and 90th percentiles exceeds 10%, the opinions of the 90th percentile seem to matter far more.

One cannot deny that Gilens’ work identifies a differential in political influence based on income or that this is an impressive and careful work of scholarship. Nonetheless, the methodology for measuring the preferences at different income percentiles makes problematic the multivariate analysis of different group opinions (see Gilens 2012, appendix). Gilens measures opinion as percentage in favor of change among those holding opinions, with the consequence that the greater rate of “no opinion” among the silent voices of the poor (Berinsky 2004) is ignored. The relative scarcity of opinion holding among the poor can explain some of the seeming lack of influence of opinions voiced by the poor. One could also question how much the tilt of policy toward the views of the well-to-do depends on the time period selected, given its run of Republican presidents and (after 1994) a Republican Congress. (According to Gilens, the results are similar for the 1960s supplemental data.)

Although a reanalysis is well beyond the scope of this article, I offer one illustrative graph, using Gilens’ publicly available data.8 I define upper-income issues as those favored by citizens at the 90th percentile by at least 10 points more than those at the 10th percentile and lower-income issues as those favored by citizens at the 10th percentile by at least 10 points more than those at the 90th percentile.

Figure 8 shows the predicted (probit) success rates (change versus status quo) as a function of general public opinion for the two types of issues. The two slopes are fairly flat and almost parallel at around 0.30, as if it takes 10 percentage points of added popularity to nudge the probability of a policy change by a mere three points. The figure shows a bias in favor of the status quo for each type of issue in the sense that it takes far more than a simple majority to approve a policy in order to get it enacted. The gap between the two slopes is approximately 17%, meaning that, holding general public opinion constant, high-income issues have about a 17-percentage-point greater likelihood of passage.

As with the studies of roll call voting, party politicians’ positions, and state policies discussed in the previous section, an obvious interpretation is that politicians either choose to ignore or are blind to the demands of less affluent citizens, seeing only the positions favored by the affluent. I suggest a different interpretation that allows politicians to respond to public opinion generally, perhaps only modestly distinguishing between views of the rich and poor or between informed and uninformed voters. Consider the possibility that the main reason why lower-income issues are harder to pass than upper-income issues is not that politicians choose to ignore lower-income

7Brunner et al. (2013) overcome the measurement problem by exploiting referendum results in rich and poor precincts represented by California state legislators. They find little difference in the degree to which representatives heed the preferences of rich and poor in their roll call voting. 8Gilens’ data are available at http://www.russellsage.org/research/data/economic-inequality-and-political-representation. The data analyzed here are from version DS52.

www.annualreviews.org • Income Inequality and Policy Responsiveness 25

A nn

u. R

ev . P

ol it.

S ci

. 2 01

5. 18

:1 1-

29 . D

ow nl

oa de

d fr

om w

w w

.a nn

ua lr

ev ie

w s.

or g

A cc

es s

pr ov

id ed

b y

U ni

ve rs

ity o

f C

al if

or ni

a -

R iv

er si

de o

n 05

/2 8/

19 . F

or p

er so

na l u

se o

nl y.

PL18CH02-Erikson ARI 2 April 2015 7:12

0 25 50 75 100 0

0.25

0.50

0.75

1

Percent favoring change Pr

ed ic

te d

pr ob

ab ili

ty o

f c ha

ng e

90th percentile favors at least 10% more

10th percentile favors at least 10% more

Figure 8 Probability of change by percent favoring change, for high-income issues and low-income issues. Data are from Martin Gilens, available at http://www.russellsage.org/research/data/economic-inequality-and- political-representation.

views but rather that lower-income issues face a more challenging political environment. Despite the sizeable numerical representation of liberal organizations that try to help the disadvantaged (Grossman 2012), there is little doubt that the interest group structure in Washington serves the interests of the affluent more than those of the poor. This interpretation resembles that of Gilens & Page (2014) in that powerful interest groups can thwart public opinion. Unlike Gilens & Page (2014), however, this interpretation allows public opinion across income groups to have some (if unequal) influence.9

STATUS QUO BIAS AND A PRESSURE FOR LIBERALISM?

The most obvious fact about policy choices in the United States is the bias for the status quo rather than change. Gilens’ (2012) analysis makes clear that for a policy to be popular is no guarantee that the policy will come to pass (see also Enns et al. 2014). From the status quo bias plus the upper- income bias to representation, one might infer that when legislation does pass, it is endorsed by the rich and pushes policy in the conservative direction. That inference would be wrong. Sometimes haltingly, policy moves mainly leftward, not rightward. Let us try to understand this seeming paradox.

Given the status quo bias, the policies that do become law are likely to be favored by public opinion. In fact, major policies such as Medicare, the War on Poverty, key civil rights bills, plus the Reagan and Bush tax cuts were passed only when polls showed plurality support. The Affordable Care Act (Obamacare) stands out as an arguable exception, with its subsequent political fallout.

Despite contrary examples (e.g., the long-term decline in the progressivity of the federal income tax), Congress much more frequently passes liberal than conservative legislation. The reasons are complex (see, e.g., Baumgartner & Jones 2015). Erikson et al. (2002) examined hundreds of laws that were deemed “major” by Mayhew (1991) and classified them as liberal, conservative, or ideologically neutral. Liberal laws outnumbered conservative laws by a count of 113 to 9, or a

9Gilens & Page (2014) claim that only interest group opinion and wealthy opinion (the “elites” at the 90th income percentile) matter, at the expense of the average citizen at the 50th income percentile. Evaluation of this provocative claim, which deserves methodological scrutiny, is beyond the scope of the present article.

26 Erikson

A nn

u. R

ev . P

ol it.

S ci

. 2 01

5. 18

:1 1-

29 . D

ow nl

oa de

d fr

om w

w w

.a nn

ua lr

ev ie

w s.

or g

A cc

es s

pr ov

id ed

b y

U ni

ve rs

ity o

f C

al if

or ni

a -

R iv

er si

de o

n 05

/2 8/

19 . F

or p

er so

na l u

se o

nl y.

PL18CH02-Erikson ARI 2 April 2015 7:12

ratio of more than 10 to one; the exceptions occurred when Republicans held the presidency plus at least one house of Congress. (The ratio is 60 liberal to 4 conservative in terms of legislation to expand/contract benefits or services or deal with taxation.) If we did an accounting of public opinion in advance of these mainly liberal laws, the strong expectation is that in most instances the poor would be more supportive and the rich more resistant.

If major policy change is hard (the status quo bias) and moves almost always in the liberal direction, how can it be reconciled with the phenomenon of an upper-class bias to policy making? There is an easy resolution of the paradox. Shifting societal demands cause the flow of policy demands to move mainly in the liberal direction, with (presumably) the poor at least being passively supportive.10 Meanwhile, affluent interests resist and are often successful, at least in the short run. The degree of public pressure for change helps to determine the amount of change. However, on less publicly visible or controversial issues of public policy and on those without obvious ideological content, the views of the affluent (and educated) elite face little resistance.11

CONCLUSIONS

This article has dealt with the seeming paradox that government seems responsive to public opinion, yet the views of the rich appear to count more. For readers of a progressive bent, I offer a modest nudge of optimism. Public opinion does matter and the press of public opinion has moved policy leftward, not rightward, on average. When policy questions are a matter of rich versus poor, the opinions of the rich are influential beyond their numbers and capable of slowing the pace of reformist legislation. The degree of this upper-class bias should continue to be a matter for future research.

An important puzzle is why the poor are so ineffectual politically. The poor dilute their political influence by patterns of nonvoting and political inattention. As we have seen, some studies that attempt to measure the relative influence of income groups go further and claim that the influence of the poor is about zero. It remains to be seen if this claim holds up to further statistical scrutiny. Future research should continue to focus on whether politicians actually ignore the preferences of less affluent voters, and if so, why.

What could be done to politically awaken the economically disadvantaged? Obviously lowering barriers to voting and other forms of participation would help, as would reducing the role of money in politics. Would it be possible to mobilize the poor for political action? Could government activity empower the poor, perhaps as a by-product of specific programs designed to help (Mettler & Soss 2004, Bruch et al. 2010)? What if a genuine social movement were to arise from the poor to demand further redistribution? Would there be a political collision with the established order?

DISCLOSURE STATEMENT

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

10The minimum wage offers food for thought. Several of Mayhew’s (1991) major laws were minimum-wage increases. These are always popular, with the least affluent most in favor. On the one hand, inflation persistently lowers the minimum wage in terms of real dollars. On the other hand, liberal legislation restores (or partially restores) its initial value. Raising the minimum wage draws resistance from conservative business forces, contributing to the status quo bias. But reducing or eliminating the federal minimum wage is never on the table. 11As Gilens (2012) shows, the relative success of rich and poor voters varies with type of issue. The poor actually appear more influential on social welfare, whereas the rich are particularly dominant on foreign policy and taxes.

www.annualreviews.org • Income Inequality and Policy Responsiveness 27

A nn

u. R

ev . P

ol it.

S ci

. 2 01

5. 18

:1 1-

29 . D

ow nl

oa de

d fr

om w

w w

.a nn

ua lr

ev ie

w s.

or g

A cc

es s

pr ov

id ed

b y

U ni

ve rs

ity o

f C

al if

or ni

a -

R iv

er si

de o

n 05

/2 8/

19 . F

or p

er so

na l u

se o

nl y.

PL18CH02-Erikson ARI 2 April 2015 7:12

ACKNOWLEDGMENTS

I thank Matt Kantrowitz for his research assistance. I am very grateful to the following scholars for their helpful comments on earlier drafts: Larry Bartels, Sarah Bruch, Martin Gilens, Kathleen Knight, Jan Leighley, Jonathan Nagler, Eric Schickler, Bob Shapiro, Laura Stoker, Elizabeth Washbrook, and Chris Wlezien.

LITERATURE CITED

Bartels L. 2008. Unequal Democracy: The Political Economy of the Gilded Age. Princeton, NJ: Princeton Univ. Press

Bartels L, Heclo H, Hero RE, Jacobs LR. 2005. Inequality and American governance. See Jacobs & Skocpol 2005, pp. 88–155

Baumgartner FR, Jones BD. 2015. The Politics of Information: Problem Definition and the Course of Public Policy in America. New York: Cambridge Univ. Press

Berinsky AJ. 2004. Silent Voices: Public Opinion and Political Participation in America. Princeton, NJ: Princeton Univ. Press

Bhatti Y, Erikson RS. 2011. How poorly are the poor represented in the U.S. Senate? See Enns & Wlezien 2011a, pp. 223–46

Bonica A, McCarty N, Poole K, Rosenthal H. 2013. Why hasn’t democracy slowed rising inequality? J. Econ. Perspect. 27(3):103–24

Brooks C, Manza J. 2013. A broken public? Americans’ responses to the Great Recession. Am. Sociol. Rev. 78(5):727–48

Bruch SK, Ferree MM, Soss J. 2010. From policy to polity: democracy, paternalism, and incorporation of disadvantaged citizens. Am. Sociol. Rev. 75(2):205–26

Brunner E, Ross SL, Washington E. 2013. Does less income mean less representation? Am. Econ. J. Econ. Policy 5(2):53–76

Burstein P. 2014. American Public Opinion, Advocacy, and Policy in Congress. What the Public Wants and What It Gets. New York: Cambridge Univ. Press

Corak M. 2013. Income inequality, equality of opportunity, and intergenerational mobility. J. Econ. Perspect. 27(3)79–102

Dawson M. 1995. Behind the Mule: Race and Class in African-American Politics. Princeton, NJ: Princeton Univ. Press

Downs A. 1958. An Economic Theory of Democracy. New York: Harper Enns PK, Kelly NJ, Morgan J, Volscho T, Witko C. 2014. Conditional status quo bias and top income shares:

how U.S. political institutions have benefited the rich. J. Polit. 76(2):289–303 Enns P, Wlezien C, eds. 2011a. Who Gets Represented? New York: Russell Sage Found. Enns P, Wlezien C. 2011b. Group opinion and the study of representation. See Enns & Wlezien 2011a,

pp. 1–26 Erikson RS, Mackuen MB, Stimson JA. 2002. The Macro Polity. New York: Cambridge Univ. Press Erikson RS, Wright GC, McIver JP. 1993. Statehouse Democracy: Public Opinion and Policy in the American

States. New York: Cambridge Univ. Press Flavin P. 2012. Income inequality and policy representation in the American states. Am. Polit. Res. 40:29–59 Gilens M. 2009. Preference gaps and inequality in representation. PS: Polit. Sci. Polit. 42(2):335–41 Gilens M. 2012. Affluence and Influence: Economic Inequality and Political Power in America. Princeton, NJ:

Princeton Univ. Press Gilens M, Page BI. 2014. Testing theories of American politics: elites, interest groups, and average citizens.

Perspect. Polit. 12(3):564–81 Grofman B. 2004. Downs and two-party convergence. Annu. Rev. Polit. Sci. 7:25–46 Grossman M. 2012. The Not-So-Special Interests: Interest Groups, Public Representation, and American Governance.

Stanford, CA: Stanford Univ. Press Hacker JS, Pierson P. 2010. Winner-take-all politics: public policy, political organization, and the precipitous

rise of top incomes in the United States. Polit. Soc. 38:152–204

28 Erikson

A nn

u. R

ev . P

ol it.

S ci

. 2 01

5. 18

:1 1-

29 . D

ow nl

oa de

d fr

om w

w w

.a nn

ua lr

ev ie

w s.

or g

A cc

es s

pr ov

id ed

b y

U ni

ve rs

ity o

f C

al if

or ni

a -

R iv

er si

de o

n 05

/2 8/

19 . F

or p

er so

na l u

se o

nl y.

PL18CH02-Erikson ARI 2 April 2015 7:12

Jacobs LR, Page BI. 2005. Who influences U.S. foreign policy? Am. Polit. Sci. Rev. 99:107–23 Jacobs LR, Skocpol T, eds. 2005. Inequality in American Democracy. New York: Russell Sage Found. Lax JR, Phillips JH. 2009. Gay rights in the states: public opinion and policy responsiveness. Am. Polit. Sci.

Rev. 103(Aug.):367–76 Leighley JE, Nagler J. 2014. Who Votes Now? Demographics, Issues, Inequality, and Turnout in the United States.

Princeton, NJ: Princeton Univ. Press Mayhew DR. 1991. Divided We Govern. New Haven, CT: Yale Univ. Press McCall L. 2013. The Undeserving Rich: American Beliefs about Inequality, Opportunity, and Redistribution. New

York: Cambridge Univ. Press Meltzer AH, Richards SF. 1981. A rational theory of the size of government. J. Polit. Econ. 89(5):914–27 Mettler S, Soss J. 2004. The consequences of public policy for democratic citizenship: bridging policy studies

and mass politics. Perspect. Polit. 2(1):65–73 Neckerman KM, ed. 2004. Social Inequality. New York: Russell Sage Found. Page BI, Jacobs L. 2009. Class War. Chicago: Univ. Chicago Press Piketty T, Saez E. 2014. Inequality in the long run. Science 244:(6186):838–43 Rigby E, Wright GC. 2011. Whose statehouse democracy? Policy responsiveness to poor versus rich con-

stituents in poor versus rich states. See Enns & Wlezien 2011a, pp. 189–222 Rigby E, Wright GC. 2013. Political parties and representation of the poor in the American states. Am. J.

Polit. Sci. 57(3):552–65 Schlozman K, Verba S, Brady HE. 2012. The Unheavenly Chorus: Unequal Political Voice and the Broken Promise

of American Democracy. Princeton, NJ: Princeton Univ. Press Shapiro RY. 2011. Public opinion and American democracy. Public Opin. Q. 75(5):982–1017 Soroka SN, Wlezien C. 2008. On the limits of political inequality and representation. PS: Polit. Sci. Polit.

41(2):319–27 Soroka SN, Wlezien C. 2010. Degrees of Democracy and Politics, Public Opinion and Policy. Cambridge, UK:

Cambridge Univ. Press Stimson JA. 1999. Public Opinion in America: Moods, Cycles, and Swings. Boulder, CO: Westview. 2nd ed. Stimson JA. 2011. The issues in representation. See Enns & Wlezien 2011a, pp. 347–60 Ura JD, Ellis CR. 2008. Income, preferences and the dynamics of policy responsiveness. PS: Polit. Sci. Polit.

41(Oct.):785–94 Winters JA, Page BI. 2009. Oligarchy in the United States? Perspect. Polit. 7:731–51 Wlezien C. 1995a. Dynamic representation: the case of U.S. spending on defense. Br. J. Polit. Sci. 26:81–103 Wlezien C. 1995b. The public as thermostat: the dynamics of public preferences for spending. Am. J. Polit.

Sci. 39:981–1000 Wlezien C, Siroka SN. 2011. Inequality in policy responsiveness? See Enns & Wlezien 2011a, pp. 285–310 Zaller J. 1992. The Nature and Origins of Mass Opinion. Cambridge, UK: Cambridge Univ. Press Zaller J. 2004. Floating voters in U.S. presidential elections, 1948–2000. In Studies in Public Opinion, ed. WE

Saris, PM Sniderman, pp. 166–214. Princeton, NJ: Princeton Univ. Press

www.annualreviews.org • Income Inequality and Policy Responsiveness 29

A nn

u. R

ev . P

ol it.

S ci

. 2 01

5. 18

:1 1-

29 . D

ow nl

oa de

d fr

om w

w w

.a nn

ua lr

ev ie

w s.

or g

A cc

es s

pr ov

id ed

b y

U ni

ve rs

ity o

f C

al if

or ni

a -

R iv

er si

de o

n 05

/2 8/

19 . F

or p

er so

na l u

se o

nl y.

PL18-FrontMatter ARI 6 April 2015 13:4

Annual Review of Political Science

Volume 18, 2015Contents

A Conversation with Hanna Pitkin Hanna Pitkin and Nancy Rosenblum � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � 1

Income Inequality and Policy Responsiveness Robert S. Erikson � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � �11

How Do Campaigns Matter? Gary C. Jacobson � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � �31

Electoral Rules, Mobilization, and Turnout Gary W. Cox � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � �49

The Rise and Spread of Suicide Bombing Michael C. Horowitz � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � �69

The Dysfunctional Congress Sarah Binder � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � �85

Political Islam: Theory Andrew F. March � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � 103

Borders, Conflict, and Trade Kenneth A. Schultz � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � 125

From Mass Preferences to Policy Brandice Canes-Wrone � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � 147

Constitutional Courts in Comparative Perspective: A Theoretical Assessment Georg Vanberg � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � 167

Epistemic Democracy and Its Challenges Melissa Schwartzberg � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � 187

The New Look in Political Ideology Research Edward G. Carmines and Nicholas J. D’Amico � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � 205

The Politics of Central Bank Independence José Fernández-Albertos � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � 217

What Have We Learned about the Resource Curse? Michael L. Ross � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � 239

v

A nn

u. R

ev . P

ol it.

S ci

. 2 01

5. 18

:1 1-

29 . D

ow nl

oa de

d fr

om w

w w

.a nn

ua lr

ev ie

w s.

or g

A cc

es s

pr ov

id ed

b y

U ni

ve rs

ity o

f C

al if

or ni

a -

R iv

er si

de o

n 05

/2 8/

19 . F

or p

er so

na l u

se o

nl y.

PL18-FrontMatter ARI 6 April 2015 13:4

How Party Polarization Affects Governance Frances E. Lee � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � 261

Migration, Labor, and the International Political Economy Layna Mosley and David A. Singer � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � 283

Law and Politics in Transitional Justice Leslie Vinjamuri and Jack Snyder � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � 303

Campaign Finance and American Democracy Yasmin Dawood � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � 329

Female Candidates and Legislators Jennifer L. Lawless � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � 349

Power Tool or Dull Blade? Selectorate Theory for Autocracies Mary E. Gallagher and Jonathan K. Hanson � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � 367

Realism About Political Corruption Mark Philp and Elizabeth Dávid-Barrett � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � 387

Experiments in International Relations: Lab, Survey, and Field Susan D. Hyde � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � 403

Political Theory as Both Philosophy and History: A Defense Against Methodological Militancy Jeffrey Edward Green � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � 425

The Empiricists’ Insurgency Eli Berman and Aila M. Matanock � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � 443

The Scope of Comparative Political Theory Diego von Vacano � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � 465

Should We Leave Behind the Subfield of International Relations? Dan Reiter � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � 481

Indexes

Cumulative Index of Contributing Authors, Volumes 14–18 � � � � � � � � � � � � � � � � � � � � � � � � � � � 501

Cumulative Index of Article Titles, Volumes 14–18 � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � 503

Errata

An online log of corrections to Annual Review of Political Science articles may be found at http://www.annualreviews.org/errata/polisci

vi Contents

A nn

u. R

ev . P

ol it.

S ci

. 2 01

5. 18

:1 1-

29 . D

ow nl

oa de

d fr

om w

w w

.a nn

ua lr

ev ie

w s.

or g

A cc

es s

pr ov

id ed

b y

U ni

ve rs

ity o

f C

al if

or ni

a -

R iv

er si

de o

n 05

/2 8/

19 . F

or p

er so

na l u

se o

nl y.

1. ar:
1. logo: