State Party Position Taking: Elites Listening to Elites
Gerald C. Wright
Indiana University
Elizabeth Rigby
George Washington University
Abstract
We measure the economic and social policy preferences of state electorates (Annenberg 2000 and 2004 Election Surveys) and the state parties’ stances on the same policy dimensions (Project Vote Smart NPATs) to see whether the inclusion of mass partisanship changes the recent conclusions of governmental responsiveness to the affluent over middle and low income constituents (Gilens 2005, Bartels 2008, Fiorina, et al. 2008, Rigby and Wright 2013). This has been implied, but not tested, in recent work (Garand 2010, Enns 2015). We look at state party responsiveness to income group preferences within the parties and find that (a) the preferences of the affluent continue to dominate with the important exception of Republicans on social policy, and (b) that the consequences of underrepresentation of partisans’ preferences by their parties’ platforms varies systematically by party and policy dimension. Lower income Democrats are underrepresented on social policy—the state Democratic parties are relatively more liberal and aligned with the preferences of high income Democrats. Lower income Republicans are underrepresented on economic policy where the state parties are relatively more conservative because they align with the preferences of affluent Republicans.
Paper prepared for delivery at the 2016 American Political Science Association conference, Philadelphia, PA, September 1-4, 2016.
Introduction
Most thinking about democracy assumes the ideal of government that is responsive “to the preferences of its citizens, considered as political equals”(Dahl 1971). How far we are in practice from this ideal is one of the central themes in contemporary political analyses. The apparent consensus of political science, at least as it pertains to the U.S. until recently, has been that (a) elected officials worry about reelection and (b) this leads them to pay attention to their constituents. The conclusion from a lot of empirical work has been the public opinion matters, at least on issues salient to the public(Erikson, et al. 1993, Canes-Wrone, et al. 2002, Erikson, et al. 2002, Ansolabehere and Jones 2010). That relatively upbeat view has been challenged by recent analyses which have reframed the traditional question of “does public opinion matter?” Instead, these researchers have asked “whose opinions matter?” Using quite different methodologies (Bartels 2008, Gilens 2005, Gilens 2012, Gilens and Page 2014, Rigby and Wright 2013) they find that the rich or affluent have a lot more say than the middle class,and the poor have nodetectible independent influence on legislators or public policy at all. Interestingly, the message of these works is consistent with a long line of elite theorists from Lippman (1922) to Schattschneider (1960) to Converse (1964), all of whom argue in different ways that the mass electorate is not up to the task of providing meaningful public control over public policy. However, this new research on income inequality has drawn a lot of attention because the findings rest on the same kinds of evidence that researchers have used for years to suggest that government is reasonably responsive to public opinion.
The most recent wrinkle in the line of work examining income inequality and governmental responsiveness brings party and partisanship into play. Of course partisanship is central to understanding mass attitudes and preferences as well as the policy behavior of elected officials, and some recent work argues that partisanship is so important that it largely swamps inequalities based on income (Garand 2010, Enns 2015). The idea here is that it is not the rich who are overrepresented so much as the partisans whose candidates are elected(Wright 1989b, Clinton 2006, Brunner, et al. 2011). Such a conclusion has very different normative implications—ones that we believe are less troubling than the argument that the wealthy have most of the power and the poor have virtually none. In this paper weexamine the argument that bringing parties and partisanship into the analysis mitigates the problem of undue influence by the wealthy, but doing so with methods and data much more in line with the long line of research on representation. Wefind that the parties are primarily responsive to the wealthy—with one significant exception--but that the degree of underrepresentation varies by party and policy dimension.
The paper proceeds as follows, we briefly outline the development of research on who is represented which provides the foundation for our analysis of within party income group effects on state party positioning. The next section describes the data, measures and estimation strategy that we employ, which is followed by four short analyses: a replication of unequal income group representation, a demonstration of the importance of partisans for party positioning, assessing within-party equality of income group influence, and finally an exploration of the results of the previous analysis to show how party and issue dimension condition misrepresentation of partisans by their own parties.
Does the Public Matter? Conflicting Views
Virtually everyone believes that democracy is preferable to most of the alternatives, but there is less than total agreement on what constitutes a democracy. Without getting into the thicket of various strands of democratic theory, we will simply posit that most recognizable visions of democratic government involve government being responsive to the wishes of the governed. “Unless mass views have some place in the shaping of politics, all the talk about democracy is nonsense.”(Key 1961) Establishing whether there is such linkage between the public and government has, of course, been the work of many scholars and writers because the question of the role of the public in policy making has such momentous normative importance. Some influential writers like Walter Lippman(1922) dwell on the limitation of people’s cognitive abilities to deal with the complexities of the world, arguing that instead we construct artificial images of the world that are hopelessly subjective and biased. The informational limitations of the average citizen were documented in the first systematic election surveys by Paul Lazarsfeld and colleagues (Lazarsfeld, et al. 1948, Berelson 1954), by authors of The American Voter(Campbell, et al. 1960) and even more thoroughly in their exhaustive coverage of available public opinion resources Delli Carpini and Keeter (1996) demonstrated that if Americans have opinions on many issues, they are founded on woefully little knowledge. This echoed in broad outlines the seminal report by Phillip Converse(1964)on the lack of ideological thinking in the mass public, arguing that the average voter’s attitudes on even important issues of the day were best characterized as “non-attitudes.” Thus, the early and most empirical work since has cast a picture of the average citizen as disinterested, uninformed with claims to issue stances that are generally fleeting, and certainly not grounded in any kind of systematic framework or set of principles. Thus a lot of empirical evidence can me marshalled to support the quite unflattering portrait of the electorate inherited from writers from Lippman(1922), Schattschneider (1960) and others.[1]
The fact that the average survey respondent falls well short of some idealized democratic citizen has not stopped efforts to assess whether government is responsive to public opinion. The classic study here is Miller and Stokes’ “Constituency Influence in Congress”(1963) which sought to determine whether and how members of Congress aligned their attitudes and roll call behavior with opinion in their districts on race, social welfare and foreign policy issues. While that initial effort has some methodological problems (Achen 1977, 1978, Erikson 1978, 1981) subsequent efforts generally concluded, that in spite of the public’s lack of interest, information and real attitudes, there exists considerable correspondence between districts’ ideological orientations and the roll call voting of members of Congress(Erikson and Wright 1980, Wright and Berkman 1986, Erikson and Wright 2000, Ansolabehere, et al. 2001, Canes-Wrone, et al. 2002, Erikson and Wright 2009, Ansolabehere and Jones 2010, Hollibaugh, et al. 2013, Canes-Wrone 2015, Hall 2015, Sulkin, et al. 2015, Miler 2016) and even state legislatures (Hogan 2004, 2008, Birkhead 2015). Beyond legislative voting, studies of state policy (Erikson, et al. 1993) and national policy making over time (Erikson, et al. 2002) add up to a large body of work which indicates that indeed, even without attentive, well-informed citizens, elected officials are reasonably responsive to public preferences. Interesting, as the methods and data sources have developed in quantity and sophistication, the evidence for responsiveness appears to be even stronger (Ansolabehere and Jones 2010).
All of this seems to be kind of good news for democracy. It appears that we achieve Key’s goal of public opinion having an impact on elected officials and policy and does so without meeting the high requirements of an informed, attentive, and principled electorate. But virtually all of these studies have considered the question in one form or another of “does public opinion matter” and as a consequence for the most part treats public opinion as a single organic entity. That is, for example, members of Congress have a district, and many models assume the appropriate representation is the median (or more often mean) voter. And given this question, the evidence indicates that public opinion does indeed matter. Some exceptions to this have looked at different constituencies, asking the question of “whose opinions matter” with the idea that responsiveness may be not to some magical median voter, but rather may favor strategically significant sub-constituencies. Some have looked and found partisan biases in representation(Wright 1989b, Clinton 2006, Brady 2007, BAFUMI and HERRON 2010). Given the high levels of party polarization today, it is of little surprise that partisans in a constituency are better represented when their representative or senator is of the same party. This is not the “democratic” ideal of “continued responsiveness of government to the preferences of its citizens, considered as political equals” (Dahl 1971). However, given a significant choice (ideological distance) between candidates and parties—which many consider a necessary condition of popular control—and the candidates are more or less equal distance from the modal voter, it is pretty much impossible for elected officials to be fully congruent with the preferences of the median or mean voter.[2]
The more bothersome news comes from the studies mentioned earlier with their take on question of “whose opinions matter?” In Martin Gilens (2005, Fiorina, et al. 2008, Gilens 2011)mapped the preferences of citizens on over 1,800 policy questions and matched these to policies passed by the national government. He reports an overwhelming pattern in which national policies are much more likely to change consistent with the preferences of the affluent (90th percentile in income) than those of the median wage earners or the poor (10th percentile). Larry Bartels comes to the same conclusion in examining the responsiveness of Senate roll call voting to different income groups: Senators are quite responsive to their better off constituents (upper third), much less so to middle income constituents and virtually not at all to the least well-off of those they supposedly represent. Building on these findings, Elizabeth Rigby and I (Rigby and Wright 2013) focused on the states, and in particular positioning of state parties. We found a pattern of relative influence that largely mirrored those of Gilens and of Bartels. In studies that systematically contrast the influence of elites, experts, organized interests and public opinion, the public comes out way behind in terms of its relative influence (Jacobs and Page 2005, Gilens and Page 2014). The rich clearly have disproportionate influence, the middle income group’s influence much less and less consistent, and the independent influence of the poor is virtually is non-existent.More recently, work has extended the question of relative responsiveness in parliamentary systems and the findings echo what has been found in the U.S.: Legislators do much better at heeding the wishes of the well-heeled (Stadelmann, et al. 2015).
The Role of Parties
Three economists offer an interesting and plausible challenge to this overall conclusion of the underrepresentation of the poor. Brunner et al. (2010) use referenda voting of precincts in California, stratified by income, and relate these to state legislators’ voting on same referenda. Their study compares high and low income precincts’ voting across 77 referenda, and they compare this to the voting on these by the members of the state Assembly and Senate. They focus on the place of parties in representation, and they conclude there is overrepresentation of the poor when Democrats win and overrepresentation of high income folks when Republicans win; thus parties swamp differences among income groups in terms of misrepresentation.
Peter Enns (2015)argues that income groups preferences are so similar that we achieve pretty good representation “by coincidence,” although this argument is strongly disputed by Gilens (2015). Most relevant for our purposes is Enns’ argument that such coincidental representation is quite unlikely for partisans since their preferences are so divergent. The implication we draw is that since people are sorted by partisanship much more effectively than by income, and similarly because officials presumably are more aligned with their own partisans, that we should find minimal, if any differences in representation of income groups within the parties.
This brings us to the central question of this analysis. Does inequality of representation vanish when we control for the party affiliation of citizens as well as that of elected officials? The setting for this study ispositions of the states parties on two policy dimensions: economic and social issues. In an earlier analysis we established that the poor are not well represented, and indeed not represented at all in terms of independent influence on the positions of the parties (Rigby and Wright 2013). Our hypothesis is that controlling for party identification will reveal inequality of responsiveness to be minimal.
This equality of party responsiveness to partisans’ preferences across income groups could occur through two mechanisms. First, it might be that after party/issues sorting, income is no longer related to issue preferences. If so, then empirically we cannot distinguish differential responsiveness because there is no longer any systematic within party variance in issue preferences associated with income. The second possibility is that the parties are responsive to their partisans’ preferences regardless of income. That is, the ideal of equal influence could be achieved within parties.
There is a counter hypothesis that the same sort of resource differences that give the wealthy and better educated an edge in virtually every type of political participation operates within parties just as it does everywhere else. This pessimistic conjecture has going for it that within the context of U.S. politics there are no significant cases that we know of that run counter to the exhaustive finding of Schlozman, Verba and Brady (1999, 2012)that greater resources translate in to more effective political voice.
Data and Measures.
Our assessment of the impact inequalities in party responsiveness to different income groups requires measures, at the state level, of mass preferences as well as the positions of the parties on same issue dimensions.
Preferences of the State Electorates. For the mass data we are using the National Annenberg election surveys of 2000 (N=58,373) and 2004 (N=81,422). Thedata collected include a large number of survey respondents from the 48 continental states, with all butfive states having at least 500 respondents across the two survey years.[3] The limitation of theseAnnenberg survey data is that respondents were interviewed across a series of rolling, cross-sectionalsurveys in which different sets of (somewhat overlapping) questions were asked of each cross-section. Togenerate measures of policy preferences across these surveys, we pooled the surveys for2000 into one dataset and the surveys from 2004 into a second merged dataset. For eachpooled dataset we identified a set of commonly asked policy items in each survey. These included 30items from the 2000 surveys and 28 items from the 2004 surveys—all asking about respondents’ positionson a range of issues: economic, health and social welfare,social/moral, energy and environment, campaign finance, and legal/ethical concerns. Since allrespondents were not asked all questions, we imputed the missing data using imputation by multiplechained equations (see Royston 2005), which predicts survey participants’ responses based on theirpreferences among items they were asked.
Following our earlier work(Rigby and Wright 2011, 2013), we extract two dimensions from the Annenberg surveys, one set of issues to tap the dimension of economic policy preferences and one tapping social issues. The former includes respondents’ preferences on tax policy, health care, social welfare and K-12 school funding. The latter encompassed abortion, the death penalty, gay rights, school prayer and gun control. The specific items used are listed in Table 1 of the appendix. The methods used are straightforward. After the imputation of missing values for questions that were not asked on specific surveys, we factor analyzed the items constituting the two dimensions separately for 2000 and 2004, using the first dimension (with eigenvalues from 1.58 to 3.22 and second eigenvalues below .50.[4] The factor scores are standardized (mean=0, std dev =1.0). To get state and income groups averages we combined the 2000 and 2004 survey files for an overall working N of 136,056.