Whose Preferences Pass
Abstract
Gilens and Page (2014) tested four theories of American politics on 1,779 survey-polled proposed policy changes from 1981 to 2002. With the preferences of the 50th income percentile, the preferences of the 90th income percentile and an interest-group alignment index in one model, the 50th-percentile coefficient was .03 and not significant, against .76 and .56 for the other two. The two income groups' preferences correlate at .78, and the model's R-squared is .074. The authors write that average citizens often get the policies they want as coincidental beneficiaries. Critics dispute what the regression can separate and what counts as winning. On the 185 cases where the two groups' majorities took opposite sides, both sides report that the middle-income side got its preferred outcome 47 percent of the time and the affluent side 53 percent. A 2024 systematic review of 25 studies and 1,163 estimates finds a positive income gradient with wide variation and states that the degree of differential responsiveness remains highly contested. Baumgartner and colleagues, studying lobbying on 98 issues, report that 60 percent of campaigns failed to change policy. The studies that relate preferences to policy outcomes do not use campaign contributions as a variable. The Gilens-Page authors state that their evidence does not indicate that the average citizen always loses out.
2.1 What each study measured
This chapter sets one dataset's published results beside the published responses to them. The studies measure different things, and the table states the measure and the limit of each as the authors or later reviewers wrote it.
Gilens and Page, Bashir, Enns and Branham, Soroka and Wlezien work from the same Gilens dataset of polled proposals. Bartels and Lax, Phillips and Zelizer analyze Senate roll calls. Elkjaer and Iversen use international survey data on fiscal policy. Baumgartner and colleagues start from interviews with lobbyists. The preference studies (Gilens and Page, Bashir, Enns, Bartels, Lax and colleagues, Elkjaer and Iversen) relate preferences to policy outcomes. None of them uses campaign contributions as a variable.
| Study | Measured | Stated limit |
|---|---|---|
| Gilens and Page 2014 | Adoption within four years of 1,779 polled proposals, 1981 to 2002, against 50th and 90th percentile support and an interest-group alignment index, with a correction for measurement error | Polled, salient proposals only; the 90th percentile is not the very rich; R-squared .074; the data cannot separate agenda blocking from rejection |
| Bashir 2015 | Replication of the main table, a simulation of the estimator, and a count of outcomes where the groups disagree | Gilens published a reply in 2016 |
| Enns 2015 | Relative support across income groups, a counterfactual for adoption, and strong partisans in 1999 to 2002 surveys | Uses the Gilens data; the partisan window falls under a Republican Congress; cannot test off-agenda cases |
| Bartels 2006 | Senate roll-call responsiveness by constituent income third, 1989 to 1994; income growth by party of president, 1948 to 2001 | In Bartels's words, a single governmental institution in a single six-year period |
| Lax, Phillips and Zelizer 2019 | 49 Senate roll calls, 2001 to 2015, three measures of unequal representation | Roll calls only; the authors call the responsiveness regressions messy to interpret |
| Elkjaer and Iversen 2020 | Preferred levels of fiscal and redistributive policy in advanced democracies | Fiscal and redistributive policy only; the authors state that the result does not mean inequality is adequately addressed |
| Elkjaer and Klitgaard 2024 | Coding of 1,163 published estimates from 25 studies | Reports published estimates and cannot correct their data |
| Page, Bartels and Seawright 2013 | Preferences and activity of 104 wealthy Chicago-area residents in 2011 | Pilot; one metropolitan area; 37 percent response rate; no policy outcomes |
| Baumgartner and colleagues 2009 | 98 issues identified through lobbyist interviews; each side's measured resources against its outcome | Every issue had organized advocates; measures lobbying, not citizens' preferences |
2.2 The Gilens-Page model
Gilens and Page analyzed 1,779 survey-polled proposed policy changes from 1981 to 2002 and tested whether adoption within four years followed the preferences of the 50th income percentile, the preferences of the 90th income percentile, or a Net Interest-Group Alignment index. With all three in one model, the 50th-percentile coefficient was .03 and not significant, against .76 for the 90th percentile and .56 for interest groups.
The paper tests four theories in one equation: majoritarian electoral democracy, economic-elite domination, majoritarian pluralism and biased pluralism. Of 1,923 policy questions that met four criteria, 1,779 also had income breakdowns, no constitutional amendment or Supreme Court ruling, and a clear adoption outcome. Preferences at the 50th income percentile stand for average citizens and preferences at the 90th, about $146,000 in household income in 2012 dollars, stand for economic elites. The interest-group index draws on the Fortune "Power 25" lists plus ten high-lobbying industries. In Table 3, model 4, the elite coefficient is .76 and the group coefficient .56.
Evidence: Exhibit 693
The authors summarize the result in the abstract: "economic elites and organized groups representing business interests have substantial independent impacts on U.S. government policy, while average citizens and mass-based interest groups have little or no independent influence." The sentence concerns independent influence with the other predictors held constant. Popular write-ups of the paper use stronger headline labels.
Exhibit 693 · Peer-reviewed article
Testing Theories of American Politics: Elites, Interest Groups, and Average Citizens
economic elites and organized groups representing business interests have substantial independent impacts on U.S. government policy, while average citizens and mass-based interest groups have little or no independent influence.
Tests four theories (Majoritarian Electoral Democracy, Economic-Elite Domination, Majoritarian Pluralism, Biased Pluralism) in one model on 1,779 survey-polled proposed policy changes, 1981-2002 (1,923 met four criteria; 1,779 also had income breakdowns, no constitutional amendment or Supreme Court ruling, and a clear adoption outcome). Preferences at the 50th income percentile stand for average citizens and at the 90th (about $146,000 household income in 2012 dollars) for economic elites; interest groups enter as a Net Interest-Group Alignment index built from the Fortune 'Power 25' lists plus ten high-lobbying industries. Outcome: whether the change was adopted within four years. Table 3, model 4: citizens .03 (not significant), elites .76, groups .56, R-squared .074. Table 2: citizen-elite preference correlation .78 after measurement-error correction; groups-citizens .04 (mass-based .12, business -.10). Table 4: mass-based groups .24, business groups .43, citizens .05. The authors call the test 'tentative and preliminary' (p. 564), state that their evidence 'does not indicate that in U.S. policy making the average citizen always loses out' (p. 572; see also p. 570), describe citizens as often 'coincidental beneficiaries' of elite wins (p. 573), report that narrow pro-change majorities got change about 30 percent of the time and 80 percent majorities about 43 percent (p. 573), and state the data cannot distinguish elite theories, cannot test non-economic elites, and cannot tell agenda blocking from rejection (pp. 574, 576). Journal pagination used for locators.
- Source
- Martin Gilens and Benjamin I. Page, Perspectives on Politics 12(3), pp. 564-581
- Date
- September 2014
- Quoted from
- Abstract, p. 564
- Grade
- Peer-reviewed
- Document
- https://doi.org/10.1017/S1537592714001595
- Accessed
- 2026-11-09
- Cited by
- Findings citing this exhibit
In the Gilens-Page data the preferences of the 50th and 90th income percentiles correlate at .78 after measurement-error correction, and the model's R-squared is .074. The authors write that average citizens often get the policies they want as coincidental beneficiaries, and that their evidence does not indicate that the average citizen always loses out.
The paper's own text qualifies the result. The authors call the test "tentative and preliminary" (p. 564) and state that average citizens often win policies they favor as coincidental beneficiaries of elite wins (p. 573). They report that narrow pro-change majorities got change about 30 percent of the time and 80 percent majorities about 43 percent. The preference correlation of .78 is reported in Table 2 of the paper.
Evidence: Exhibit 693
The paper reports three further limits. The polled set omits many narrow policies such as one-firm tax breaks (p. 574). The measure of interest groups covers a small share of the groups that lobby. The data cannot test non-economic elites. The model explains 7.4 percent of the variation in adoption, and Bashir notes that this is under 10 percent.
2.3 The statistical critique
Bashir reproduced the Gilens-Page estimates and simulated the estimator. With the 50th and 90th percentile preferences correlated at .78, a true median-income coefficient of about 0.4 was estimated as essentially zero in over 20 percent of trials.
Bashir reproduced the Gilens-Page main table from the authors' input diagram and replication data. He then simulated the estimator with the 50th and 90th percentile preferences correlated at .78. A true median-income coefficient of about 0.4 came out as essentially zero in over 20 percent of trials, and at lower correlations the problem disappeared. Bashir states that his review does not show that the wealthy and business lack advantages.
Evidence: Exhibit 697, Exhibit 698
The paper's correlation of .78 sets the conditions of the simulation. Lax, Phillips and Zelizer describe the same difficulty in Senate data, calling their responsiveness regressions messy to interpret because of multicollinearity. The systematic review in section 2.8 finds that extreme results come mostly from models that enter several income groups at once.
Gilens answered Bashir in Research & Politics in April 2016. The 2024 review describes the reply as arguing that Bashir's critique is misplaced.
Exhibit 697 · Peer-reviewed article
Testing Inferences about American Politics: A Review of the 'Oligarchy' Result
average Americans have received their preferred policy outcome roughly as often as elites have when the two groups have disagreed with each other.
Reproduces Gilens and Page Table 3 using the authors' AMOS input diagram and replication data, then simulates. With the enforced 50th/90th correlation of .78, a true median-income coefficient of about 0.4 is estimated as essentially zero in over 20 percent of trials; at lower correlations the problem disappears. Descriptive reanalysis of the same dataset: in 185 cases where the two income groups' majorities are on opposite sides, the median-income side gets its preferred outcome 47 percent of the time; at 80 percent-plus support for change, change is enacted 47 percent of the time for median-income support and 52 percent for elites; when neither the rich nor the average favor change, change still occurs 23 percent of the time. Bashir states that the 'oligarchy' label comes from journalistic accounts, not Gilens and Page, and that his review does not show the wealthy and business lack advantages.
- Source
- Omar S. Bashir, Research & Politics 2(4), October-December 2015
- Date
- October-December 2015
- Quoted from
- Abstract, p. 1
- Grade
- Peer-reviewed
- Document
- https://doi.org/10.1177/2053168015608896
- Accessed
- 2026-11-09
- Cited by
- Findings citing this exhibit
2.4 When the two groups disagree
In the 185 of 1,779 Gilens-Page cases where the median-income and affluent majorities took opposite sides, the median-income side got its preferred outcome 47 percent of the time and the affluent side 53 percent. Gilens and Page accept these figures.
Bashir counted the cases where the median-income and affluent majorities were on opposite sides. In those 185 cases the median-income side got its preferred outcome 47 percent of the time and the affluent side 53 percent. Gilens and Page report the same two figures in their 2016 reply to critics. Elkjaer and Klitgaard report that Branham, Soroka and Wlezien (2017) found the same split and that disagreement of this kind occurs in about 1 case in 10.
Evidence: Exhibit 697, Exhibit 699, Exhibit 700
Both sides use these figures. They differ on how to weight the cases. Gilens and Page argue that slim majorities should not be counted the same as overwhelming ones.
Gilens and Page report that when only the affluent strongly supported a change, at 75 percent support or 25 percent opposition, it was adopted 46 percent of the time, and when only the middle class strongly supported it, 24 percent of the time. When the affluent strongly opposed a change the middle class did not, adoption was 4 percent, and in the reverse case 40 percent.
The tabulation comes from the authors' May 2016 essay. With the other group held constant, the authors report that strong affluent support raises adoption by about 25 points and that strong middle-class support has no positive association.
Evidence: Exhibit 699, Exhibit 695
Gilens and Page
Slim majorities should not be counted the same as overwhelming ones. At strong support, 75 percent in favor or 25 percent opposed, a change favored only by the affluent was adopted 46 percent of the time and one favored only by the middle class 24 percent. At strong opposition, a change the affluent opposed and the middle class did not was adopted 4 percent of the time, and in the reverse case 40 percent.
The critics
Bashir finds that at 80 percent or more support for change, the change was enacted 47 percent of the time for median-income support and 52 percent for elite support. When neither group favored change, change still occurred 23 percent of the time. Enns notes that the "twice as likely" comparison rests on a 5.1 percent subset of the data.
The 47 and 53 percent figures and the 46 and 24 percent figures describe the same dataset. The first set counts every case where the groups disagree. The second set counts the cases where support is strong. Enns's response concerns the size of the subset behind the second set.
Exhibit 699 · Authors' response (opinion essay)
Critics argued with our analysis of U.S. political inequality. Here are 5 ways they're wrong
Economic elites and interest groups can shape U.S. government policy — but Americans who are less well off have essentially no influence over what their government does.
The authors' own summary of their critics and replies. Concedes the 47/53 majority-win-rate figures from Bashir and Branham et al. Argues that slim majorities should not equal overwhelming ones. Using a 75 percent / 25 percent strong-support rule: when only the affluent strongly support a change it is adopted 46 percent of the time, when only the middle class strongly supports it 24 percent; when the affluent strongly oppose and the middle class does not, 4 percent; the reverse, 40 percent. With the other group held constant, strong affluent support raises adoption by about 25 points and strong middle-class support has no positive association. The authors say their 90th-percentile proxy may attribute to the affluent influence that belongs to the truly rich, citing Page, Bartels and Seawright.
- Source
- Martin Gilens and Benjamin I. Page, The Washington Post (Monkey Cage), 23 May 2016
- Date
- 23 May 2016
- Quoted from
- opening paragraph
- Grade
- Reported
- Document
- https://web.archive.org/web/2016/https://www.washingtonpost.com/news/monkey-cage/wp/2016/05/23/critics-challenge-our-portrait-of-americas-political-inequality-heres-5-ways-they-are-wrong/
- Accessed
- 2026-11-09
- Cited by
- Findings citing this exhibit
- Note
- Opinion essay by the study authors; the primary record is the dataset and Gilens 2015/2016 replies.
2.5 Relative policy support
Enns found that support for proposed policies among the 50th and 90th income percentiles correlates at r=.94, and at r=.68 among the 18 percent of policies where the two groups differ by more than 10 points. About 66 percent of adoption values predicted from middle-income preferences fall inside the 95 percent intervals of the values predicted from affluent preferences.
Enns uses the Gilens data, which he counts as 1,836 survey questions, to ask whether a gap between the groups' preferences implies unequal representation. In his abstract he writes that "even on those issues for which the preferences of the wealthy and those in the middle diverge, policy ends up about where we would expect if policymakers represented the middle class and ignored the affluent." He also affirms the status quo bias in the data and warns against an overly sanguine portrait of inequality.
Evidence: Exhibit 694, Exhibit 695, Exhibit 696
Enns
"Scholars cannot treat preference gaps as synonymous with unequal representation" (Enns 2015, p. 1073). Enns concedes in a note that Gilens could be right if policies the middle prefers and the affluent oppose are kept off the agenda and never polled, which the data cannot test. He also concedes that the truly rich may influence policy at the expense of both groups.
Gilens
Gilens agrees that congruence without influence is "democracy by coincidence." He writes: "There simply is not enough coincidence of policy outcomes and middle-class preferences to justify the conclusion that middle-income Americans are likely to be satisfied with the policies their government adopts" (p. 1065). Gilens argues that the cases where policy departs from middle-class preferences are important, highly salient issues.
Enns and Gilens agree that the two income groups often get the policies they favor and that coincidence of this kind carries political implications. They disagree on how often it occurs.
Exhibit 694 · Peer-reviewed article
Relative Policy Support and Coincidental Representation
even on those issues for which the preferences of the wealthy and those in the middle diverge, policy ends up about where we would expect if policymakers represented the middle class and ignored the affluent.
Re-analyzes the Gilens data (Enns counts 1,836 survey questions). Preference gaps between income groups do not imply unequal representation when both groups rank policies similarly. Correlation of 50th and 90th percentile support is r=.94 unadjusted; r=.68 among the 18 percent of policies where the groups differ by more than 10 points. About 66 percent of predicted adoption values built from middle-income preferences fall inside the 95 percent intervals of the affluent-based values. Partisan analysis uses only surveys from 1999-2002 (policies enacted 2000-2004): strong Republicans' support predicts adoption (b=.79, se .25), strong Democrats' does not (b=.10, se .17); strong-partisan support correlates only r=.43 (r=.21 where gaps exceed 10 points). Enns notes the period was one of Republican control of Congress. He also affirms the status quo bias finding and warns against an overly sanguine portrait of inequality.
- Source
- Peter K. Enns, Perspectives on Politics 13(4), pp. 1053-1064 (Reflections Symposium)
- Date
- December 2015
- Quoted from
- Abstract, p. 1053
- Grade
- Peer-reviewed
- Document
- https://doi.org/10.1017/S1537592715002315
- Accessed
- 2026-11-09
- Cited by
- Findings citing this exhibit
Exhibit 696 · Peer-reviewed reply
The Insufficiency of 'Democracy by Coincidence': A Response to Peter K. Enns
There simply is not enough coincidence of policy outcomes and middle-class preferences to justify the conclusion that middle-income Americans are likely to be satisfied with the policies their government adopts.
Gilens agrees with Enns that congruence without influence is 'democracy by coincidence' and that it has political implications. He argues that the cases where policy departs from middle-class preferences are important, highly salient issues; that Enns's 'relative policy support' is vague; that Enns overlooks characteristics of the data and makes erroneous calculations; and that coincidence is no substitute for responsiveness. Only the first page (p. 1065) was read, as a page image from the publisher; pp. 1066-1071 were not available. Later literature (Elkjaer and Klitgaard 2024, p. 320) quotes p. 1070: coincidental representation is a 'pale, counterfeit, simulacrum of democracy'.
- Source
- Martin Gilens, Perspectives on Politics 13(4), pp. 1065-1071
- Date
- December 2015
- Quoted from
- p. 1065
- Grade
- Peer-reviewed
- Document
- https://doi.org/10.1017/S1537592715002327
- Accessed
- 2026-11-09
- Cited by
- Findings citing this exhibit
2.6 The party record
Several studies in this chapter break their results out by party. The findings run in both directions. The table lists each result with the period and institution it covers.
In the 1999 to 2002 surveys of the Gilens data, Enns found that the preferences of Strong Republicans predicted policy adoption (b=.79) and the preferences of Strong Democrats did not (b=.10). Enns describes the period as one of Republican control of Congress.
The strong-partisan test uses only surveys from 1999 to 2002, with policies enacted from 2000 to 2004. In that window the support of Strong Republicans and Strong Democrats correlates at r=.43, and at r=.21 where the gaps exceed 10 points.
Evidence: Exhibit 694
Bartels analyzed Senate roll calls from 1989 to 1994 and reported responsiveness to constituents in the middle and upper income thirds and no evidence of responsiveness to the bottom third. His abstract reports that the bottom fifth receives little or no weight, especially from Republican senators.
Bartels describes the pattern as holding "for a single governmental institution in a single six-year period." Elkjaer and Klitgaard report that Bhatti and Erikson (2011) could not replicate the Senate result on new data because of multicollinearity. They also report that Bartels finds Democrats attaching relatively more weight to middle-class preferences and both parties responding most strongly to the affluent.
Evidence: Exhibit 704, Exhibit 705, Exhibit 703
Lax, Phillips and Zelizer analyzed 49 Senate roll-call votes from 2001 to 2015 and found that affluent influence is overstated and depends on partisanship. The poor get what they want more often from Democrats, and the rich more often from Republicans, but only if Republican constituents side with the rich.
The authors write that responsiveness regressions are messy to interpret because of multicollinearity and that other measures show smaller disparities. The study covers roll calls and has no measure of agenda control.
Evidence: Exhibit 701
| Source | Period and institution | Democrats | Republicans |
|---|---|---|---|
| Enns 2015 | Surveys 1999 to 2002; Republican control of Congress | Strong Democrats' preferences did not predict adoption (b=.10) | Strong Republicans' preferences predicted adoption (b=.79) |
| Bartels, abstract | Senate, 1989 to 1994 | Per Elkjaer and Klitgaard, more weight on middle-class preferences than Republicans | Bottom fifth receives little or no weight, "especially from Republican senators" |
| Lax, Phillips and Zelizer | Senate, 49 roll calls, 2001 to 2015 | The poor get what they want more often | The rich get what they want more often, if Republican constituents side with the rich |
| Elkjaer and Klitgaard, US studies | Review of published estimates | Overrepresent the rich over the poor in about 2 of 3 cases | Overrepresent the rich over the poor in 8 of 10 cases |
| Gilens 2012, as reported in the review | Polled proposals, national | Overrepresent the affluent on most issues | Overrepresent the affluent on most issues; more responsive to public opinion overall |
| Rigby and Wright 2013, as reported in the review | State party platforms | Respond most strongly to high-income preferences, especially in unequal states | Respond about equally to middle- and high-income preferences |
The Senate studies and the survey study cover different periods and institutions and use different measures. Maks-Solomon and Rigby (2020), as reported in the review, find that both parties respond more to their rich co-partisans. Lax, Phillips and Zelizer find that the Republican pattern holds only where Republican constituents side with the rich.
Bartels also reports an outcome pattern by party of president. From 1948 to 2001, real family income at the 20th percentile grew 2.6 percent a year under Democratic presidents and 0.6 percent under Republican presidents, and families at the 95th percentile did equally well under both. Bartels attributes the gap in a working paper to unemployment that was 30 percent lower and GDP growth that was 30 percent higher under Democratic presidents. According to their abstract, Blinder and Watson (2016) find the economy performed better under Democratic presidents on most measures and attribute the gap mainly to more benign oil shocks, better productivity growth and a more favorable international environment. Income growth by party of president is an outcome and does not measure whose stated preferences became policy.
Exhibit 704 · Journal article (APSA)
Is the Water Rising? Reflections on Inequality and American Democracy
there is no evidence of responsiveness to the views of constituents in the bottom third of the income distribution.
Summarizes the Senate responsiveness regressions: estimates for 1989-1994 roll calls by income third show responsiveness to middle- and upper-third constituents and none to the bottom third. Notes the pattern is 'for a single governmental institution in a single six-year period'. Figure 2 and note 2 give Census-based real family income growth under presidents, 1948-2001: 2.6 percent a year at the 20th percentile and 2.1 percent at the 95th under Democratic presidents; 0.6 percent at the 20th under Republican presidents, with affluent families doing equally well under both. Bartels treats each president's influence as beginning one year after taking office. Describes Gilens's then-unpublished work as a parallel finding.
- Source
- Larry M. Bartels, PS: Political Science & Politics, January 2006, pp. 39-42
- Date
- January 2006
- Quoted from
- p. 40
- Grade
- Peer-reviewed
- Document
- https://apsanet.org/Portals/54/files/Task%20Force%20Reports/PSJan06Bartels.pdf
- Accessed
- 2026-11-09
- Cited by
- Findings citing this exhibit
Exhibit 706 · Working paper abstract
Partisan Politics and the U.S. Income Distribution (Russell Sage Foundation working paper abstract; basis of chapter 2 of Unequal Democracy)
On average, families at the 95th percentile of the income distribution have experienced identical income growth under Democratic and Republican presidents, while those at the 20th percentile have experienced more than four times as much income growth under Democrats as they have under Republicans.
Reports large, consistent differences in real pre-tax family income growth under Democratic and Republican presidents in the post-war United States. Families at the 95th percentile saw identical growth under both parties and families at the 20th percentile more than four times as much under Democrats. Bartels attributes the gap to unemployment (30 percent lower under Democratic presidents on average) and GDP growth (30 percent higher). This is an outcome by party of president; it does not measure whose stated preferences became policy.
- Source
- Larry M. Bartels, Princeton University, for the Russell Sage Foundation
- Date
- 2005
- Quoted from
- Abstract
- Grade
- Peer-reviewed
- Document
- https://www.russellsage.org/node/4351
- Accessed
- 2026-11-09
- Cited by
- Findings citing this exhibit
Exhibit 707 · Peer-reviewed article
Presidents and the US Economy: An Econometric Exploration
Rather, it appears that the Democratic edge stems mainly from more benign oil shocks, superior total factor productivity (TFP) performance, a more favorable international environment, and perhaps more optimistic consumer expectations about the near-term future.
Finds the US economy performed better under Democratic presidents on most measures. Attributes the gap mainly to more benign oil shocks, better total factor productivity, a more favorable international environment and perhaps more optimistic consumer expectations, not to more expansionary monetary or fiscal policy. Known from the EconPapers abstract page. Relevant as the strongest published qualification of how to read Bartels's party-of-president income-growth pattern.
- Source
- Alan S. Blinder and Mark W. Watson, American Economic Review 106(4), pp. 1015-1045
- Date
- April 2016
- Quoted from
- Abstract (Blinder and Watson, Am Econ Rev 106(4): 1015-1045), via Crossref
- Grade
- Peer-reviewed
- Document
- https://doi.org/10.1257/aer.20140913
- Accessed
- 2026-11-09
- Cited by
- Findings citing this exhibit
Exhibit 701 · Peer-reviewed article
The Party or the Purse? Unequal Representation in the US Senate
We find that affluent influence is overstated and itself contingent on partisanship—party trumps the purse when senators have to take sides. The poor get what they want more often from Democrats. The rich get what they want more often from Republicans, but only if Republican constituents side with the rich.
Evaluates income bias and partisan bias together on 49 Senate roll-call votes, 2001-2015. Uses three measures (responsiveness, congruence, taking-sides). Per Elkjaer and Klitgaard 2024 (p. 321), the authors call the responsiveness regressions 'messy to interpret' because of multicollinearity (p. 928) and find smaller disparities with the other measures.
- Source
- Jeffrey R. Lax, Justin H. Phillips and Adam Zelizer, American Political Science Review 113(4), pp. 917-940
- Date
- November 2019
- Quoted from
- Abstract (publisher page citation_abstract metadata)
- Grade
- Peer-reviewed
- Document
- https://doi.org/10.1017/S0003055419000315
- Accessed
- 2026-11-09
- Cited by
- Findings citing this exhibit
2.7 Other measures, other policy areas
Elkjaer and Iversen compared preferred levels of fiscal and redistributive policy with policy in advanced democracies, including the United States, and found that policy aligns with middle-class interests. Their simulations show that models of policy change on preferences for change can credit the rich with influence when the rich are only better informed.
Elkjaer and Iversen used International Social Survey Programme data and a new dataset on who gains from fiscal policy. They write that "middle-class power has remained remarkably strong over time, even as inequality has risen" (p. 254). The authors add that this result does not mean inequality is adequately addressed and that the poor and lower middle class depend on coalitions with higher classes.
Evidence: Exhibit 702
The study covers fiscal and redistributive policy and does not use the Gilens data. The authors also write that critics who elevate the middle class to a sometimes coequal of the upper class do not explain why the rich so often get their way (p. 257). Their simulation shows how a model of policy change on preferences for change can attribute influence to a better-informed group.
Exhibit 702 · Peer-reviewed article
The Political Representation of Economic Interests: Subversion of Democracy or Middle-Class Supremacy?
middle-class power has remained remarkably strong over time, even as inequality has risen.
Argues that models regressing policy change on preferences for change (the Gilens and Bartels set-up) can credit the rich with influence when they are only better informed about short-run fiscal policy, and shows this on simulated data. Re-tests with preferred levels of spending, using International Social Survey Programme surveys across advanced democracies including the United States and a new dataset on who gains from fiscal policy. Finds policy aligns with middle-class interests. The scope is fiscal and redistributive policy, not Gilens's full range of 1,779 issues. The authors state that the middle-class result does not mean inequality is adequately addressed, and that the poor and lower middle class depend on coalitions and the generosity of higher classes. They also write that critics elevating the middle class to a sometimes coequal with the upper class do not explain why the rich so often get their way (p. 257).
- Source
- Mads Andreas Elkjaer and Torben Iversen, World Politics 72(2), pp. 254-290
- Date
- April 2020
- Quoted from
- Abstract, p. 254
- Grade
- Peer-reviewed
- Document
- https://doi.org/10.1017/S0043887119000224
- Accessed
- 2026-11-09
- Cited by
- Findings citing this exhibit
2.8 The systematic review
A systematic review of 1,163 estimates from 25 studies found a positive income gradient in responsiveness with wide variation. The predicted probability of an extreme pro-rich result is .59 in models that enter several income groups together and .08 in models run separately.
The review appeared online on 11 October 2021 and in the June 2024 issue of Perspectives on Politics. The extreme result is a positive, significant rich coefficient with a negative lower-income coefficient. Of the models producing it, 96 percent enter several income groups together, and 9 of the 25 studies use only such models.
Evidence: Exhibit 703
In the United States studies in the same review, both parties respond more to the rich than to the poor. Democrats overrepresent the poor relative to Republicans, with study-level probabilities of .27 against .12, and partisanship does not change the finding that the rich are overrepresented relative to the middle class.
The review reports that Republicans overrepresent the rich over the poor in 8 of 10 cases and Democrats in about 2 of 3. It finds that differentials are weaker in the United States than in other democracies in published work.
Evidence: Exhibit 703
The review's abstract states: "Often, the preferences of income groups overlap in a way that yields roughly equal representation, and even when income groups have opposing preferences, representational disparities are limited and better explained by partisanship than affluence." The authors conclude that the degree of differential responsiveness "remains highly contested."
Exhibit 703 · Peer-reviewed systematic review
Economic Inequality and Political Responsiveness: A Systematic Review
Often, the preferences of income groups overlap in a way that yields roughly equal representation, and even when income groups have opposing preferences, representational disparities are limited and better explained by partisanship than affluence
Codes 1,163 responsiveness estimates from 25 studies. Finds a positive income gradient overall with large variation. The predicted probability of an extreme pro-rich result (rich coefficient positive and significant, lower-income coefficient negative; high-versus-low comparisons) is .59 for models that enter several income groups together and .08 for models run separately; 96 percent of the models producing it enter several groups together. Differentials are weaker in the United States than in other democracies in published work. In US studies both parties respond more strongly to the rich than to the poor; Democrats overrepresent the poor relative to Republicans (study-level probabilities .27 against .12); Republicans overrepresent the rich over the poor in 8 of 10 cases and Democrats in about 2 of 3. Partisanship does not change the rich-over-middle result. More than one-third of studies (9 of 25) use only multivariate income-group models. The authors conclude the degree of differential responsiveness 'remains highly contested'.
- Source
- Mads Andreas Elkjaer and Michael Baggesen Klitgaard, Perspectives on Politics 22(2), pp. 318-337 (online 11 October 2021)
- Date
- June 2024
- Quoted from
- p. 318
- Grade
- Peer-reviewed
- Document
- https://doi.org/10.1017/S1537592721002188
- Accessed
- 2026-11-09
- Cited by
- Findings citing this exhibit
2.9 The lobbying counter-case
Baumgartner and colleagues studied 98 issues drawn from interviews with registered lobbyists and report that 60 percent of lobbying campaigns failed to change policy. Measured resources explain less than 5 percent of the difference between successful and unsuccessful efforts.
The project was funded by the National Science Foundation (grants SBR-9905195 and SES-0111224). The issues came from interviews conducted from February 1999 to October 2002, so every issue in the sample had organized advocates on at least one side. A pre-publication manuscript reports 214 sides, 2,221 advocates and that the side with more resources won between 45 and 56 percent of the time, a relationship that is not statistically significant.
Evidence: Exhibit 689, Exhibit 709
Mechanism: Lobbying and purchased policy
The authors write in the manuscript that existing policy already reflects past power, so lobbying outcomes depend on changes in mobilization and not on levels: "If the wealthy are so certain to win, they should already have won in a previous iteration of the process" (draft chapter 1, p. 32). The synopsis states that the Washington system has a strong bias toward the status quo and that the lobbying community reflects elite interests. The manuscript calls the resulting equilibrium "quite unfair" in many cases.
The study measures resources of advocacy sides against outcomes on issues with organized advocates. It does not measure citizens' preferences against policy, so it answers a different question from Gilens and Page. The two studies overlap on the status quo. Gilens and Page found that even 80 percent majorities got change only about 43 percent of the time, and they cite Baumgartner and colleagues for a "modest tendency" of outcomes to favor the side with greater resources (p. 568). The Gilens-Page authors also note that narrow one-firm tax breaks are underrepresented in their polled set (p. 574), which is the territory where lobbying studies sit.
Hacker and Pierson published in the same September 2014 issue of Perspectives on Politics. They contrast a view built on elections and the median voter with one in which policy and organized groups are central, and they apply the second to rising top incomes. Their 2010 article in Politics & Society argues that organizational and policy change explains the rise of top incomes better than the median voter. Their method is argument and history and is not a preference-to-policy regression.
Exhibit 689 · Book synopsis
Lobbying and Policy Change: Who Wins, Who Loses, and Why (synopsis)
The authors find that resources explain less than five percent of the difference between successful and unsuccessful efforts.
Book not read; synopsis from the authors' site. Based on 98 issues.
- Source
- Baumgartner, Berry, Hojnacki, Kimball and Leech, University of Chicago Press (authors' project page)
- Date
- 2009
- Quoted from
- Synopsis
- Grade
- Peer-reviewed
- Document
- https://fbaum.unc.edu/books/lobby/lobbying-synopsis.htm
- Accessed
- 2026-11-10
- Cited by
- Findings citing this exhibit
Exhibit 709 · Pre-publication book manuscript
Advocacy and Policy Change, pre-publication manuscript of Lobbying and Policy Change (19 July 2008)
If the wealthy are so certain to win, they should already have won in a previous iteration of the process.
Draft pagination, not the printed book. Reports 214 'sides' on 98 issues and 2,221 advocates (23 per case). Chapter 1 reasons that existing policy already reflects past power, so lobbying outcomes depend on changes in mobilization, not levels. Chapter 10 Table 10.5: the side with more resources won between 45 and 56 percent of the time, with no statistically significant relationship. The same chapter calls the status-quo equilibrium 'quite unfair' in many cases.
- Source
- Baumgartner, Berry, Hojnacki, Kimball and Leech (manuscript hosted by the authors)
- Date
- 19 July 2008
- Quoted from
- draft Chapter 1, p. 32
- Grade
- Peer-reviewed
- Document
- https://fbaum.unc.edu/books/lobby/Advocacy_July_19_2008.pdf
- Accessed
- 2026-11-09
- Cited by
- Findings citing this exhibit
Exhibit 710 · Peer-reviewed article
Winner-Take-All Politics: Public Policy, Political Organization, and the Precipitous Rise of Top Incomes in the United States
they miss the profound role of government policy in creating this “winner-take-all” pattern
Argues that political science accounts of rising inequality that rest on the median voter in electoral politics miss the role of government policy and of the organizational transformation of American politics. Their own account links major policy shifts that raised the standing of those at the top to organized efforts by resourceful private interests. The companion trade book, Winner-Take-All Politics (Simon & Schuster, 2010), has chapters titled 'The Politics of Organized Combat', 'A Tale of Two Parties' and 'Democrats Climb Aboard' (table of contents as listed in a library record).
- Source
- Jacob S. Hacker and Paul Pierson, Politics & Society 38(2), pp. 152-204
- Date
- 2010
- Quoted from
- Abstract (Hacker and Pierson, Politics & Society 38(2): 152-204), via Crossref
- Grade
- Peer-reviewed
- Document
- https://doi.org/10.1177/0032329210365042
- Accessed
- 2026-11-09
- Cited by
- Findings citing this exhibit
- Note
- Quote taken from the abstract as supplied by OpenAlex; confirm against the publisher page. The publisher abstract was seen only as copied by OpenAlex; no wording is quoted until checked against the publisher page.
Exhibit 711 · Peer-reviewed article
After the 'Master Theory': Downs, Schattschneider, and the Rebirth of Policy-Focused Analysis
These benefits are illustrated through brief examinations of two of the biggest changes in American politics over the last generation: asymmetric partisan polarization and the growing concentration of income at the top.
Published in the same issue as Gilens and Page 2014. Contrasts a Downsian view (elections and the median voter) with a Schattschneider view in which policy and organized groups are central. Applies the second view to asymmetric partisan polarization and to income concentration at the top.
- Source
- Jacob S. Hacker and Paul Pierson, Perspectives on Politics 12(3), pp. 643-662
- Date
- September 2014
- Quoted from
- Abstract (Hacker and Pierson, Perspectives on Politics 12(3): 643-662), via Crossref
- Grade
- Peer-reviewed
- Document
- https://doi.org/10.1017/S1537592714001637
- Accessed
- 2026-11-09
- Cited by
- Findings citing this exhibit
- Note
- The publisher abstract was seen only as copied by OpenAlex; no wording is quoted until checked against the publisher page.
2.10 The wealthy survey
Page, Bartels and Seawright interviewed 104 wealthy Chicago-area residents in 2011 and found them much more conservative than the general public on taxation, economic regulation and social welfare. Of the respondents, 58 percent identified as Republican and 27 percent as Democratic, leaners included.
The refined sample of 83 came from a list of wealth-holding households in about the top 4 percent, narrowed toward the top 1 percent. The response rate was 37 percent, mean wealth $14,006,338 and median wealth $7,500,000. Wealthy Democrats in the sample were more conservative on economic issues than Democrats in the general public.
Evidence: Exhibit 708
The abstract reports that the respondents "are extremely active politically and that they are much more conservative than the American public as a whole with respect to important policies concerning taxation, economic regulation, and especially social welfare programs." The survey measures preferences and activity and has no policy outcomes. The authors warn of sampling error in a small pilot confined to one metropolitan area and offer the rival reading that wealthy preferences may reflect better information.
Exhibit 708 · Peer-reviewed article
Democracy and the Policy Preferences of Wealthy Americans
We find that they are extremely active politically and that they are much more conservative than the American public as a whole with respect to important policies concerning taxation, economic regulation, and especially social welfare programs.
Pilot interview survey of wealthy Chicago-area residents in 2011 (NORC interviewers): 104 interviews in all, 83 in the refined sample drawn from the Wealthfinder 'rank A' list (about the top 4 percent of wealth-holding households) narrowed toward the top 1 percent; response rate 37 percent; mean wealth $14,006,338, median $7,500,000; mean income $1,040,140. 58 percent identified as Republican and 27 percent as Democrat (leaners included). Wealthy Democrats were more conservative on economic issues than Democrats in the general public, and wealth and party had independent effects. It measures preferences and political activity, not policy outcomes. The authors warn of sampling error in a small pilot confined to one metropolitan area and offer the rival reading that wealthy preferences may reflect better information. Read from the Northwestern-hosted PDF.
- Source
- Benjamin I. Page, Larry M. Bartels and Jason Seawright, Perspectives on Politics 11(1), pp. 51-73
- Date
- March 2013
- Quoted from
- Abstract, p. 51
- Grade
- Peer-reviewed
- Document
- https://doi.org/10.1017/S153759271200360X
- Accessed
- 2026-11-09
- Cited by
- Findings citing this exhibit
2.11 Questions the sources leave open
The Gilens-Page elite coefficient attaches to the 90th income percentile, about $146,000 in household income. No cited source measures the policy preferences of the top 1 percent across issues and years, so the effect cannot be assigned to the very rich.
In their May 2016 essay Gilens and Page state that their 90th-percentile proxy may attribute to the affluent influence that belongs to the very rich, and they cite the Chicago survey. Enns concedes that the truly rich may influence policy at the expense of both groups.
Evidence: Exhibit 693, Exhibit 699, Exhibit 695, Exhibit 708
The Gilens-Page data cannot show whether proposals the middle class favors and the affluent oppose fail by being kept off the agenda or by being voted down.
Gilens and Page state that the data cannot distinguish agenda blocking from rejection (pp. 574, 576). Enns states in his reply that this is a case his analysis cannot test.
Evidence: Exhibit 693, Exhibit 695
The 47 and 53 percent and the 46 and 24 percent figures come from Bashir's reanalysis and the Gilens-Page essay of May 2016. Gilens's 2016 reply to Bashir and the Branham, Soroka and Wlezien article are described through Elkjaer and Klitgaard 2024 and Elkjaer and Iversen 2020, and each finding that rests on them says so.
The published record covers both major parties in each party-conditional result listed in section 2.6. The studies that relate preferences to outcomes do not test campaign contributions.
Key findings
- Gilens and Page estimated a median-income coefficient of .03, not significant, against .76 for the 90th income percentile and .56 for interest groups, on 1,779 polled proposals (2.1).
- The two groups' preferences correlate at .78, the model's R-squared is .074, and the authors write that citizens often win as coincidental beneficiaries (2.2).
- Bashir's simulation returned a near-zero median-income estimate in over 20 percent of trials where the true coefficient was about 0.4 (2.3).
- Where the median-income and affluent majorities disagree, the outcomes split 47 and 53 percent, and Gilens and Page accept the figures (2.4).
- At strong support the authors report adoption of 46 percent for affluent-only and 24 percent for middle-class-only support (2.5).
- Enns finds the two groups' policy support correlates at r=.94 and at r=.68 where they differ (2.6).
- In 1999 to 2002 only Strong Republicans' preferences predicted adoption, under Republican control of Congress (2.7).
- Bartels found Senate responsiveness to the middle and upper thirds and none to the bottom third, 1989 to 1994 (2.8).
- Lax, Phillips and Zelizer found the poor do better with Democrats and the rich with Republicans in 2001 to 2015 (2.9).
- Elkjaer and Iversen found fiscal policy aligns with middle-class interests (2.10).
- A review of 1,163 estimates finds a positive income gradient and a predicted extreme result of .59 against .08 depending on model type (2.11).
- In US studies both parties respond more to the rich than to the poor (2.12).
- Baumgartner and colleagues report 60 percent of lobbying campaigns failed to change policy across 98 issues (2.13).
- A 2011 survey of 104 wealthy Chicago-area residents found them much more conservative than the public on taxes, regulation and welfare (2.14).
- No cited source measures the preferences of the top 1 percent across issues and years (2.15).
- The Gilens-Page data cannot separate agenda blocking from rejection (2.16).