Deciphering the Effects of Decentralization on Water Rights: State to Urban Inconsistencies in Bolivia, Kyle G. Webber   

Iceland's Environmental Saga: Motivations for Sustainable Action and Belief, Kate Gregory

Accountability for Income-Biased Representation, Patrick W. Buhr


Deciphering the Effects of Decentralization on Water Rights: State to Urban Inconsistencies in Bolivia, Kyle G. Webber

(Top of Page)

Link to Full Thesis:


Iceland's Environmental Saga: Motivations for Sustainable Action and Belief, Kate Gregory

(Top of Page)

Link to Full Thesis:


Accountability for Income-Biased Representation, Patrick W. Buhr

(Top of Page)


A growing body of research into representation has indicated that Representatives tend to favor certain constituencies when making roll-call voting decisions. These constituencies are those which are most effective at impacting the Representative’s reelection—those who vote, communicate their preferences, are attentive to the Representative’s behavior, and contribute to their reelection. Differences on these measures tend to fall along historical lines of discrimination regarding race, gender, and sexual orientation. Does the same effect extend to constituents of different income levels? And, if so, are disadvantaged groups able to improve their representation or hold their Representative accountable? Numerous previous studies indicate that lower-income constituents have less policy representation than those with higher income. This study adds two novel findings to this literature. First, low-income constituents are less effective than high-income constituents at improving their representation through traditional means, such as voting or contacting their Representative directly. Second, low-income constituents are able to hold their Representatives accountable for policy decisions, but only when Representatives exhibit extreme bias. Within the margins, higher-income constituents are more effective influencing elections. 


“Duty, gratitude, interest, ambition itself,” argues James Madison in Federalist Paper No. 57, “are the chords by which [representatives] will be bound to fidelity and sympathy with the great mass of the people” (Madison 1788). Recent studies of representation, however, have discovered the outsize impact of ambition—taken to mean a reelection incentive—in shaping Representatives’ decisions (Mayhew 1974, Fenno 1978, Lijphart 1997).  

If Representatives’ primary goal is to win reelection, their decisions will be largely designed to maximize their margins of victory. This may lead them to prioritize constituents with higher voting turnouts (Lijphart 1997), higher likelihoods of swing-voting (Bartels 2008), or those who are better at communicating their preferences directly or through interest groups (Miller and Stokes 1963; Verba, Schlozman, Brady 1995).  

A growing body of research has indicated that these motivations tend to fall along historical patterns of discrimination. This has led to black and Latino constituents having lower rates of policy representation than white constituents (Griffin and Newman 2008). Women are less likely to have effective representation than men (Griffin, Newman, and Wolbrecht 2012). Homosexual constituents are less likely to have their preferences enacted into law than heterosexual constituents (Lax, Phillips, and Krimmel 2009). 

Does an income distinction carry the same consequences? Numerous studies have found that low-income constituents are less likely than high income constituents to have their preferences enacted into law (Gilens 2012, Flavin 2012, Rigby and Wright 2011). Few studies, however, have explicitly analyzed the mechanisms by which disadvantaged groups can improve representation, and whether these attempts are successful at holding Representatives electorally accountable.  

This paper will seek to answer three questions: 

  1. Is income level a salient political division? 

  1. Do Representatives exhibit bias in their roll-call voting in favor of any income level? 

  1. Do mechanisms exist for disadvantaged groups to improve representation? Are these groups effective at holding Representatives accountable? 

By comparing policy agreement between survey respondents and Representatives, I find that there are salient political differences between the three income levels, and that Representatives have clear preferences for constituents of higher incomes. Additionally, this paper includes two novel findings. I find that disadvantaged groups are able to improve representation through voting, but do not have the same power as other groups in expanding representation through other means. Additionally, low-income constituents have outsized influence at the extremes of policy representation, but high-income constituents continue to have the largest influence within the margins.  

Theory and Expectations 

Numerous studies have illustrated the influence of socioeconomic factors on policy representation (Bartels 1998; Verba, Schlozman, and Brady 1995). These factors fall along historical lines of discrimination, combining with unequal representation based on race (Griffin and Newman 2005), gender (Griffin, Newman, and Wolbrecht 2012), and sexual orientation (Lax, Phillips, and Krimmel 2009). These differences have the effect of reinforcing unequal representation, further entrenching social divisions through policy. 

The current paradigm holds that these representational inequalities are the results of differences in ‘voting power’—the different abilities of any given individual to influence the outcome of an election (Bartels 2008). This creates a bias toward constituencies with higher rates of voter turnout (Burnham 1987) and higher rates of swing voting (Bartels 1998). These divisions will tend to create selective incentives for Representatives to prioritize different constituencies over one another (Rosenstone and Hansen 1997, Verba and Nie 1972, Wolfinger and Rosenstone 1980).  

One of these divisions is income level, where lower turnout among lower-income constituents leads to undesirable policy (Piven and Cloward 1988, Gilens 2011). However, some studies have indicated that increased voting behavior has a correlation with better representation (Lijphart 1997, Fiorina 1974). This would suggest that increased voting by disadvantaged groups would—to an extent—improve representation.  

Likewise, Representatives are only able to agree with their constituents on policy if they are aware of their constituents’ preferences. Those constituents who are more effective and likely to communicate their preferences should be enjoy better representation. Numerous studies have pointed to this mechanism as another contributor to income-based representational inequality, as upper-class constituents are likely to be better informed and more effective at communicating their preferences (Miller and Stokes 1963; Verba 2003; Verba, Schlozman, and Brady 1995). Increased attention to current events and increased contact with a Representative’s office has the potential to improve representation for disadvantaged groups.  

Likewise, fundraising has become an essential aspect of modern political campaigns (Powell 2015). Representatives are forced to constantly fundraise in order to fend off challengers and prepare for an expanded and costly election cycle (Heberlig and Larson 2014). This creates an incentive for Representatives to prioritize campaign donors, as these constituents have an outside influence on their reelection success. 

Likewise, recent research has illustrated that many Representatives are themselves from the upper-class. This may influence the way they perceive legislation (Carnes 2013) and may shift policy in favor of the wealthy (Griffin, Newman and Wolbrecht 2012). 

Finally, numerous studies have illustrated that constituents form opinions of their Representatives based on the roll-call votes they cast (Ansolabehere, Snyder, and Stewart 2001, Wright 1978). These constituents are then likely to hold their Representatives accountable for these votes (Erickson 1971; Canes-Wrone, Brady, and Cogan 2002). Therefore, I expect to find that Representatives with low policy agreement tend to face higher electoral penalties.  


Even by the standards of social scientists, socioeconomic class is a difficult concept to quantify. It includes one’s profession, tastes, speech patterns, and neighborhood. The most salient factor, however, is one’s income. According to the Pew Research Center (Brown 2016), there are substantial cultural rifts between those receiving less than 67% of the median household income (low-income), between 67% and 200% of the median household income (middle-income), and receiving 200% or above the median household income (high-income). While not encompassing of all the complexities of class, these distinctions should serve as effect proxies for the class divides between respondents. 

Independent Variable 

This paper’s method of measurement for policy agreement is the win-ratio for groups of constituents on notable pieces of legislation. In the Cooperative Congressional Election Survey (CCES), a large-scale nationally stratified survey administered through the website YouGov, respondents are asked their preference on five to twelve1 prominent pieces of legislation considered by the U.S. House of Representatives. These preferences are then compared to the actual roll call votes cast by that respondent’s Member of Congress. This produces a ratio of the Member’s policy agreement with each individual constituent.2 These ratios can then be averaged over every member of a group to produce a decimal value between 0 and 1, which can then be multiplied by 100 to produce a percentage value.  

This method has four primary benefits. First, it focuses exclusively on representation from a policy perspective. Policy representation—as opposed to, for example, securing appropriations—forces Representatives to make zero-sum decisions, with incentives changing in different contexts. Second, the CCES only considers prominent legislation. This increases both the likelihood that constituents will have formed coherent preferences, and the potential impact of the policy in question. Third, the CCES includes between 30,000-50,000 respondents. This allows for comparisons within a single congressional district, as well as increases the statistical confidence in our findings. Finally, creating win-ratios allows for exact comparisons across groups. This method, therefore, can be considered an accurate barometer of how much a Representative prioritizes actually translating a specific group’s preferences into law. 

This measure also has its drawbacks. Representation is a multifaceted concept. Constituents think of their Representatives as advocates, local celebrities, ombudsmen, and even personal friends (Ansolabehere and Jones 2013). This measure of representation does not include legislation that a Representative introduces, nor how much federal funding a Representative brings to his or her district. It certainly does not include a Representative’s success as a civic or moral leader. For the purposes of this investigation, however, the win-ratio provides the clearest available measure of policy representation.  

Dependent Variable 

The dependent variable in this investigation is the electoral consequences that Representatives face for their voting decisions. Electoral outcomes are measured as the percentage change in the margin of victory for the incumbent’s party from the previous and current elections. These values are then subtracted from the average of all districts represented by the same party in the Congress proceeding the election. This measure is therefore able to measure relative changes in margins of victory (as opposed to a categorical variable of whether or not an incumbent was reelected), and minimizes the impact of national trends through adjusting to the national two-party vote. 

Electoral margins are of course subject to many variables. Much research has been done investigating how the state of the economy (Sides and Vavreck 2012), the quality of a challenger (Bond, Covington and Fleisher 1985), media attention (Schudson 2002), and campaign spending by outside groups (Farrar-Myers and Skinner 2012) may affect reelection margins. By comparing Representatives against members of their own party as well as against their previous electoral performances, this measure likely encapsulates many of these exogenous variables.   

Descriptive Characteristics 

This investigation also includes numerous descriptive qualities, such as a respondent’s income, voting status, and attention to current events. These identifiers are self-reported through the CCES by the respondents and treated as categorical variables.  

It is worth noting that CCES does not have a means to verify the accuracy of these reports. It is therefore possible that some respondents have misreported their status in these variables in favor of more socially acceptable answers. Any misreported descriptors, however, will decrease the magnitude of any effect, leading toward the null hypothesis.  

There are three prominent questions addressed in this investigation: the political differences between different income levels, the preferences toward any given income a Representative may exhibit, and the mechanisms and effectiveness of disadvantaged groups at holding their Representatives accountable.  

The first question—the political differences between incomes—will be addressed through averages on three survey questions: a measure of general ideology, a measure of economic preferences, and a measure of social ideology. General ideology will be measured as the average place on a seven-point party identification question. Economic policy is measured by the average response to a question over whether to cut spending to decrease the national deficit. Higher scores indicate an increased preference to cut spending. Finally, social ideology will be measured by opinions on abortion, with increased scores indicating a preference toward greater legality and fewer restrictions.  

Representatives’ selective decisions will be measured by taking the average win-ratio for the three income categories. This will be further broken down by party affiliation, to investigate ideological influence in policy decisions.  

There are four mechanisms of improving representation which will be considered: voting, knowledge of current events, whether a respondent has contacted a Representative’s office within the last year, and whether a respondent has donated to a political campaign in the last two elections. The interactions of these mechanisms for each group will be explored through a multivariate regression model. Finally, the win-ratios of all three categories will be considered against the relative change in margin of victory for the incumbent’s party. This measure will provide a final, outside measure of the relative effectiveness of each group in influencing electoral outcomes. 


My investigation proceeds in three steps. First, I seek to illustrate the differences between low-, middle-, and high-income respondents. I then investigate to what extent, if any, Representatives hold selective preferences for one group over another. Finally, I examine four common mechanisms of improving representation and test the relative impacts of changes in policy agreement on changes in electoral success.  


In order for Representatives to make discriminatory decisions, and for citizens to hold Representatives accountable, respondents of different income levels must hold distinct political preferences from one another. This will be measured through three tests: general ideology, economic preferences, and social preferences.  

Figure 1. 

Figure 1. illustrates that increases in income correspond to more conservative general ideological preferences, while decreases in income correspond to more liberal ideological preferences. Low-income respondents were 0.2 points more liberal than middle-income respondents, and over 0.3 points more liberal than high-income respondents.  

Figure 2.  

Columns represent the average response on a 100-point scale of whether cut spending or raise taxes to finance a budget deficit.  

Likewise, Figure 2. demonstrates that low-income respondents are 4% more averse to cutting domestic spending to finance a budget deficit than were middle-income respondents, and 5% more averse than high-income respondents.  

Figure 3. 

Average response on a four-point scale of whether to keep abortion completely legal with no restrictions, legal with some restrictions, partially legal with many restrictions, or completely illegal. Higher scores indicate preferences toward legality and lack of restrictions. 

Finally, Figure 3. indicates that, despite appearing more liberal in general and economic ideology, low- and middle-income respondents were more conservative on social issues such as abortion. 

Selective Representation 

Having established that constituents of different incomes hold divergent political beliefs, I now examine whether Representatives exhibit selective biases in how they represent different constituents.  

Figure 4.  

Average policy agreement for three income groups in the 110th Congress. Error bars represent 95% confidence intervals. 

As is evident, a clear bias exists in favor of higher-income respondents. Examining Congress as a whole, high-income respondents tend to enjoy 2 points higher policy representation than low-income respondents, and 1 point higher than middle-income respondents. All results are statistically significant to a 95% confidence interval.  

Figure 5. 

Figure 5 further examines the apparent selective representation, breaking down Representatives by party affiliation. Democrats as a whole had better policy representation, although this is to be expected from the majority party (Cox and McCubbins 1993). While Republican Representatives continued to prefer higher income respondents over low-income respondents, Democratic Representatives exhibited a slight preference for low- and middle-income respondents. However, the bias exhibited by Democratic Representatives was less drastic than that exhibited by Republican Representatives, potentially explaining why low- and middle-income respondents tended to be disadvantaged in general. 

Accountability Mechanisms 

There are four mechanisms I will examine that may mitigate the bias exhibited in the previous section: voting, attention to current events, contacting their Representative directly, or becoming involved politically.  

Figure 6. 

Figure 6 demonstrates that there are rewards for voters of all income categories. High-income respondents received the most relative rewards, with high-income voters receiving 4 points higher representation than high-income non-voters. Low-income respondents also experienced gains, with voters receiving just over 2 points better representation than nonvoters. However, the gains received by low-income voters were not enough to reach either category for middle-income respondents, who experienced only a slight representational benefit from voting. 




Figure 7. 

Likewise, as Figure 7 indicates, low-income respondents received incremental improvements in representation for increased attention to current events, with high-awareness respondents averaging 3 points higher than respondents with no awareness. This gain, however, was not enough for low-income respondents to reach representational equity with middle- or high-income respondents of any awareness. These categories—middle- and high-income—did not exhibit a consistent link between awareness and representation.  

Figure 8 

The third mechanisms by which disadvantaged groups may influence their Representative is through contacting their Representative directly. This can be done either through calling the Representative’s office, writing a letter or signing a petition, or attending a town-hall. There were notable benefits to doing so for middle- and high-income respondents, with those having contacted an office receiving 2 point and 1.5 point improvements, respectively. For low-income respondents, however, those having contacted their Representative actually received slightly lower policy agreement.  

Figure 9. 

The final method for disadvantaged groups to improve representation is through donating to a political campaign directly. There was some slight representational benefit to doing so for middle-income respondents, with donors receiving 1 point higher policy agreement than nondonors. For low-income and high-income respondents, however, there was no link.  

Table 1. Multivariate Regression by Income 





Voting Status 







Current Events 







Contact with Office 





















Multivariate Regression model between policy agreement and voting status, current events awareness, contact with office and donations. *p<0.1, **p<0.01, ***p<.0.001 

As Table 1 demonstrates, the only statistically significant method of improving representation for low-income respondents is through increased voting. However, the gains from voting continue to be highest for high-income constituents. Low-income respondents have the highest rewards for increased media attention, but this change is insignificant. Middle- and high-income respondents have very high correlations between representation and contact with their representative’s office—there was little correlation for low-income respondents. Finally, no income level received notable correlations between political donations and representation. 




Political Consequences 

Finally, I examine whether Representatives face electoral consequences for differences in their roll-call voting.  

Figure 10 indicates that, for all income levels, there is a positive relationship between policy agreement and relative change in incumbent party’s vote margin.  

There are two striking observations from this data. First, low-income constituents have outsized impacts on electoral consequences at the extremes of representation, with Representatives agreeing with low-income constituents above 0.85 receiving disproportionate benefits, and Representatives agreeing with low-income constituents below 0.3 receiving disproportionate costs.  

Within the margins, however, all three income levels appear to advance semi-linearly.3 High income respondents have the highest slope, followed by middle-income respondents, and with low-income respondents with the lowest slope.  

This indicates that, within the margins of 0.4 and 0.8, Representatives receive a higher marginal vote share for representing constituents of increased incomes. Low-income respondents, therefore, are able to hold their Representatives accountable, but only outside the normal range of representation.  


The purpose of this investigation was to investigate the representational differences between constituents of different incomes, how constituents can improve their representation, and whether constituents hold Representatives accountable for differences in policy agreement.  

This study was not a perfect measure of accountability. The Congresses examined both included heavy Democratic majorities, and did not examine nonelection years. Both of these factors would have worked in favor of low-income policy agreement, possibly diminishing the significance of any findings.  

Likewise, each piece of legislation was treated as equal. There was no measure for the issue’s salience between groups, the media attention paid toward the bills, or whether the legislation was ultimately enacted.  

Finally, both elections examined featured huge partisan swings. These swings had many exogenous influences, such as the Global Financial Crisis or the Tea Party movement, which were not accounted for. Likewise, district characteristics were not accounted for, nor were challenger quality and differences in turnout between groups. 

There were several prominent findings. As many scholars have noted, large representational gaps exist between constituents of different incomes. These effects are compounded by party, with Democrats agreeing more with low-income constituents, and Republicans agreeing with high-income constituents.  

Low-income constituents are able to improve their representation through voting, but are ineffective at improving agreement through other means.  

Finally, low-income respondents have a disproportionate effect on electoral outcomes at the margins, yielding huge rewards for extremely high representation and huge punishments for extremely low representation. Within, the margins, however, higher-income constituents are still more effective at holding their Representatives accountable.  

Both of these novel findings hold significant normative value. How can disadvantaged groups improve representation if the mechanisms do not work equally for different groups? How can these groups hold their Representatives accountable if their influence is only effective at the margins? Are these findings consistent with the principles of democracy, or in the intent behind a republican form of government? 

If, as Martin Luther King famously stated, “the moral arc of the universe is long, but it bends toward justice,” those groups disadvantaged by biases in representation must be capable of altering the system. As this investigation demonstrates, that capacity is rarely realized.  


Ansolabehere, Stephen, James M. Snyder, and Charles Stewart. "Candidate Positioning in U.S. House Elections." American Journal of Political Science 45, no. 1 (2001): 136.  

Bartels, Larry M. "Electoral Continuity and Change, 1868–1996." Electoral Studies17, no. 3 (1998): 301-26.  

Bartels, Larry M. Unequal Democracy: The Political Economy of the New Gilded Age. New York: Russell Sage Foundation, 2008. 

Bond, Jon R., Cary Covington, and Richard Fleisher. "Explaining Challenger Quality in Congressional Elections." The Journal of Politics 47, no. 2 (1985): 510-29.  

Brown, Anna. "What Americans Say It Takes to Be Middle Class." Pew Research Center RSS. 2016. Accessed April 27, 2016.

Burnham, Walter Dean. "Elections as Democratic Institutions." Society 24, no. 4 (1987): 38-48. doi:10.1007/bf02695792. 

Canes-Wrone, Brandice, David W. Brady, and John F. Cogan. "Out of Step, Out of Office: Electoral Accountability and House Members' Voting." Am. Pol. Sci. Rev. American Political Science Review 96, no. 01 (2002). doi:10.1017/s0003055402004276. 

Carnes, Nicholas. White-collar Government: The Hidden Role of Class in Economic Policy Making. 

Cox, Gary W., and Mathew D. McCubbins. Legislative Leviathan: Party Government in the House. Berkeley: University of California Press, 1993. 

Enns, Peter, and Christopher Wlezien. Who Gets Represented? New York: Russell Sage Foundation, 2011. 

Erickson, Robert S. "The Electoral Impact of Congressional Roll Call Voting." The American Political Science Review 65, no. 4 (1971): 1018.  

Farrar-Myers, Victoria A., and Richard Skinner. "Super Pacs and the 2012 Elections." SSRN Electronic Journal SSRN Journal.  

Fenno, Richard F. Home Style: House Members in Their Districts. Boston: Little, Brown, 1978. 

Fiorina, Morris P. Representatives, Roll Calls, and Constituencies. Lexington, MA: Lexington Books, 1974. 

Flavin, Patrick. "Does Higher Voter Turnout among the Poor Lead to More Equal Policy Representation?" The Social Science Journal 49, no. 4 (2012): 405-12. 

Gilens, Martin. Affluence and Influence: Economic Inequality and Political Power in America. Princeton, N.J: Princeton University Press, 2012. 

Griffin, John David, and Brian Newman. Minority Report: Evaluating Political Equality in America. Chicago: University of Chicago Press, 2008. 

Griffin, John D., Brian Newman, and Christina Wolbrecht. "A Gender Gap in Policy Representation in the U.S. Congress?" Legislative Studies Quarterly 37, no. 1 (2012): 35-66. 

Heberlig, Eric S., and Bruce A. Larson. "U.S. House Incumbent Fundraising and Spending in a Post- Citizens United and Post- McCutcheon World." Political Science Quarterly 129, no. 4 (2014): 613-42.  

Lax, Jeffrey R., and Justin H. Phillips. "Gay Rights in the States: Public Opinion and Policy Responsiveness." Am Polit Sci Rev American Political Science Review103, no. 03 (2009): 367. 

Lijphart, Arend. Thinking about Democracy: Power Sharing and Majority Rule in Theory and Practice. London: Routledge, 2008. 

Mayhew, David R. Congress: The Electoral Connection. New Haven: Yale University Press, 1974. 

Miller, Warren E., and Donald E. Stokes. "Constituency Influence in Congress."Am Polit Sci Rev American Political Science Review 57, no. 01 (1963): 45-56. 

Munro, Gary. "Who Votes? Raymond E. Wolfinger and Steven Rosenstone New Haven: Yale University Press, 1980, Pp. X, 158." Can J Pol Sci Canadian Journal of Political Science 14, no. 03 (1981): 672. 

Polsby, Nelson W., Frances Fox Piven, and Richard A. Cloward. "Why Americans Don't Vote." Contemporary Sociology 17, no. 6 (1988): 784.  

Rosenstone, Steven J., John Mark Hansen, and Donald R. Kinder. "Measuring Change in Personal Economic Well-Being." Public Opinion Quarterly 50, no. 2 (1986): 176. 

Schickler, Eric, and Frances E. Lee. The Oxford Handbook of the American Congress. Oxford: Oxford University Press, 2011. 

Schudson, Michael. "The Business of Media: Corporate Media and the Public Interest." The Information Society 18, no. 4 (2002): 307-08. 

Sides, John, and Lynn Vavreck. The Gamble: Choice and Chance in the 2012 Presidential Election. 

Verba, Sidney, and Norman H. Nie. Participation in America: Political Democracy and Social Equality. New York: Harper & Row, 1972. 

Verba, Sidney, Kay Lehman Schlozman, and Henry E. Brady. Voice and Equality: Civic Voluntarism in American Politics. Cambridge, MA: Harvard University Press, 1995.