Ideological Incongruence and Trust in Congress

Ideological Incongruence and Trust in Congress Justin H. Kirkland jhkirkland@uh.edu Kevin K. Banda bandak@missouri.edu Abstract Citizens perceive of their legislators as agents acting on their behalf and, much like other principal-agent dynamics, should trust Congress at higher levels the more their legislators behave as citizens wish. While citizens beliefs about the similarity between themselves and their legislators should be affected by their legislators behavior in Congress, their trust in the institution should be more powerfully informed by their perceptions of similarity than by actual legislative behavior. We test this conception of trust in Congress using survey data from the 2008 Cooperative Congressional Election Study and find support for our theory. Citizens trust in Congress is strongly affected by their perception of the ideological distance between themselves and their representatives. We further find no evidence that legislators roll call voting affects citizens trust in Congress after accounting for citizens perceptions of ideological distance. Department of Political Science, University of Houston Department of Political Science, University of Missouri

Democratic governments cannot function efficiently if the citizens they govern are distrustful of government institutions. The trust citizens hold toward their government is fundamentally affected by the actions taken by the government. At its core, democracy requires that citizens exchange some degree of consent for coercion in the form of government action. Citizens who trust government should be willing to cede greater authority to the government, which in turn should allow the government to take more actions because office holders who are trusted should find it easier to successfully act than should those who are distrusted (Hetherington, 1998, 2005). Extant research suggests that citizens respond to government activity by altering the degree to which they trust the government. Scholars also report results suggesting that trust varies in response to a number of factors such as the behavior of political actors, the partisanship of individual citizens, and the performance of the national economy. Despite much scholarly interest in the behavioral and attitudinal effects of trust in government and the causes and consequences of citizen and elite policy preferences, we do not yet know how citizens perceptions of the ideological dispositions of elected officials affect the proclivity of citizens to express trust in the institutions of which these officials are members. Most recent work on trust in government has focused on economic performance or government transparency as explanations for political trust (see Stimson (2004) and Tolbert and Mossberger (2006) for examples). In contrast to these aggregate arguments, we offer an individual level theory that argues that citizens perceptions of the ideological distances between themselves and their legislators is a key determinant of the degree to which they report trusting Congress. Citizens who feel proximal to their legislators view themselves as being similar to these legislators and should therefore be more willing to trust Congress as a whole. They should do so because they believe that ideologically proximal legislators are more likely to behave in accordance with the citizens own preferences. 1

We test our theory of the formation of trust in Congress with data from the 2008 Cooperative Congressional Election Study. We find that perceived ideological distance between citizens and their representatives informs citizens trust in Congress. As citizens view their legislators as being increasingly ideologically distant from themselves, they express lower levels of trust in Congress. Furthermore, these trust attitudes appear to be driven more powerfully by citizens perceptions than by the similarity between legislators actual voting behavior and the preferences of citizens. Our finding that perceptions dominate reality in affecting trust attitudes suggest that scholars of public opinion should take care when examining the ways that elite behavior informs the attitudes of citizens. 1 Political Trust Scholars typically conceptualize trust as an evaluative orientation directed at the political system as a whole (Stokes, 1962; Easton, 1965). Miller (1974) argued that this evaluation is driven in large part by the normative expectations that citizens hold about how the government has and should operate. Trust can be further divided into political and social trust. The former refers to trust felt towards government, politicians, and institutions while the latter refers to interpersonal trust. We focus on political trust. Political trust is an important concept for several reasons. First, citizens who feel that the government is trustworthy are more likely to comply with and consent to government actions and regulations (Levi, 1988; Tyler, 1990; Scholz and Pinney, 1995; Scholz and Lubell, 1998a,b; Levi, 1997; Tyler, 1998). A government that is not trusted may find its behavior constrained because citizens will not grant it as much leeway. Governments that are trusted, on the other hand, may operate more freely. Trusted governments, then, may be granted more discretion by citizens in the policy making process. Second, citizens who exhibit low levels of political trust and high levels of political efficacy are more likely to participate in politics 2

than are other citizens (Easton, 1965; Gamson, 1968; Bandura, 1982). Third, distrust tends to generate negative evaluations of incumbents (Sigelman, Sigelman, and Walkosz, 1992; Hetherington, 1998, 2005). It furthermore affects individuals vote choices by encouraging citizens to support challengers and, when they are electorally viable, third party candidates (Aberbach, 1969; Rosenstone, Behr, and Lazarus, 1984; Hetherington, 1999). Additionally, research on descriptive representation implies that when a representative looks like (or shares obvious physical traits with) a constituent, that constituent is more likely to participate in politics (Bobo and Gilliam, 1990) and trust government (Abney and Jr., 1981; Howell and Fagan, 1988; Gay, 2002; Scherer and Curry, 2010). Finally, aggregate levels of political trust in the U.S. are generally low (Rosenstone et al., 1993) in part due to low levels of social capital (Keele, 2007) and views of political institutions, especially Congress, have become increasingly negative over the past several decades (Craig, 1993; Hibbing and Theiss-Morse, 1995). Taken as a whole, the existing literature suggests that political trust can affect both the attitudes and the behaviors of citizens. Research on descriptive representation suggests that when a representative looks like (or shares obvious physical traits with) a constituent, that constituent is more likely to participate in politics (Bobo and Gilliam, 1990) and trust government (Abney and Jr., 1981; Howell and Fagan, 1988; Gay, 2002; Scherer and Curry, 2010). While we do not wish to conflate descriptive with substantive representation, we do wish to extend this logic beyond obvious physical characteristics and consider how perceptions of ideological similarity influence trust in government institutions. 1.1 The Origins of Political Trust Scholars suggest that feelings of trust in government are influenced by a number of factors. Presidential approval (Hetherington, 1998) and citizens perceptions of the president s 3

personal characteristics (Citrin and Green, 1986) both contribute to the formation of trust attitudes. Citizens tend to view the president as the face of the national government, so when they approve of the president and attribute positive characteristics to him they also tend to express higher levels of trust in the government. More generally, the behavior of office holding politicians along with the perceptions citizens hold about the performance of government appear to inform the degree of trust citizens feel towards government. The actions taken by both the president and Congress along with the performance of the government affect trust in government at the national level (Erber and Lau, 1990; Craig, 1993; Keele, 2005). Scandals decrease political trust (Chanley, Rudolph, and Rahn, 2000), likely in part because of the high levels of attention given to them by the news media (Orren, 1997). The state of the national economy also influences trust; when citizens believe that the economy is strong, trust in government increases (Citrin and Green, 1986; Hetherington, 1998; Chanley, Rudolph, and Rahn, 2000). Trust similarly decreases when citizens become less optimistic about the economy. Finally, higher crime rates also appear to lead to lower levels of trust in government (Mansbridge, 1997). Citizens should be more likely to express feelings of trust when their preferred party controls the government or an institution such as Congress (Citrin, 1974; Keele, 2005) because they filter information in a partisan manner (Stokes, 1966; Kunda, 1990; Taber and Lodge, 2006) and make use of partisan heuristics. They use partisan stereotypes and party cues as information shortcuts when forming and altering attitudes about political figures (Page, 1978; Conover, 1981; Wright and Niemi, 1983; Hamill, Lodge, and Blake, 1985; Hurwitz, 1985; Granberg, Kasmer, and Nanneman, 1988; Jacoby, 1988; Riggle et al., 1992; Rahn, 1993), so partisan control of an institution should inform the way citizens view the institutions. A citizen whose party controls both chambers of Congress, for example, should trust Congress to a greater extent than should a citizen whose party is in the minority. This is because the citizen whose party is in control can reasonably expect Congress to generate policies that are 4

generally congruent with her preferences. A citizen whose party is in the minority, on the other hand, can reasonably expect the legislature to produce consistently displeasing policy outcomes. 1.2 Political Trust and Representation Elected officials are charged with representing their constituents before government, but they also symbolize government to their constituents. That is, for many constituents in a district or state, their elected officeholder is one of the primary people they associate with government generally and the institution of which the officeholder is a part specifically. The more an officeholder represents the interests of a citizen, the more she should trust the construct the officeholder embodies. We begin by assuming that the representative-constituent relationship is similar to a traditional principal-agent relationship. That is, principals (constituents) elect an agent (representatives) to act on their behalf. These agents wield expertise the principal cannot, but are subject to performance evaluations and monitoring from the principal in the form of elections. The agent in this circumstance has many principals, but each grants her a great deal of discretion. Monitoring costs are high because people pay little attention to politics (Campbell et al., 1960; Converse, 1964) and generally do not understand political debates (Delli Carpini and Keeter, 1996), but agents who behave against the wishes of the principal are often punished electorally (Canes-Wrone, Brady, and Cogan, 2002), perhaps because their incongruent behavior is communicated to citizens by the news media and the officeholders political opponents. This implies that principals invest agents with some degree of discretion, thus allowing them to act without being directly monitored at all times. Constituents elect legislators to represent their policy concerns before government and develop an expectation that the 5

legislator will do just that. This further implies that agents who are perceived to have satisfied those expectations more frequently are more likely to earn trust from their principals. In other words, legislators who are perceived as being more ideologically proximal to their constituents should be trusted to a greater extent by those constituents. We argue that the perceived ideological distance between citizens and their legislators is a key determinant of the degree to which citizens believe that Congress is trustworthy. We make this argument because citizens who believe themselves to be more ideologically proximal to their legislators should be increasingly willing to cede authority to the legislator, who they believe will behave in accordance with the citizen s own preferences. This in turn should lead citizens to feel higher levels of trust in Congress. In this sense, legislators act as a sort of symbolic figurehead of the institution in the minds of citizens. The traditional proximity model as explicated by Downs (1957) provides us with a useful theoretical outline of the process citizens go through when deciding the degree to which they trust Congress. In a Downsian world, citizens are attracted to parties and candidates with whom they believe they are ideologically similar. In other words, citizens prefer candidates and parties they think are closer to them (but see Rabinowitz and Macdonald, 1989, for a different set of theoretical expectations). We argue that the distance citizens perceive between themselves and their legislators should affect the degree to which they trust Congress. Citizens who feel that their interests are being represented, i.e. those who feel they are close to their legislators, should express higher levels of trust than should those who feel their interests are not being represented. To put it another way, the level of trust in Congress expressed by a citizen should increase as the ideological distance they perceive between themselves and their legislator decreases. Thus, our theory yields the following hypothesis: H 1 : As the perceived ideological distance between citizens and their legislators increases, citizens levels of trust in Congress will decrease. 6

1.2.1 Legislative Behavior, Perceptions, and Political Trust Our argument to this point has focused on citizens perceptions of the distance between their own ideologies and those of their legislators. But how are these perceptions formed? Citizens perceptions of the ideological distance between themselves and their legislators should be informed by legislators behavior while in office. While it is true that citizens generally know little about politics in large part due to low levels of political interest (Campbell et al., 1960) and tend to misunderstand contemporary policy debates (Delli Carpini and Keeter, 1996), it is also the case that (1) the news media often covers votes on salient legislation and (2) legislators try to establish reputations over time by making use of position-taking techniques (Mayhew, 1974). These reputations may be ideological in nature. Citizens may therefore respond to signals from the news media and from elites when assessing the ideological location of their legislators. In other words, citizens should view legislators who vote contrary to their preferences as being ideologically distant from them compared to those whose votes are consistent with citizens preferences. Put more formally: H 2 : As a legislator s behavior becomes increasingly divergent from a citizen s preferences, their perceptions of the legislator s ideological distance should increase. We further argue that legislators voting behavior should not inform citizens attitudes of trust in Congress, at least when controlling for citizens perceptions of ideological similarity. This is because citizens perceptions of legislators should dominate the actual behavior of those legislators. In other words, this behavioral congruence of legislators the degree to which the a legislator s voting behavior is similar to the preferences of a citizen should not inform the trust a citizen expresses towards Congress once that citizen s perceptions of the legislator are accounted for. Citizens perceptions of ideological congruence should be built up over time as legislators represent the districts or states in which citizens reside, advertise, and use various position taking strategies. A legislator s voting behavior, on the other 7

hand, can be transitory in nature and any deviation from the norm they have established at least on salient legislation should be reported by the news media. Changes in voting behavior should be followed by a changes in citizens perceptions of legislators. In other words, citizens beliefs about the ideological dispositions of legislators should be more important than legislative behavior in generating trust attitudes. The hypothesis generated by this argument follows: H 3 : Behavioral congruence should not affect citizens level of trust in Congress when perceptions are accounted for. 2 Research Design We use data from the 2008 Cooperative Congressional Election Study (CCES) to test our theory. These data come from the University of Missouri team content module of 1,000 CCES participants (see Richardson Jr., Konisky, and Milyo, 2012). Respondents were asked to answer a variety of questions about trust in government including the degree to which they trusted the U.S. Congress. Responses were categorized into four levels ranging from zero to three in which the lowest levels of trust in Congress were represented by the lowest numerical value. The CCES asked respondents to place themselves on a zero to one hundred point scale with 0 representing extremely liberal and 100 representing extremely conservative. The survey also asked respondents to place a variety of political figures on this scale including their member of the U.S. House and their Senators. We used these scales to measure respondents perceptions of ideological distance between themselves and their legislators. Our measure of perceived ideological distance is simply the absolute difference between a respondent s self placement and her placement of her legislators. This measure ranges from zero perfect congruence to one hundred absolute incongruence. 8

Our theory specifies that a respondent s perceptions of ideological incongruence with her legislators will influence her level of trust in Congress and that behavioral divergence between a legislator and a citizen s stated preferences will have little effect on trust in Congress. In other words, ideological incongruence only matters so long as respondents perceive it. Fortunately, respondents to the 2008 CCES expressed their positions on eight salient roll call votes in the prior session of the U.S. Congress. 1 We constructed a measure of behavioral divergence by adding the number of times a respondent s member of Congress voted the way the respondent wished and divided that total by the number of responses the respondent actually provided. Because some respondents stated preferences for roll call votes more often than others, the denominator in this percentage varies across respondents. We control for respondents sex using a dummy variable coded one if the respondent was female and zero if they were male, age in years, religiosity using a six point ordinal scale for which higher values indicated higher levels of religiosity, expectations for the national economy a five point scale for which higher values indicate worsening economic perceptions, education using a six point ordinal scale in which higher values indicate higher levels of educational educational achievement, race and ethnicity using dummies indicating whether or not a respondent identifies as black or Latino, and respondents degree of political interest using a three point scale on which higher values indicated higher levels of interest. We also include the ideologies of respondents as measured by a five point scale on which higher values indicate increasing levels of ideological conservatism. Finally, we control for respondents party identities and the partisanship of their legislators. The focus of our theory is on perceived ideological incongruence, which should be correlated with co-partisanship, thus we include an interaction term that estimates the effect of party identification when a 1 These votes were for the withdrawal of troops from Iraq, increasing the minimum wage, federal funding for embryonic stem cell research, a bill allowing U.S. agents to eavesdrop on foreign operatives without a court order, health insurance for children in low income families, consideration of an amendment against gay marriage, federal assistance for homeowners facing foreclosures, and the extension of NAFTA to include Peru and Columbia. 9

Table 1: Summary Statistics for Ideological Placement, Trust in Congress, and Control Variables Variable Mean Standard Deviation N Trust in Congress 1.684 0.680 861 Ideological distance to House Member 32.349 27.489 495 Ideological distance to senior Senator 33.727 28.154 483 Ideological distance to junior Senator 34.370 28.695 513 Roll Call divergence with U.S. House Member 0.546 0.283 995 Age 47.5 16.14 893 Religiosity 2.246 1.695 889 Education 3.363 1.494 1000 Interest in politics 1.401 0.615 885 Expectations for the national economy 4.505 0.669 889 Self-reported ideology 3.195 1.195 840 Respondent is black 0.061 0.239 1000 Respondent is Latino 0.063 0.243 1000 Copartisanship House Member 0.529 0.499 918 Senior Senator 0.502 0.500 910 Junior Senator 0.513 0.500 897 Note: There were 454 women in the sample and 439 men. There were also 448 Democrats and 410 Republicans. There were 107 independents/non-partisans in the sample. N represents the number of respondents who provided usable responses. respondent and her representative are of the same party. Table 1 provides summary statistics for each of our variables of interest. The dependent variable in our analysis is our four category measure of trust in Congress for which higher values indicate higher levels of political trust. We model political trust in Congress using ordered logistic regression. Ordered logit includes cutpoints that provide unique intercepts based on the probability that an observation will fall into each of the observable categories. Thus, a negative coefficient in an ordered logistic regression implies that the as an independent variable increases, the probability that a respondent falls into the lowest category of the dependent variable also increases, while the probability of a respondent falling into the higher categories of the dependent variable decreases. 10

We begin our analysis by demonstrating that perceived ideological distance is related to trust in Congress. We follow that up with a series of models that estimate the influence of roll call divergence on perceived ideological distance, and subsequently the effect of both roll call divergence and perceived ideological distance on trust in Congress. Again, our expectations are that perceived ideological distance influences trust in Congress, and that controlling for perceptions, actual roll call divergence will play little roll in structuring trust attitudes. We estimate separate models for the perceived distance between a respondent and each of her members of Congress, and provide a single unified model in the supplemental appendix. 3 Results Table 2 presents results from an ordinal logistic regression predicting respondents degree of trust in the U.S. Congress. Higher values of the dependent variable indicate higher levels of trust. The key variable of interest is the respondents perceived ideological distance from their legislator. As control variables, we include a dummy variable indicating whether the respondent is a Democrat, a dummy variable indicating whether the respondent is a Republican, a dummy variable indicating whether a respondent s legislator is a Democrat, a dummy variable indicating whether the respondent is female, the respondent s age in years, the respondent s reported level of religiosity, the respondent s expectations for the national economy, the respondent s interest in politics, the respondent s education level, dummy variables indicating the respondent s race, the respondent s ideology on a 7-point scale, and interactions of the respondent s party identification and that of their representative. These interaction terms allow us to capture the change in a respondent s trust in Congress when her representative is of the same party as she is. Column (1) presents results using the perceived ideological distance between a respondent and her House member. Columns (2) and (3) use the distance between a respondent and her senior and junior Senators, respectively. 11

In each of the three models, the effect of a respondent s perceptions of the distance between herself and her legislator is negative and significantly related to trust in the U.S. Congress. This indicates that as a respondent perceives of a greater distance between herself and her member of Congress, she trusts the entire institution of Congress less. 2 The only other consistently significant predictors of trust in Congress are a respondent s interest in politics and the Democratic respondent identifier variable. A respondent s interest in politics is consistently negatively associated with trust in Congress indicating that controlling for the other variables in the model, respondents more interested in politics trusted Congress less. The coefficient on the Democratic respondent identifier indicates that on average, Democratic respondents trusted Congress more than nonpartisan respondents. 3 Ordinal logistic regressions predict the probability of a response falling into each of the observable set of categories, thus a negative coefficient implies than when an independent variable increases the probability of an observation being in a low category increases while the probability of being in a high category decreases. The baseline rates of falling into each category are determined by the model s cutpoints. Since these nuances can be hard to interpret numerically, we provide predicted probability plots for the lowest level of trust in Congress as the respondent s perceptions of the ideological distance between her and her House member increase. Figure 1 presents these predicted probabilities for a white, male, Democratic respondent with a Democratic representative. 45 The respondent s age, interest in politics, education, expectations of the national economy, and religiosity were all kept 2 In the supplemental appendix, we provide a model with all three legislators included in the same regression. Despite the extremely high levels of multicollinearity this inclusion creates, ideological distance between a respondent and her House member remains a negative and significant predictor of trust in Congress. 3 The difference between the Democratic respondent and Republican respondent coefficient is statistically significant in all three models. Thus, Democratic respondents systematically trust Congress at higher levels than Republican respondents controlling for the other covariates in the model. 4 The predicted probability plots using the results from the senior and junior Senator models are nearly identical and can be furnished upon request. 5 We provide plots for changes in the perceived ideological congruence between a respondent and her senators in the supplemental appendices. 12

Table 2: Ordinal Logistic Regression Predicting Trust in the U.S. Congress as a Function of Perceived Ideological Distance from a Legislator U.S. House Member Senior Senator Junior Senator Ideological distance to legislator(s) -0.016* -0.012* -0.012* (0.004) (0.004) (0.004) Respondent is Democrat 1.902* 2.109* 2.582* (0.865) (0.738) (0.371) Respondent is Republican 0.567 0.515 0.974 (0.863) (0.724) (0.847) Representative is Democrat 0.943 1.088 0.917 (0.928) (0.827) (0.940) Respondent is female 0.311 0.266 0.255 (0.195) (0.191) (0.193) Age in years -0.001-0.004-0.0001 (0.006) (0.006) (0.006) Religiosity 0.061 0.099 0.057 (0.062) (0.062) (0.061) Interest in politics -0.522* -0.472* -0.509* (0.190) (0.181) (0.191) Education -0.033-0.050-0.021 (0.064) (0.062) (0.063) Expectations of the national economy -0.036-0.166-0125 (0.166) (0.153) (0.156) Respondent s ideology -0.329* -0.299* -0.370 (0.119) (0.119) (0.120) Respondent is black 0.849* 0.713 0.980* (0.424) (0.414) (0.429) Respondent is Latino 0.598-0.157 0.308 (0.479) (0.413) (0.455) Democrat Respondent X Democrat Representative -1.171-1.481-1.352 (0.967) (0.882) (0.999) Republican Respondent X Democrat Representative -0.346-0.871-0.716 (0.984) (0.879) (0.990) Cutpoint 1 0.376-0.431 0.288 (1.346) (1.166) (1.288) Cutpoint 2 3.109* 2.184 3.137* (1.349) (1.167) (1.292) Cutpoint 3 5.859* 5.031* 6.117* (1.432) (1.262) (1.408) N 503 493 519 Pseudo R-Squared 0.554 0.547 0.555 Note: Cell entries are estimated coefficients generated by an ordinal logit model. Standard errors are reported in parentheses. * p-value 0.05 (two-tailed test) 13

Probability of 'Hardly Ever' Trusting Congress as Respondent's Perception of Ideological Distance to U.S. House Member Increases Predicted Probability of 'Hardly Ever' Trusting Congress 0.0 0.2 0.4 0.6 0.8 0 20 40 60 80 100 Proximate To Representative Distant From Representative Perceived Ideological Distance Figure 1: Predicted Probability of Hardly Ever Trusting Congress for White, Male, Democrat as Perceived Ideological Distance Increases. Dotted Lines Represent 95% Confidence Intervals. at their mean values. Respondents falling into the lowest category of trust in Congress reported that they Hardly Ever trust the U.S. Congress. Not surprisingly, virtually no respondents placed themselves in the highest level of trust in Congress making the probability of observing anyone in that category essentially zero, regardless of citizens perceptions of ideological distance from their legislators. As the figure shows, an increase in the perceived ideological distance between a respondent and her legislator increases from its minimum observed value to its maximum observed value, the probability that that respondent will report Hardly Ever trusting the U.S. Congress increases from 0.28 to 0.54. 14

As respondents perceived of a greater distance between themselves and their House member, those respondents systematically trusted the U.S. Congress less. While this may seem intuitive, the bulk of research on trust in government institutions has focused on the performance of government in structuring citizens trust attitudes, and the research that has focused on representation s effect on trust has focused exclusively on descriptive representation. Here we have evidence that dyadic policy congruence plays a key role in the development of trust attitudes, and that representation can influence trust in Congress even when perceptions of the economic conditions, and thus, government performance are poor. 3.1 Roll Call Similarity and Perceived Congruence Our theory suggests that trust in Congress is based on a citizen s perceptions of her legislators behavior, and that controlling for those perceptions, the actual behavioral similarities between a respondent and her legislator will have little effect on that trust. In other words, a legislator s behavior influences her constituents perceptions of that legislator, and those perceptions influence trust. However, the translation of a legislator s behavior into her constituents perceptions is likely to be less than perfect. To test this assertion, we develop measures of behavioral divergence between a survey respondent and her members of Congress. The 2008 CCES asked respondents their positions on eight salient roll call votes that took place in the prior session of U.S. Congress. To measure behavioral divergence, we simply add up the number of times a respondent voted for a bill in different way than each of her legislators respectively. We then divide that sum by the number of votes the respondent provided. 6 This gives us a proportion of roll call divergence in which high values represent high levels of actual policy incongruence and low values represent low levels of policy incongruence. 6 Thus, each respondent has a behavioral and perceived divergence measure for each of her legislators. 15

The first step in testing our assertions regarding the relationship between behavioral divergence, perceived ideological distance, and trust in Congress is to establish that the actual behavioral similarities between legislators and their constituents influence constituents perceptions of ideological distance. Table 3 presents the results of an OLS regression in which the dependent variable is the perceived ideological distance between a CCES respondent and each of her members of the U.S. Congress. As independent variables in the model we include all of the control variables included in Table 2. 7 We also include an independent variable measuring the roll call divergence between a respondent her legislator. Across each of the models, as the roll call divergence between a respondent and her legislator increases, the respondent s perception of the ideological distance between herself and her legislator also increases. This pattern exists controlling for partisanship and whether respondents are of the same party as their representatives. Again, the only other consistent predictors of perceived ideological distance is a respondent s interest in politics. Thus, it appears that respondents beliefs about their representatives are informed by roll call congruence with their legislators. To provide some context for the magnitude of these coefficients, Figure 2 plots the predicted value of the perceived ideological distance between a respondent and her House member as the roll call divergence between them increases. As roll call divergence moves from its minimum value to its maximum value, the predicted value of perceived ideological distance moves from 15.34 to 26.14. This is roughly equivalent to a jump from the 35th percentile to the 50th percentile in perceived ideological distance. Thus, we see that roll call divergence plays an important role in predicting perceived ideological distance, but does so imperfectly. Even a jump from the minimum to the maximum level of roll call divergence does not move perceived divergence to anything near its maximum value. This is of course a function of where we set the remaining control variables (at their means), but that is precisely the 7 We incorporate all of the controls from our earlier models because we intend to use the models in Table 3 in an instrumental variable approach later in the paper. 16

Table 3: OLS Regression Predicting Perceived Ideological Distance Between a Respondent and her Representatives in the 2008 CCES U.S. House Member Senior Senator Junior Senator Roll Call Divergence 10.789* 27.625* 22.819* (4.164) (4.431) (4.230) Respondent is Democrat -13.589 18.474* 26.511* (7.242) (6.920) (6.856) Respondent is Republican 2.465-4.873 6.672 (6.935) (6.772) (6.771) Representative is Democrat -4.889 15.102 21.386* (7.877) (7.889) (7.825) Ideology 3.198* -1.229 2.846* (1.191) (1.225) (1.153) Education 0.939 0.964 0.223 (0.652) (0.635) (0.614) Interest in Politics 4.184* 4.261* 6.080* (1.920) (1.874) (1.865) Perceptions of the National Economy -3.793* -3.113* 0.098 (1.636) (1.554) (1.500) Respondent is Black -9.377* -14.086* -5.924 (4.241) (4.181) (4.101) Respondent is Latino 5.504-0.148-2.925 (4.730) (4.173) (4.431) Respondent is Female -0.915-0.740 4.891* (2.012) (1.977) (1.908) Respondent s Age -0.002-0.074-0.018 (0.064) (0.625) (0.061) Respondent s Education 0.939 0.963 0.223 (0.652) (0.635) (0.614) Respondent s Religiosity 0.660 0.433-0.305 (0.636) (0.628) (0.586) Democrat Respondent X Democrat Representative 1.919 15.102* -44.235* (8.628) (7.889) (8.401) Republican Respondent X Democrat Representative 12.385 22.391* 6.712 8.473 (8.464) (8.442) Intercept 40.251* 35.159* 3.452 (12.335) (11.578) (11.779) N 511 500 527 R 2 0.399 0.455 0.493 Adjusted R 2 0.381 0.438 0.478 F -value 21.95 * 27.01* 33.20* Note: Cell entries are estimated coefficients generated by an OLS regression model. Standard errors are reported in parentheses. * p-value 0.05 (two-tailed test) 17

Predicted Perceived Ideological Differences between Respondent and House Member as Roll Call Differences Increase Perceived Ideological Differences 5 10 15 20 25 30 35 0.0 0.2 0.4 0.6 0.8 1.0 Highly Similar Voting Highly Different Voting Roll Call Differences Figure 2: Predicted Perceived Ideological Differences Between Respondent and her House Member as Roll Call Differences Increase for Democratic and Republican Respondents to the 2008 CCES. point. A number of additional variables play a role in the development of citizens perceptions of their legislators, implying that their perceptions of the ideological distance between themselves and their representatives are based on more than actual policy incongruence. Having demonstrated that actual behavioral divergence measured as roll call divergence between a representative and her constituents influences those constituents perceptions of the ideological distance between them, we next address the question of whether perceived ideological distance or roll call divergence influence levels of trust in the U.S. Congress. 18

To do so, we present three ordinal logistic regressions predicting citizens reported level of trust in the U.S. Congress in Table 4. These models are identical to the model in Table 2, except that we now include our measure of roll call divergence between a respondent and her members of Congress along with a respondent s perceived distance. 8 These results suggest that the perceived ideological divergence between a respondent and her legislator remains a negative and statistically significant predictor or trust levels even when controlling for roll call divergence. In each of our models, even after including actual behavioral divergence between a respondent and her legislator, the respondent s perceptions of that difference are what carry predictive power. In two of the three models, roll call divergence is not a significant predictor trust in Congress despite being a significant predictor of perceived distance. 9 This supports our expectation that actual congruence influences perceptions of congruence which subsequently influence trust in Congress. 10 To again provide some context to the size of these coefficients, Figure 3 presents the predicted probability of a respondent falling into the lowest category of trust in Congress as perceived ideological distance and roll call divergence increase using the coefficients from the House Member model in Table 4. All the other independent variables in the model are held at the same levels as Figure 1 and the opposing measure of ideological distance is held at its mean. Even after controlling for actual differences in the Roll Call preferences of respondents 8 There is good reason to be concerned that in our models of trust in Congress, perceived distance is actually caused by trust in Congress, rather than the reverse. Using the residual values from the regression in Table 3, we have performed the Wu-Hausman test for endogeneity between perceived divergence and trust. The test reveals that endogeneity is not a concern for our models of trust in Congress that include a respondent s House member and junior Senator, but there are reasons to be concerned about endogeneity for our model including a respondent s senior Senator. Given the high levels of predictive success in Table 3 (an R 2 of 0.41), we are confident that roll call divergence works as a quality instrument for perceived divergence, and thus, are confident that the Wu-Hausman test is a useful way to evaluate endogeneity. We discuss these results more in the supplemental appendix. 9 The coefficients on our roll call divergence measure are generally larger than those on the perceived congruence measure due to the difference in the scales of the variables. Perceived distance runs from 0 to 100, while roll call distance runs from 0 to 1. 10 We have tested the parallel slopes assumption of ordinal logistic regression in each of our ordinal models. These tests consistently indicate that treating trust in Congress as ordinal is preferable to treating trust in Congress as multinomial. 19

Table 4: Ordinal Logistic Regression Predicting Respondent s Trust in the U.S. Congress in the 2008 CCES House Member Senior Senator Junior Senator Perceived Distance -0.016 * -0.008* -0.012* (0.004) (0.004) (0.004) Roll Call Divergence 0.009-1.567* 0.066 (0.415) (0.456) (0.454) Respondent is Democrat 1.898 * 2.199* 2.580* (0.873) (0.746) (0.853) Respondent is Republican 0.567 0.471 0.984 (0.865) (0.732) (0.8489) Representative is Democrat 0.943 1.017 0.926 (0.928) (0.835) (0.942) Ideology -0.329 * -0.231-0.971* (0.121) (0.122) (0.121) Education -0.033-0.052-0.021 (0.063) (0.062) (0.063) Interest in Politics -0.522* -0.485* -0.510* (0.190) (0.182) (0.191) Perceptions of the National Economy -0.036-0.188-0.125 (0.166) (0.154) (0.156) Respondent is Black 0.849 * 0.736 0.981* (0.424) (0.416) (0.428) Respondent is Latino 0.598-0.198 0.308 (0.479) (0.415) (0.454) Respondent is Female 0.311 0.220 0.258 (0.195) (0.192) (0.195) Respondent s Age 0.001-0.003-0.001 (0.006) (0.006) (0.006) Respondent s Education -0.033-0.052-0.022 (0.063) (0.062) (0.063) Respondent s Religiosity 0.061 0.084 0.057 (0.062) (0.062) (0.061) Democrat Respondent X Democrat Representative -1.167-1.641-1.354 (0.984) (0.891) (1.001) Republican Respondent X Democrat Representative -0.348-0.611-0.739 (0.989) (0.891) (1.003) Cutpoint 1 0.379-0.875 0.316 (1.352) (1.183) (1.302) Cutpoint 2 3.112 * 1.785 3.166 (1.356) (1.181) (1.306) Cutpoint 3 5.562 * 4.651 6.145 (1.438) (1.274) (1.421) N 503 493 519 AIC 881.50 882.57 879.56 Pseudo R 2 0.554 0.553 0.555 Note: Cell entries are estimated coefficients generated by an Ordinal Logistic regression model. Standard errors are reported in parentheses. * p-value 0.05 (two-tailed test) 20

and their representatives, increases in respondents perceptions of the ideological distance between themselves and their representatives increase the probability of respondents Hardly Ever trusting Congress. A move from the minimum level of perceived ideological divergence to the maximum perceived ideological distance increases the probability of being in the lowest category of trust in Congress from roughly 0.29 to 0.74. Alternatively, increases in roll call divergence barely move the predicted probability of falling into this lowest category. A move from the minimum to the maximum roll call divergence increases the probability of being in the lowest category of trust in Congress from 0.41 to 0.40. 11 Thus, our empirical evidence provides support for each of our hypotheses. First, the perceived ideological distance between a respondent and her legislators influences her trust in Congress. Second, actual behavioral divergence between a respondent and her legislators influences her perceptions of that distance. Finally, roll call divergence has little influence on trust in Congress in two of our three models after controlling for a respondent s perceptions of the ideological distance between herself and her legislators, indicating that the influence of roll call divergence on trust in Congress is expressed through a respondent s perceptions of ideological distance. This provides strong support for our theory regarding the influence of the dyadic representational relationship between a respondent and her legislator on her trust in Congress. Individual legislators can improve trust in Congress by acting in greater congruence with their constituents preferences, but only so long as their constituents actually perceive of that congruence. 11 We have also run these models interacting these incongruence measures with both interest in politics and education. These additional model complexities do not effect the results of our models in important ways. Perceptions of ideological distance matter uniformly across educational and interest levels. 21

Probability of Hardly Ever' Trusting Congress as Perception of Ideological Differences Increase Holding Roll Call Differences Constant Probability of Hardly Ever' Trusting Congress as Roll Call Differences Increase Holding Perception of Ideological Differences Constant Probability of Hardly Ever' Trusting Congress 0.0 0.2 0.4 0.6 0.8 1.0 Probability of Hardly Ever' Trusting Congress 0.0 0.2 0.4 0.6 0.8 1.0 0 20 40 60 80 100 0.0 0.2 0.4 0.6 0.8 1.0 Perceived Ideological Divergence Roll Call Divergence Figure 3: The Predicted Probability of Being in the Lowest Level of Trust in the U.S. Congress. As average perceived ideological distance to respondent s set of legislators increases the probability of a respondent being in the lowest category of trust in Congress increase. The same probably is virtually unchanged by Roll Call voting differences. 22

4 Conclusion In this research, we argue that trust in Congress is formed in response to the principalagent relationship between citizens and their legislators. As legislators are perceived of as being more faithful agents before Congress, their principals trust the entire institution of Congress more. When those agents are perceived as poor representatives, principals trust the entire institution less. We also hypothesized that citizen s perceptions of the ideological distance between themselves and their legislators was grounded in the actual behavior of those legislators, but imperfectly so, and that the actual behavior of legislators only influences trust in Congress through citizen perceptions. Thus, legislators may increase the trust citizens have in Congress by advertising their representative activities. Legislative behavior can influence trust in Congress, but only if citizens are aware of how congruent that behavior is to citizens preferences. Empirical evidence from the 2008 CCES strongly corroborates our expectations. As a respondent s perception of ideological distance between herself and her legislators increases, so too does her trust in the U.S. Congress. This relationship holds true whether our analysis employs any of a respondent s legislators, and controlling for a host of alternative covariates. Additionally, when citizens perceptions of ideological distance are pitted against the actual behavioral divergence between citizens and their legislators, only perceptions are consistently predictive of trust in Congress. Thus, our theoretical account appears to be on firm empirical ground. A citizen s perceptions of the ideological distance between herself and her member of Congress are driven in part by the behavior of the member. A citizen s level of trust in Congress is in turn strongly affected by her perception of the ideological distance between herself and her legislators. Prior research on trust in government has focused largely on the macro-level and centered on trust in the federal government as a whole. Our research suggests that trust in government 23

institutions respond to predictable stimuli. Thus, this indicates that trust in Congress, the President, the Supreme Court, and the federal government more generally are all connected in fundamental ways, but still harbor distinct and important variance. Our results also suggest that trust in these governmental institutions may increase if citizens believe their representatives are more ideologically proximal to them, even if those representatives are not actually more similar to them. Further, our research indicates that perceptions dominate actual behavior in the formation of attitudes and evaluations of political institutions and actors. Thus, scholars of public opinion need to carefully consider citizens perceptions and how they may differ from the behavior of political actors when studying evaluations of government and political figures. Moving forward, a number of potential extensions and further tests are worth considering. For example, we rely on survey respondents perception of ideological distance to their legislators. Experimental work revealing the actual behavior of legislators may help further isolate the causal effects of perceptions of ideological proximity, legislative behavior, and updates to perceptions of ideological behavior on trust attitudes for Congress rather than simply relying on respondents beliefs about that proximity. Additionally macro-level public opinion work (e.g. Stimson, 2004) seems to imply that there are few systematic differences between trust in the branches of government and approval of the branches of government. It would be of interest to scholars of public opinion and representation to determine whether or not this remains true at the micro-level. Additionally, work on representation and district makeup has also consistently shown that homogeneous legislative districts are easier for legislators to represent and that the preferences of citizens from homogeneous districts are better reflected in the behavior of their legislators (Gerber and Lewis, 2004; Buchler, 2005). Our research in turn shows that the quality of policy congruence between legislators and their constituents has important implications for trust in Congress. Combined these two pieces of research suggest that constructing homogeneous, uncompetitive Congressional 24

Districts is one way to increase citizens trust in Congress. Citizens in homogeneous districts would be more likely to be well-represented by their legislators, and by extension, would be more likely to trust Congress. Finally, if legislators are aware of how ideological proximity affects citizens evaluations of their chamber, it would be of interest to scholars of legislative politics to observe how changes in trust and approval of Congress are answered by legislative activity. 25

References Aberbach, Joel D. 1969. Alienation and Political Behavior. The American Political Science Review 63(March): 86 99. Abney, F. Glenn, and John D. Hutcheson Jr. 1981. Race, Representation, and Trust: Changes in Attitudes After the Election of a Black Mayor. Public Opinion Quarterly 45(1): 91 101. Bandura, Albert. 1982. Self-Efficacy Mechanism in Human Agency. American Psychologist 37(February): 122 147. Bobo, Lawrence, and Franklin Gilliam. 1990. Race, Sociopolitical Participation, and Black Empowerment. American Political Science Review 84(June): 377 393. Buchler, Justin. 2005. Competition, Representation and Redistricting The Case against Competitive Congressional Districts. Journal of Theoretical Politics 17(4): 431 463. Campbell, Angus, Philip E. Converse, Warren E. Miller, and Donald E. Stokes. 1960. The American Voter. New York: Wiley. Canes-Wrone, Brandice, David W. Brady, and John F. Cogan. 2002. Out of Step, Out of Office: Electoral Accountability and House Members Voting. The American Political Science Review 96: 127 140. Chanley, Virginia A., Thomas J. Rudolph, and Wendy M. Rahn. 2000. The Origins and Consequences of Public Trust in Government: A Time Series Analysis. Public Opinion Quarterly 64(3): 239 256. Citrin, Jack. 1974. Comment: The Political Relevance of Trust in Government. American Political Science Review 68(September): 973 988. Citrin, Jack, and Donald P. Green. 1986. Presidential Leadership and the Resurgence of Trust in Government. British Journal of Political Science 16(October): 431 453. Conover, Pamela Johnston. 1981. Political Cues and the Perception of Political Candidates. American Politics Quarterly 9: 427 448. Converse, Philip E. 1964. The Nature of Belief Systems in Mass Publics. In Ideology and Discontent, ed. David E. Apter. Ann Arbor: University of Michigan Press. Craig, Stephen C. 1993. The Malevolent Leaders: Popular Discontent in America. Boulder, CO: Westview. Delli Carpini, Michael X., and Scott Keeter. 1996. What Americans Know About Politics and Why it Matters. New Haven: Yale University Press. Downs, Anthony. 1957. An Economic Theory of Democracy. New York: Harper and Row. 26