Class and Unequal Representation in Latin America: Linking Descriptive and Symbolic Representation

Transcription

1 Class and Unequal Representation in Latin America: Linking Descriptive and Symbolic Representation Tiffany D. Barnes Associate Professor University of Kentucky Gregory W. Saxton Ph.D. Candidate University of Kentucky *Paper Prepared for the Annual Meeting of the American Political Science Association, September 2, 2017, in San Francisco, California.

2 Abstract How does the near-exclusion of working class citizens from government influence citizens perceptions of representation? We argue that when groups of people are continually denied access to representation, citizens are unlikely to believe that their interests are represented by political parties or by parliament. By contrast, more inclusive institutions that incorporate members of the working class foster support for representative bodies, particularly among working class citizens and among citizens with the resources and interests to monitor government. Using a multilevel analysis of 18 Latin American countries a region plagued by a crisis of representation we show greater inclusion of the working class is associated with stronger attachments to agents of representation and better evaluations of legislative performance. These findings have implications for strengthening democracy in Latin America, as they indicate that more diverse governments may be key to deepening citizens attachments to representative bodies and mitigating the crisis of representation.

3 Democracy rests on the idea that power should reflect the will of the people. In practice, however, democracies vary dramatically in their representativeness, with different groups in society facing varying levels of inclusion (Morgan and Meléndez 2016; Taylor-Robinson 2010). In particular, working class people are dramatically underrepresented in democratic governments across the globe, as politicians are drawn from a narrow set of elites (Best 2007; Carnes 2013). In Latin America, for instance, despite that more than sixty percent of employees are blue-collar workers, fewer than twelve percent of politicians are drawn from the working class (Carnes and Lupu 2015a). This political exclusion of the working class calls into question one of the fundamental principles of democracy. How does the exclusion of the working class influence citizens perceptions of and satisfaction with representation? We provide the first study of workers symbolic representation. Despite that previous research has investigated the policy consequences of class representation (e.g., Carnes 2012; 2013; Carnes and Lupu 2015a) and evaluated whether citizens are willing to support working class candidates (Carnes and Sadin 2015; Carnes and Lupu 2016), no study has considered whether descriptive representation of the working class improves satisfaction with representation. We argue that when groups of people are continually denied access to representation, citizens are unlikely to believe that their interests are represented by the political system. In particular, given that political parties and legislatures are the primary conduits of representation (Mainwaring et al., 2006; Morgan 2011; Taylor-Robinson 2010), for people to feel represented they need to identify with these key agents of representation and believe they represent their best interests in the policy-making process. When elected officials are drawn from a narrow set of elites, this undermines the core principals of a representative democracy, causing all citizens and particularly those from the working class to question the extent to which elected officials can identify with their concerns or will have their interests in mind. Thus, we anticipate that the exclusion of working class members weakens citizens Class and Unequal Representation 1

4 ties to political parties and undermines trust in and satisfaction with the legislature. By contrast, greater inclusion of the working class transmits pro-egalitarian messages (e.g., Morgan and Buice 2013) that serve to strengthen citizens satisfaction with parties and legislatures. Nonetheless, for descriptive representation of the working class to incite symbolic representation, citizens must be aware of legislators working class backgrounds and be able to observe legislators influence on the policy-making process. Unlike other descriptive characteristics such as gender, race, or ethnicity that incite symbolic representation (Alexander 2012, 2015; Hayes and Hibbing forthcoming), it may be difficult to identify legislators class backgrounds. As a result, a number of resources (e.g., access to information and education) and a high level of political interest are necessary for citizens to determine the class of representatives and effectively monitor politics, yet citizens access to these resources and interest in politics varies dramatically (Rubenstein 2007; Taylor-Robinson 2010). Whereas some citizens can easily acquire and process political information, others lack the capacity and/or motivation to monitor government activity. To the extent that descriptive representation of the working class engenders symbolic representation, we thus expect this relationship to be stronger among citizens with the most resources to monitor politicians and those who express strong interest in politics. To evaluate support for our argument, we turn to a multilevel analysis of elite and public opinion surveys from 18 Latin American countries, which exhibit substantial variation in the level of working class representation in national legislatures. We use data from the University of Salamanca (USAL) to identify legislators with working class backgrounds, and individual-level data from the Latin American Public Opinion Project s (LAPOP) AmericasBarometer to evaluate citizens beliefs about representation (Carnes and Lupu 2015a). Using measures that capture both affective attachments to parties and evaluative questions of policy-making performance, we find that as the share of working class legislators increases, citizens perceptions of representative institutions Class and Unequal Representation 2

5 improve. In particular, citizens become more likely to identify with and trust political parties, on average, and they are more likely to say that political parties listen to them and represent their interests. Additionally, all citizens are more trusting of the legislature, and they are more likely to approve of the legislature s job performance when the working class is better represented. Although this relationship extends to all citizens including those from the working class it is strongest among those individuals with the greatest interest and capacity to monitor government, thus bolstering confidence in our results. That is, if the observed relationship between class descriptive representation and symbolic representation were spurious we would not expect it to hold among a subset of citizens from the working class, nor would we expect it to be stronger among citizens with the highest level of political awareness. Latin America offers an ideal setting to evaluate our hypotheses. To begin with, Latin America hosts substantial variation in the level of working class representation in national legislatures. Workers access to legislative office ranges from a low of three percent in Brazil to upwards of 25 percent in Honduras (Carnes and Lupu 2015a). This variation is necessary to evaluate our theorized link between class representation and citizens beliefs about representation or symbolic representation. Equally important, members of the working class make up the vast majority of the labor force in Latin America between 60 and 85 percent. 1 In many Latin American countries, poor and working class citizens have suffered a history of social and political exclusion (Taylor-Robinson 2010), and in most countries, they remain dramatically underrepresented today (Carnes and Lupu 2015a). Our findings suggest that this underrepresentation has contributed to a crisis of representation in Latin America during the last two decades, wherein citizens display widespread disenchantment with and rejection of parties and legislatures (Mainwaring 2006, 16; see also, Luna 2016; Mudde 2015; Siavelis 2016). 1 Source: International Labor Organization (ILO), see Carnes and Lupu (2015a). Class and Unequal Representation 3

6 Dissatisfaction with central agents of representation has implications for democratic consolidation and the quality of democracy more generally (Luna and Zechmeister 2005; Mainwaring et al. 2006). When representative linkages between citizens and the state break down, electoral participation declines, electoral volatility increases, citizens turn to anti-establishment political outsiders, and the entire party system risks collapse (Mainwaring et al. 2006; Mayorga 2006; Morgan 2011; Morgan and Meléndez 2016; Tanaka 2006). Our findings demonstrate that political parties can address this crisis and enhance citizens trust in political institutions and feelings of representation more generally by incorporating members of historically marginalized groups. A common criticism of descriptive representation is that incorporating members of marginalized groups, such as the working class, into representative bodies could pose a danger to democracy by putting unqualified representatives into power (Mansbridge 1999; 2015). The findings from this study contribute to a growing body of research that empirically evaluates this criticism. Whereas others find that descriptive representatives are at least as effective (Carnes and Lupu 2015b; Volden et al. 2015) and qualified (Nugent and Krook 2015; Murray 2015) as their colleagues, the findings from our research further rebut this criticism demonstrating that descriptive representation of the working class actually strengthens democracy by enhancing symbolic representation for working class and non-working class citizens alike. Indeed our findings show that when more legislators come from working class backgrounds, citizens attachments to the key agents of democratic representation political parties and legislatures are substantially enhanced. As established political parties struggle to retain their relevance, and party system linkages decay in the volatile electoral climates across Latin America (Leongómez 2006; Lupu 2014; Morgan 2011; Roberts 2012), our work suggests that parties seeking to strengthen ties with constituents should recruit more members of the working class. Not only do workers exert a distinct effect on the Class and Unequal Representation 4

7 policymaking process (Carnes 2013; Carnes and Lupu 2015a), but they also wield a profound influence on symbolic representation, thus bolstering the health of democratic institutions. Why Perceptions of Representation Matters For democratic institutions to consolidate and function properly, citizens need to view democracy as the only game in town (Linz and Stepan 1996). Yet, when citizens do not feel like they are being represented, this lack of symbolic representation (Pitkin 1967) poses problems for democratic consolidation. Crises of representation have led to democratic backsliding in Venezuela and Peru (Mainwaring 2006), for example, and are associated with complete party system collapses in other countries in the region (Morgan 2011). As a matter of fact, in some of the countries suffering from a crisis in representation, confidence in representative institutions has declined precipitously (Mainwaring 2006). Between 2001 and 2008 in Peru, for instance, the percentage of people expressing confidence in political parties decreased from about 24 percent to around 11 percent in Likewise, for most of the early 2000s, Ecuadorians trust in the legislature hovered between 7 and 12 percent. 2 Yet, similar trends are also observed in countries with more stable democracies such as Chile (Morgan and Meléndez 2016). These low levels of satisfaction have serious implications for democracy particularly in Latin America where the crises of representation often precipitated democratic setbacks (Mainwaring et al. 2006) and even in more entrenched democracies where dissatisfaction with representation has given rise to populist antiestablishment leaders (Mudde 2015). A growing body of literature shows that political institutions play an important role in shaping citizens feelings and attitudes about democracy and representation. At the broadest level, 2 Source: Latinobarometro. AmericasBarometer data, LAPOP did not start surveying most Latin American countries until Class and Unequal Representation 5

8 more proportional electoral systems are shown to increase satisfaction with democracy (Anderson and Guillory 1997). When citizens see representatives who look like them, they are more likely to participate in and positively evaluate the political system (Banducci, Donovan, and Karp 2004; Barnes and Burchard 2013). For this reason greater numeric representation of women (Karp and Banducci 2008; Ulbig 2007) and racial and ethnic minorities (Hayes and Hibbing forthcoming; Scherer and Curry 2010; Tate 2001) are generally shown to foster higher levels of trust and satisfaction with the government. Although a substantial body of research has investigated the symbolic consequences of numeric representation of women and racial and ethnic minorities, scholars have not considered how numeric representation of the working class shapes citizens attachments to and evaluations of government. Herein we argue that for multiple reasons, descriptive representation of the working class may also have profound implications for symbolic representation. When white-collar representatives are left to act on behalf of the working class, this sends a signal to working class citizens that their participation in politics is not valued, and it may also lead to policy failure and even erode the legitimacy of representative institutions. Yet, the mere presence of working class representatives in decision-making bodies may have a direct effect on symbolic representation improving citizens feelings about representation (Schwindt-Bayer and Mishler 2005) and enhancing the de facto legitimacy of representative institutions for everyone (Mansbridge 1999). Moreover, where descriptive representation leads to improved policy representation, it may have an indirect effect on citizens evaluations of the government. Thus, in the following section we theorize that representation of the working class is imperative for restoring trust and satisfaction with political parties and legislatures. Parties, Legislatures, and Crises of Representation in Latin America Class and Unequal Representation 6

9 Political parties and legislatures are the primary vehicles for achieving democratic representation. Indeed, scholars have gone so far as to say that modern democracy is unthinkable without political parties (Schattschneider 1942). Political parties are central to democratic representation because they supply voters with information shortcuts, provide mechanisms for electoral accountability, and connect citizens to the state (Mainwaring et al. 2006; Mershon and Shvetsova 2013; Morgan 2011). Further, parties articulate and aggregate citizens interests, creating majorities and organizing dissent in the legislature. They implement policy objectives, and serve as symbols of identification and loyalty that allow political dissatisfaction with governmental outputs to be directed at specific institutions rather than at the state itself (Dalton and Wattenberg 2000, 6-10). Legislatures are a second, more institutional, vehicle for democratic representation (Crisp 2006; Mainwaring 2006). Legislatures are intended to be the represenative branch of government (Taylor-Robinson 2010), and interparty politics within the legislature is key to the success of programmatic policy goals (Crisp 2006). Indeed, it is in the legislature that descriptive representation generally yields substantive outcomes (Mansbridge 1999; 2015; Pitkin 1967). At its root, the crisis of representation in Latin America, as well as in other democracies around the world, is a story about disenchantment with political parties and legislatures (Luna 2016; Mainwaring 2006; Mair 1995; Tanaka 2006). Indeed, despite the importance of these institutions in Latin America, they are generally viewed with distrust (Mainwaring et al. 2006; Seligson 2007). When these agents of representation become discredited in the eyes of citizens, populist presidents may displace these institutions as representatives of the people (Taylor-Robinson 2010, 14). Scholars attribute the crisis of representation to a variety of factors. Crisp (2006), for instance, attributes the crisis to institutional factors, such as electoral systems that encourage personal vote seeking, and thus a focus on particularistic rather than programmatic benefits. Other scholars attribute the low Class and Unequal Representation 7

10 levels of confidence in parties and legislatures to weak governance and poor state performance (Mainwaring 2006; Mayorga 2006). We posit that the exclusion of the working class a group that makes up a sizeable majority of the population in Latin America contributes to the crisis of representation, even after taking state performance into consideration. Whereas members of the working class make up the vast majority of the labor force in most Latin American countries, they only make up about 12 percent of the legislatures in the region. We speculate that the crisis of representation is mitigated where this disparity is less stark. Despite competing claims about the root cause of the crisis, understanding the factors that enhance citizens attachments to and affect for political parties and legislatures is vital for representative democracy. Symbolic Representation from Descriptive Representation Normative democratic theory posits that chronic underrepresentation of historically marginalized groups is intrinsically unfair (Williams 1998), and any efforts to remedy this chronic underrepresentation should be seen as a democratic good. In Latin America, class is one of the most salient and defining cleavages in society. Not only do members of the working class have fundamentally different life chances, but they also remain on the margins in politics. Although decades of comparative research have discounted any relationship between politicians socioeconomic background and political attitudes and preferences, (Matthews 1985; Norris and Lovenduski 1995; Putnam 1976), more recent research demonstrates that class-based descriptive representation is vitally important, and that the conceptualization of class is fundamentally distinct from socioeconomic status as defined by income (Carnes and Lupu 2015a). In sharp contrast to income or education, a robust body of research indicates that how individuals earn a living (i.e., their occupation) most accurately reflects people s place in society (Brooks and Manza 2007). Individuals in different professional strata, even those with the same income, face dramatically different life Class and Unequal Representation 8

11 chances (Carnes and Lupu 2015a). As Carnes and Lupu (2015a, 4) explain: An intern at an investment bank, a skilled machinist, and a graduate student, for instance, earn similar incomes in the United States, but it would make little sense to group them in the same class. Occupational background reflects class status better than income because the dividing line between social classes in most societies revolves around the labor market, that is, how people earn a living. (Carnes 2013, 3) In Latin America, the dramatic underrepresentation of the working class has created a political playing field that many observers perceive as unequal. The legacy of elite, upper class political dominance signals that politics does not work for lower class citizens, thereby promoting a political apathy and fatalism that is hard to overcome (Taylor-Robinson 2010, 12). By contrast, political inclusion of marginalized groups sends a signal that politics represents all citizens (Kittilson and Schwindt-Bayer 2012), and it enhances the de facto legitimacy of the entire political system (Mansbridge 1999). Drawing on these insights, we posit that incorporation of the working class into formal bodies of representation will increase the perceived legitimacy of formal agents of representation. Just as citizens draw on working-class backgrounds as a useful heuristic or information shortcut when evaluating candidates and voting (Carnes and Sadin 2014), we posit the same heuristic devices structure citizens attachments with representatives. As such, increases in descriptive representation of the working class should erode some of the structural inequalities and social hierarchies that have plagued Latin American politics, and may dampen or even reverse feelings of political apathy and disillusionment. Symbolic Representation through Policy Responsiveness Beyond the direct effects of descriptive representation, legislators from the working class may also engender feelings of representation through policy responsiveness. Members of marginalized groups are more likely to represent the interests of those groups (Escobar-Lemon and Class and Unequal Representation 9

12 Taylor-Robinson 2014; Htun 2016; Schwindt-Bayer 2010; Taylor-Robinson and Heath 2003; Williams 1998) and this policy or substantive representation may strengthen citizens feelings of being effectively represented. Indeed, one of the core reasons descriptive representation is argued to be important is because of its policy consequences (Mansbridge 1999; Weldon 2002). Policy representation is fundamental for political legitimacy, and, as Mansbridge (2015, 263) explains, representatives who have themselves lived through typical working class experiences will be more likely than others to understand the nuances of issues and thus be able to represent their working class constituents more insightfully than even committed and well-meaning representatives who have not had similar experiences. Thus, not surprisingly, despite that anyone can theoretically provide substantive representation, research shows that descriptive representation of blue-collar workers is associated with higher levels of substantive representation for working class citizens. A person s class status reflects her life chances and socialization, and in Latin America, working class status is an important predictor of a whole host of political attitudes and policy preferences. Across Latin America, for instance, working class legislators are shown to have different policy priorities and preferences than their colleagues from white-collar backgrounds. Working class politicians are far more likely to favor state intervention and increased government spending for social and economic welfare programs (Carnes and Lupu 2015a). In Argentina, for example, lawmakers from the private sector have more conservative views about government spending and economic intervention than working class politicians (Carnes and Lupu 2015a). Similar patterns exist in the United States, where legislators from the working class are more likely to devote their legislative efforts to advancing legislation related to labor and employment and other economic policies as opposed to issues such as banking and finance (Carnes 2012; 2013). Moreover, when casting their votes, working class Class and Unequal Representation 10

13 legislators are more likely to prioritize constituents from the lower class and local businesses rather than big businesses, government administrators, and white-collar professionals (Carnes 2013, 101). This increased attention to policy issues that disproportionately influence the lives of working class citizens may further engender citizens satisfaction with key agents of representation. Given that descriptive representatives are more likely to provide such policy representation, descriptive representation may further foster symbolic representation through increased attention to policy issues that disproportionately influence the working class, a group that constitutes the majority of the labor force in Latin America. In sum, insights from prior research on democratic representation suggests that descriptive representation of the working class may engender greater feelings of being fairly and effectively represented, both directly (regardless of policy representation) and indirectly (via policy representation). Hypothesis 1: Increases in the percentage of legislators from the working class will improve perceptions of representation for working-class citizens. Descriptive Representation Improves Perceptions of Representation for All Citizens Although it may seem obvious that descriptive representation of the historically excluded serves to improve perceptions of representation among members of that group, theoretical and empirical insights from previous research on representation suggests these benefits should extend to all groups (Mansbridge 1999; Morgan and Buice 2013). In the abstract, scholars have suggested that increases in descriptive representation may enhance de facto legitimacy, not only for historically marginalized groups, but for the powerful ones as well (Mansbridge 1999). Research on elite cues offers further theoretical insights into how and why descriptive representation of marginalized groups may enhance perceptions of representation for all citizens. Elite behavior such as nominating more working class candidates to run for office sends important pro-egalitarian signals, promoting support for class representation (Beaman et al. 2009; Class and Unequal Representation 11

14 Hansen 1997). And, although elite behavior is likely to shape the attitudes of all citizens, it can also influence citizens with weak (or no) personal investment in the issues (e.g., Barnes and Córdova 2016; Morgan, Espinal, and Hartlyn 2008). Morgan and Buice (2013) argue, for example, that because men are less likely than women to hold firm beliefs on the role of women in leadership, they are more susceptible to cues transmitted via elite behavior. In this vein, they demonstrate that men are more supportive of female leadership in countries where elites nominated women to serve as cabinet ministers. Given that white-collar workers are unlikely to have a strong self-interest in the representation of the working class, they may rely heavily on elite-cues from party leaders and congress when assessing the quality of representation. Thus, where party leaders have incorporated members of the working class, this pro-egalitarian message is likely to extend beyond workers shaping the mass-public opinion of all citizens. Hypothesis 2: Increases in the percentage of legislators from the working class will improve perceptions of representation for all citizens. The Conditional Impact of Monitoring Capacity For descriptive representation of the working class to incite symbolic representation, citizens must be able to monitor both who is in office (i.e., if they have a working class background) and how representatives engage in the policy-making process. Yet, as we explain, monitoring representatives is costly. Whereas some citizens have access to monitoring resources, such as education and information, or have high levels of political interest allowing them to effectively monitor politics many citizens lack monitoring resources and express very little interest in politics. Without monitoring resources or political interests citizens cannot easily assess if representatives are standing for them and representing their policy preferences. As a result, we propose that the relationship between descriptive representation and symbolic representation will be strongest among Class and Unequal Representation 12

15 citizens with high levels of monitoring resources and among those expressing high levels of political interests. To begin with, the link between descriptive and symbolic representation of the working class will only be realized if citizens are familiar with the individuals elected to office and are privy to their backgrounds. Members of the working class do not possess descriptive attributes that are readily visible unlike sex, race, or ethnicity thus, it is relatively demanding for citizens to observe the class back ground of representatives. Furthermore, observing substantive representation may prove even more demanding. Programmatic policies, for example, are challenging to monitor, as they do not result in immediate tangible payoffs. Rather, it can take years for ordinary citizens to realize the benefits of programmatic policies, and policy success is often conditional on whether the bureaucracy effectively implements the policies. Even when legislators are working to represent their constituents, they frequently lack the resources to successfully pass policy proposals. Unsuccessful proposals are even more difficult to monitor than successful bills. Given these challenges, citizens must collect information and use it to evaluate both who is in office and how they behave if descriptive representation of the working class is to foster feelings of fair and effective representation. Monitoring Resources Yet, monitoring politics and particularly the representation of the working class is cognitively demanding and costly. It requires extensive resources such as education, time, money, and information and not all citizens have equal access to the monitoring resources necessary to evaluate politics (Taylor-Robinson 2010; Rubenstein 2007). In particular, citizens with fewer resources especially those with little or no education and limited access to information typically lack the resources necessary to monitor elected officials (Gallego 2015; Taylor-Robinson 2010). First and foremost, education is fundamental for monitoring the political process (Rubenstein 2007). Class and Unequal Representation 13

16 Whereas educated people can easily acquire and process sophisticated information and distill complex data necessary to evaluate politics (Rosenstone and Hansen 1993; Wolfinger and Rosenstone 1980), uneducated people have limited access to reliable information and are inhibited in their ability to process complicated information. Limited reading skills make it difficult, for example, to read basic newspaper articles, let alone to navigate the intricacies of the political system (Smith 2016). Third party monitoring may serve as an information shortcut so that individuals regardless of their monitoring capacity can have at least some access to important information about political representatives (Rubenstein 2007). The third party can sort through information, identifying the most important and salient information, and give it to stake holders (voters) in a digestible format. News media, for example, serves as a surrogate source of information, allowing uneducated and poor people to overcome their information deficit. Nonetheless, although media is critical for government oversight, it is not a perfect surrogate for monitoring representation. Whereas news media is likely to provide citizens with information about who is in office and whether the legislature successfully passes legislation that represents the working class s best interests, it is less likely to provide accurate information about whether individual legislators championed workers interests in the absence of policy adoption. Furthermore, media bias can limit the quality of this information undermining poor peoples ability to evaluate the extent to which representatives act on their behalf (Taylor-Robinson 2010). Uneducated people may be less equipped than educated citizens to critically evaluate complex news stories (Smith 2016) or to discern media bias. For these reasons, although news consumption improves individuals monitoring capacity, alone, it may not be sufficient for poor and uneducated people to overcome their monitoring deficits. In sum, we contend that for descriptive representation to foster symbolic representation, citizens must be able to monitor who is in office and the policy-making process more generally to Class and Unequal Representation 14

17 evaluate whether politicians are at least trying to represent their interests. Yet, monitoring politics is demanding and citizens access to the resources necessary to monitor politics vary dramatically. To the extent that policy representation of the working class engenders symbolic representation, we expect this relationship to be stronger among citizens with the most resources to monitor politicians. We thus posit the following hypothesis: Hypothesis 3: The positive relationship between descriptive representation of the working class and symbolic representation will be stronger among resource-rich citizens than among resource-poor citizens. Political Interest Independent of citizens capacity to monitor politics, individuals with high levels of political interest are more likely to be aware of government activities. Much of the literature on political participation and engagement, for example, conceptualizes interest as a psychological attribute that is distinct from resources (e.g., Brady, Verba, and Schlozman 1995). In contrast to people with low levels of political interest, citizens expressing high interest are more likely to expose themselves to political information and seek out opportunities to engage in political activities (Delli Carpini and Keeter 1997; Verba, Schlozman, and Brady 1995). For example, interested citizens are more likely to engage in conversations and social interactions that facilitate the dissemination of political information therefore keeping them up to speed on politics. As Delli Carpini and Keeter (1997) explain, interest in politics is a key predictor of all types of political knowledge: People who say they follow politics are more likely to know more about any aspect of politics than those who claim not to pay attention (175). Political interest is relatively stable over time much like political identity and although it is easier than ever for uninterested citizens to avoid exposure to political information, interested individuals actively seek out political information, learn more about politics, and participate at higher rates (Prior 2010). Thus, independent of their level of monitoring resources, citizens with strong political interest are more likely to be aware of who is in office and Class and Unequal Representation 15

18 how well they are representing their constituents. As a result, we anticipate that working class representation is far more likely to incite feelings of symbolic representation among citizens with high levels of political interest than among those with lower levels of interest. Hypothesis 4: The positive relationship between descriptive representation of the working class and symbolic representation will be stronger among citizens with higher levels of political interest than among those with low levels of political interest. Data and Methods To evaluate the impact of descriptive representation of the working class on symbolic representation, we employ a number of individual-level survey questions from the 2008 wave of the Latin America Public Opinion Project s (LAPOP) AmericasBarometer in 18 counties. 3 Additionally, we leverage elite-level data on legislators occupational background as a measure of working class descriptive representation. Data on occupational background comes from the second wave of the University of Salamanca s (USAL) survey of Latin American legislators. 4 Dependent Variables The dependent variable in this analysis is symbolic representation, or perceptions of representation. To measure citizens feelings of being fairly and effectively represented (Schwindt- Bayer and Mishler 2005, 407), we employ two indices. First, the political party index captures affective attachments to parties and feelings about how well they fulfill their representative function. Second, the legislature index measures evaluations of policy-making performance. Importantly, our results are remarkably robust to the decision to index these questions into one variable. Models presented in 3 A full list of countries is included in the Appendix. 4 Specifically, the survey asks, What was your primary activity [that earned you the most money] prior to being elected Deputy? Guatemala s second wave survey uses a different question wording, so we rely on the first wave. Additionally, Brazil was not included in the USAL second wave, so we rely on the third wave. Class and Unequal Representation 16

19 the Appendix show the consistency of our findings for each component of the political party and legislature indices. 5 We begin by employing four questions that are designed to tap citizens feelings and attitudes toward political parties. The first question taps citizens confidence in political parties. Respondents are asked to indicate how much they trust political parties on a scale from 1 (not at all) to 7 (a lot). The next two questions measure citizens satisfaction with parties representative function. LAPOP asks for respondents evaluations of the following: 1) Thinking of political parties in general, to what extent do your country s political parties represent their voters well; and 2) How often do political parties listen to the average person? Responses are coded such that 1=not at all and 7=a lot. Finally, LAPOP asks a question about citizens attachment to political parties: do you currently identify with a political party, yes or no? 6 To create our measure of perceptions of political party representation, we recode responses to these four questions between 0 and 1, and then we average the responses. While political parties are essential for modern democracy (Schattschnieder 1942), the crisis of representation has manifested itself in other representative institutions as well. In particular, Mainwaring et al. (2006) explain that when representation breaks down, trust in the legislature the primary policymaking body also suffers. As with political parties, LAPOP asks a number of questions meant to measure citizens feelings and attitudes about the legislature. Respondents are asked to indicate their level of trust in the legislature on a scale from 1 (none) to 7 (a lot). LAPOP also asks respondents: Thinking of Congress as a whole, without considering the political parties to which they belong, do you think the members of Congress are performing their jobs very poorly 5 In Tables B.1.a and B.1.b in the Appendix, we separately model each component of political party index and our legislature index as the dependent variable. 6 Factor analysis reveals that all for of these variables map onto a single dimension, their Cronbach s alpha scale reliability coefficient is.53. Class and Unequal Representation 17

20 (coded 1), poorly (2), fair (3), well (4), or very well (5)? Finally, respondents were asked, To what extent does the Congress accomplish what you would hope for it to do? Responses range from not at all (1) to a lot (7). 7 As with the political party measure, we rescale responses to these three questions between 0 and 1. We then average responses to create our measure of perceptions of legislature representation. Independent Variable Descriptive Representation: The key explanatory variable in this analysis is descriptive representation of the working class. Specifically, we use data from the USAL to construct a measure that captures the percentage of legislators that come from a working class background. Following Carnes and Lupu (2015a), we operationalize working class background with a legislator s occupation prior to getting elected. 8 As Figure 1 shows, the percentage of workers in the legislature varies dramatically across Latin American countries, from a low of 3 percent in Brazil to a high of 26 percent in Honduras. Working Class Citizens: To test our first hypothesis, that descriptive representation of the working class will increase perceptions of representation among the working class, we create a dummy variable that measures whether or not individuals belong to the working class. Following Carnes and Lupu (2015a), we use occupational data from LAPOP and code individuals as working class if they indicate that their main occupation or type of work is as a skilled worker, office worker, employee in the service sector, food vendor, farmhand, domestic servant, or servant. All other occupations are coded as non-working class. Monitoring Resources: Although we expect that all citizens will express higher levels of symbolic representation when greater numbers from the working class are incorporated into representative 7 These three measures all load onto a single factor. Their scale reliability coefficient is Coding rules for which occupations are classified as working class are included in the appendix. Class and Unequal Representation 18

21 Figure 1: Distribution of Dependent Variable, Percent Working Class in the Legislature Note: Figure 1 shows the distribution of legislators with working class occupational backgrounds (Carnes and Lupu 2015a) across Latin American countries. institutions, we also posit that this relationship might be conditional on citizens level of monitoring resources, such as education and access to political information. Specifically, we construct a measure of monitoring resources by combining responses to LAPOP questions about respondents highest level of education completed, as well as how often respondents watch news on the television and read news in the newspaper. For the analysis, we rescale all these responses between 0 and 1 and then take their average to create our measure of monitoring resources. 9 Political Interest: We also expect the relationship between descriptive representation of the working class and symbolic representation to be conditional on political interest. LAPOP asks 9 Cronbach s alpha=.61 Class and Unequal Representation 19

22 respondents, How much interest do you have in politics: a lot (coded 1), some (2), little (3), or none (4). For the analysis, we recoded responses such that higher values indicate more political interest. Control Variables Individual Level: In addition to monitoring resources and political interest, we also control for a number of individual-level variables that might influence individuals attitudes toward political institutions. First, we control for political ideology. Ideology is measured using a question that asks respondents to place themselves on a 10-point scale, where 1 is left and 10 is right. We then collapse responses into three nominal categories: left, center, and right. Given our theoretical expectations about weak or non-existent party attachments in countries suffering from crises of representation, we recode non-response values as a category and include them in our analysis. We also control for perceptions of the economy. While some scholars attribute the crisis of representation to poor state performance (Mainwaring 2006; Mayorga 2006), we expect that working class representation has an effect on symbolic representation independent of economic performance, and thus we control for economic evaluations. LAPOP asks respondents whether they think economic conditions in their country are the better, the same, or worse than they were 12 months ago. We recode responses such that better/same=0 and worse=1. Finally, we control for respondents age, sex, and marital status. Country Level: In addition to our battery of individual-level controls, we also account for a number of country-level variables in our analysis. To control for the possibility that the crisis of representation is caused by weak state capacity (Mainwaring 2006), or that representation is correlated with the level of political and socioeconomic development in a country (Luna and Zechmeister 2005) we include an economic measure of Gross National Income (GNI) per capita. Our results, moreover, are robust to other country-level factors that may shape perceptions of representation. Specifically in the Appendix we demonstrate the consistency of our findings when Class and Unequal Representation 20

23 controlling for: level of democracy (Freedom House), government effectiveness (World Bank Governance Indicator), party system fragmentation (Effective Number of Parties in the Legislature), and the percentage of women in the lower chamber of the legislature. 10 Method A one-way ANOVA reveals that our two dependent variables measuring symbolic representation vary significantly across the 18 countries included in this analysis (p <.001), suggesting that factors measured at the country-level explain some of the variation in the individual-level dependent variables. To account for the nested nature of the survey individuals i living in countries j as well as individual-level and country-level data, we estimate a series of multilevel models to test our hypotheses about descriptive representation of the working class. As Steenbergen and Jones (2002) explain, The goal of multilevel analysis is to account for variance in a dependent variable that is measured at the lowest level of analysis by considering information at all levels of analysis (219). We estimate a series of multilevel linear regression models using STATA Results Influence of Working Class Legislators on Perceptions of Representation Models 1 and 2 in Table 1 present the results for our political party dependent variable, and models 3 and 4 present the results for our legislature dependent variable. Models 1 and 3 present baseline results, and Models 2 and 4 contain interactions between our variables for working class descriptive representation and citizens class identity, and offer a test of our first two hypotheses. Since the signs and significance on interaction coefficients and their constituent parts must be interpreted with caution, we follow the advice of Brambor, Clark, and Golder (2006) and present the results graphically. 10 See Tables B.4.a and B.4.b in the Appendix. Class and Unequal Representation 21

24 Table 1: Impact of Working Class Descriptive Representation on Symbolic Representation Model 1 Model 2 Model 3 Model 4 Party Party Legislature Legislature % Workers in Legislature.68***.66***.69**.69** (.19) (.19) (.29) (.29) Working Class Citizen.01*** *.01 (.00) (.01) (.00) (.01) % Workers x Working Class.10* -.02 (.05) (.05) Individual Level Monitoring Resources *** -.03*** (.01) (.01) (.01) (.01) Political Interest.07***.07***.03***.03*** (.00) (.00) (.00) (.00) Income *** -.00*** (.00) (.00) (.00) (.00) Female.01**.01**.02***.02*** (.00) (.00) (.00) (.00) Ideology=Left.04***.04*** (.00) (.00) (.00) (.00) Ideology=Center.01*.01**.01***.01*** (.00) (.00) (.00) (.00) Ideology=Right.06***.06***.04***.04*** (.00) (.00) (.00) (.00) Age.00***.00*** (.00) (.00) (.00) (.00) Rural.02***.02***.02***.02*** (.00) (.00) (.00) (.00) Economic Perception Worse -.05*** -.05*** -.06*** -.06*** (.00) (.00) (.00) (.00) Married (.00) (.00) (.00) (.00) Widowed/Divorced ** -.01** (.01) (.01) (.01) (.01) Country-Level GNI per capita.49***.49***.65**.65** (.17) (.17) (.26) (.26) Observations 24,813 24,813 24,517 24,517 Number of groups Wald χ *** p<.01, ** p<.05, *.1. Multilevel regression coefficients, standard errors are in parentheses. Ideology (No Answer), & Single are excluded as reference categories. Class and Unequal Representation 22

25 In Figure 2, we use models 2 (left panel) and 4 (right panel) in Table 1 to calculate and plot the expected value of each dependent variable across the range of working class representation in our sample (from a low of three percent to a high of 26 percent). The top panel in Figure 2 offers support for Hypothesis 1: increases in the percentage of legislators from the working class improve perceptions of representation for working class citizens. For instance, a move from the lowest to the highest level of working class representation is associated with a.16 increase in the expected value of the political party index for working class citizens, or a 60 percent improvement in perceptions of representation. Similarly, an increase from the lowest to highest level of working class representation is associated with a.15 increase in the expected value on the legislature index, or a 44 percent increase in perceptions of representation. These findings demonstrate support for our theory that increases in descriptive representation of the working class will be associated with improved perceptions of representation among working class citizens. Turning next to Hypothesis 2, the bottom panels in Figure 2 display graphically the relationship between descriptive representation of working class legislators and symbolic representation for non-working class citizens. Similar to the previous set of results, we show that increases in the percentage of working class legislators are associated with improved perceptions of representation from political parties and legislatures for non-working class citizens. Taken together, the results from the top and bottom panel of Figure 2 demonstrate that working class representation is good for all citizens not merely those from the working class. That said, the relationship for non-working class citizens is slightly weaker than the relationship observed for working class citizens when respondents are asked about their affective attachments to parties. The positive and significant relationship graphed in the panels on the left in Figure 2 demonstrates that increases in the percentage of working class legislators improves perceptions of representation from political parties more for working class citizens than non- Class and Unequal Representation 23

26 Figure 2. Effect of Working Class Representation on Symbolic Representations Class and Unequal Representation 24