Racial Rhetoric in Black and White: Symbolic and Substantive References in U.S. House Speeches

Transcription

1 Racial Rhetoric in Black and White: Symbolic and Substantive References in U.S. House Speeches Bryce J. Dietrich, Matthew Hayes, Saumil Dharia, Stella L. Wancke July 21, 2017 Abstract The rhetoric legislators use to discuss issues of race is both important and understudied. In this paper, we explore the extent that the presence of black legislators influences the nature of racial dialogue in the U.S. House. We ask three questions. First, do black and white members of Congress talk about issues involving race at different rates? Second, when black and white MCs talk about race, do they do so in different ways? Third, are these rhetorical differences consequential for the evaluations and satisfaction of African-American constituents? Using data from 790,654 floor speeches in the U.S. House and data from an original survey experiment, we demonstrate that black legislators are more likely to talk about civil rights, and when they do so they employ significantly more symbolic references to African Americans struggle for civil rights in the United States. These symbolic references have a clear positive effect on the evaluations of black constituents, suggesting that if white legislators engaged in more symbolic responsiveness they could boost the satisfaction of their black constituents. Comments and suggestions welcome. We are grateful to Elizabeth Maltby and Spencer Piston as well as conference participants at MPSA for their helpful feedback. Assistant Professor of Social Science Informatics, Department of Political Science, University of Iowa Assistant Professor, Department of Political Science, Indiana University, Bloomington Ph.D. candidate, Department of Political Science, University of Missouri, Columbia Ph.D. student, Department of Political Science, University of Texas, Austin 1

2 1 Introduction The inclusion of underrepresented groups in legislatures can have wide-ranging consequences for political representation. Their presence can affect agenda-setting, substantive outcomes, as well as influence the political trust and engagement of constituents from underrepresented groups. One of the strangest arguments about the importance of legislators from underrepresented groups is their heightened ability to voice the concerns of their community. Through their shared identity, these officeholders are both more willing and able to articulate the concerns of constituents who might otherwise lack a champion in the legislature. For example, in the immediate aftermath of the shooting of Trayvon Martin in Florida, Representative Bobby Rush (D-IL) donned a hoodie on the floor of the House as part of his speech protesting racial profiling. This symbolic speech helped Rep. Rush connect with African American constituents both within his district and across the country, and expressed his solidarity with their concerns. This study is interested in whether there are systematic differences in the use of such symbolic speeches when discussing racial issues in Congress, and whether these speeches have an impact on constituents. The rhetoric legislators use to discuss issues of race is both important and understudied. Minority legislators can change the terms of debate around racial issues, participating in what Daniel Gillion calls discursive governance (Gillion, 2016). Whether it is delivering a floor speech on civil rights or highlighting the importance of police brutality, there are a number of ways a representatives rhetoric can be used to advance minority interests. When representatives use symbols to talk about substantively important issues, they empathize with the African-American community. By answering three related questions, this study explores this relationship in the U.S. House of Representatives, adding considerably to our understanding of the use of symbolism and descriptive representation more broadly. First, do black and white members of Congress (MCs) talk about issues involving race at different rates? If they do, then well established relationships found in the floor speeches of female representatives (e.g., Osborn and Mendez, 2010; Pearson and Dancey, 2011a,b), seem 2

3 to also apply to African-American representatives. Second, when black and white MCs talk about race, do they do so in different ways? Regardless of the underrepresented group, the answer to this question is noticeably lacking in the literature. This is unfortunate, because the quality of discursive governance is just as important as the quantity. Third and finally, are these rhetorical differences consequential for the evaluations and satisfaction of African American constituents? Those interested in descriptive and substantive representation often take for granted that floor speeches carry some weight with underrepresented communities. Although theoretically supported, the present study is the first to test this claim empirically. Using 790,564 floor speeches from we show that not only are black representatives more likely to talk about issues that are important to African Americans, but they do so in ways that are distinct from their white counterparts. To these representatives, Rosa Parks and Martin Luther King, Jr. are not just historic icons, but important symbols that resonate with African-Americans and link to contemporary issues. Similarly, Plessy v. Ferguson and Brown v. Board of Education are more than important Supreme Court cases they are part of the lexicon that African-Americans use to talk about modern civil rights. Our study expands the notion of discursive governance to include a variety of symbols of the struggle for civil rights in the African American community. These symbols are used to advance both the descriptive and substantive interests of minority communities by reinforcing a sense of shared identity and calling for policy change. In the sections that follow, we demonstrate that black MCs are more likely to talk about civil rights, and when doing so, they are more likely to use symbolic references to the struggle for civil rights. White MCs are less likely to talk about civil rights or make symbolic references, and tend to do so only when they have strong electoral incentives. We also show African American constituents evaluate representatives more positively when they use these symbols, especially when they are used by white representatives. Taken together, this suggests that legislators can and do use symbolic floor speeches to advance minority interests on Capitol Hill and improve the evaluations of constituents in their district. 3

4 2 Race, Floor Speeches, and the Symbolic Representation of African-Americans 2.1 The Importance of Presence for Underrepresented Groups There are myriad reasons why it is important for underrepresented groups to achieve a presence in decision-making bodies. Studies have found that the presence of women and minority legislators affects agenda-setting behavior (Bratton and Haynie, 1999; Minta and Sinclair- Chapman, 2012), bill sponsorship and cosponsorship activities (Whitby, 2002; Tate, 2003; Swers, 2005; Rocca and Sanchez, 2007), and provision of service and pork to the district (Grose, 2011). In this way, achieving descriptive representation having officials who share your identity and background is important for the substantive concerns of underrepresented groups. The presence of women and minorities is also important for its effects on constituents. Women s access to office can have a role model effect, leading to a boost in the civic engagement of female constituents (Hansen, 1997; Atkeson, 2003; Barnes and Burchard, 2012). And the presence of minority and female officeholders can improve constituent s attitudes towards elected officials (Tate, 2003; Lawless, 2004), perhaps even compensating for receiving substantively unfavorable outcomes (Hayes and Hibbing, 2016). Although the policy and instrumental value of having underrepresented groups present in decision-making bodies should not be understated, perhaps the strongest argument for their inclusion is in the domain of deliberation. Because of the shared identity between women and minority officeholders and their constituents, these legislators should be more capable in expressing the concerns of their group, even when those concerns are uncrystallized (Mansbridge, 1999). There is also empirical evidence to support this claim. Although there is some disagreement (c.f. Gerrity, Osborn and Mendez, 2007; Osborn and Mendez, 2010), female legislators are more likely to speak about women s issues in deliberation (Kathlene, 1995; Pearson and Dancey, 2011a,b), and such differences can change the terms of legislative debate (Levy, 4

5 Tien and Aved, 2001). This suggests minority and female legislators can bolster what Dan Gillion calls discursive governance. In reference to the case of African-Americans, even if deliberation does not ultimately reflect the policy concerns of underrepresented groups, a dialogue on race that takes place in government draws attention to racial inequality on the policy agenda and informs governmental officials and the American public alike of the continuing inequalities that persist (Gillion, 2016, 19). In this way, African-Americans presence in deliberation can advance the extent to which the policymaking process acknowledges the special concerns of their racial group, which might help descriptively similar constituents feel better represented. This should manifest itself in terms of boosting African-Americans feelings of symbolic representation. Symbolic representation is concerned not with who the representatives are or what they do, but how they are perceived and evaluated by those they represent (Schwindt- Bayer and Mishler, 2005). By giving a voice to the concerns of the black community, black representatives can make that community feel that they are better represented, irrespective of whether such voice actually shifts policy outputs. The key question is how discursive governance might boost symbolic representation. We argue that the rhetoric used in deliberation can serve as a powerful means through which constituents can obtain symbolic representation. Pitkin uses the example of a flag as a symbolic representation; because the flag symbolizes (suggests, evokes, arouses feelings appropriate to) the honor and majesty of the United States (Pitkin, 1967, 98), it can essentially serve as a stand-in to provide symbolic meaning in the minds of those being represented. In much the same way, symbolic rhetoric can evoke the same feelings in the minds of constituents in order to enhance their feelings of being well represented. This should be particularly true for African Americans on the issue of civil rights. The struggle for civil rights in the United States is associated with many symbols that carry meaning for African Americans (and likely for many non-black Americans). For example, evoking Rosa Parks as a symbol will not only convey the thought of a pioneer in the struggle, but also the ideals of egalitarianism and racial justice. 5

6 2.2 The Importance of Intrinsic and Extrinsic Motivation When used during legislative speech, symbols like those associated with the Civil Rights Movement and the broader struggle for racial justice can convey meaning to black constituents. When black MCs discuss symbols that are readily identifiable by the black community, they are more than simply pandering; instead, they are connecting themselves with the enduring fight for civil rights and providing symbolic representation to black constituents by reminding them that they have a champion in the halls of government. In this way, a reference to Martin Luther King, Jr., the March on Washington, or Selma, can be seen as an attempt to provide symbolic representation to black constituents. However, not all legislators are intrinsically motivated to provide symbolic representation to black constituents (Broockman, 2013). Representatives who have an intrinsic motivation use symbols like Rosa Parks and Martin Luther King, Jr. because they feel a personal attachment to the concerns of the black community. For example, a representative like John Lewis, who marched in Selma and spoke at the March on Washington, likely makes symbolic references to the struggle for civil rights because those symbols hold the same deep value to him that they do to many of his African American constituents. Making such references might benefit representatives on Election Day, but intrinsically motivated legislators would make symbolic references regardless of whether these electoral incentives existed. Indeed, John Lewis spoke about symbols of civil rights with great reverence and meaning well before he became a member of Congress. Other representatives likely use symbolism because it aligns with their strategic pursuit of reelection (Broockman, 2013, 523). Although these representative may feel some solidarity with the black community, they rarely make symbolic references unless they have an extrinsic motivation to do so. For example, many Southern Democrats represent large African American populations. In order to bolster their support among this community, these representatives should be more likely to reference issues of importance to the black community, such as civil rights. When they talk about these issues, they should be likely to use symbolic references 6

7 that will resonate with their African American constituents because of the potential electoral benefits of doing so. This is not to say that these individuals do not understand the importance of these symbols to the black community, but that legislators who are extrinsically motivated use their rhetoric to meet the representational demands of constituents rather than speaking from a deep-seated commitment to the struggle for civil rights. For example, Broockman (2013) found white representatives tend to respond slower to s from out-of-district black constituents, whereas black representatives responded at the same speed, regardless of whether the black constituent was inside or outside of their district. This comports well with several observational studies, most notably Overby and Cosgrove (1996) who found white incumbents who lost black constituents during redistricting became less sensitive to the concerns of African-Americans (540). Hutchings, McClerking and Charles (2004) reached a similar conclusion. Using voting behavior from 101st 103rd Congress, they show white representatives tend to only court black support when they might receive the maximum electoral advantage (465). Although their findings are largely restricted to high profile issues in the South, Grose (2011) reaches essentially the same conclusion when he argues black legislators may not alter their levels of substantive representation dependent on the black population of their districts to the same extent that white legislators might (23). 3 Theoretical Expectations Regarding Legislative Speech on Civil Rights Several potential mechanisms could influence what issues legislators bring to the floor of the U.S. House, and the rhetoric they employ to discuss those issues. The first such mechanism is a shared racial identity. Mansbridge (1999) argues that descriptive representatives possess a shared history and an innate advantage in voicing the concerns of their group. For this reason, we expect that black MCs should be more likely to talk about issues of importance to their 7

8 community than white members of Congress. Although there are many issues of importance to the black community, none are more prominent and central than the issue of civil rights. This leads to our first hypothesis. Hypothesis 1: Black members of Congress will be more likely to talk about civil rights than white members of Congress. Whether it is service, allocation, or certain policies, constituents demand a variety of things from their legislators. Thus, we might expect that legislators will provide speech on racial issues to the extent that their constituents and particularly those constituents who are electorally consequential demand it. Recent work has explored the extent to which constituents demand, and legislators supply, different forms of representation (Harden, 2015). However, we argue symbolic representation is also an important, yet understudied, demand African-Americans may also advance on Capitol Hill. Over the past several decades, a recurring finding is that African-American constituents place a higher premium on service and allocation responsiveness than do their white counterparts (Tate, 2003; Griffin and Flavin, 2011; Hayes, 2016). This follows directly from the electoral incentives of legislators. Similar to policy responsiveness (Miller and Stokes, 1963), legislators respond to constituent service and allocation demands in order to ensure they have support in the next electoral cycle. Symbolic issue references function in a similar way. Indeed, virtually all students of symbolic activity contend that it is electorally motivated..., suggesting legislators likely reference symbolically valuable issues like civil rights when constituents demand greater attention should be paid to African-Americans (Hill and Hurley, 2002, 220). Although we generally expect increasing the size of the black constituency will make legislators more responsive to black interests (Hutchings and Valentino, 2004, 397), we also expect this effect to be more pronounced for white representatives. Beginning with perhaps Cameron, Epstein and O Halloran (1996), scholars have shown white representatives are more likely to advance black interests when they are extrinsically motivated to do so. Even though we acknowledge some studies have found either mixed (Fleisher, 1993) or null re- 8

9 sults (Bullock, 1995), we generally expect white representatives will be more responsive to changes in the number of black voters in their district. This is especially true when it comes to high profile issues like civil rights. Indeed, Hutchings (1998) found that white southern Democrats were more likely to vote on the 1990 Civil Rights Act when they represented a large number of African-Americans. On high profile issues, like civil rights, African-American constituents do not give their representatives much leeway, making white representatives more responsive. We expand the notion of responsiveness to include speech-making. For example, African- American voters may demand both black and white MCs advance civil rights protections. However, legislatively advancing such issues may prove of little consequence since such legislation may be extraordinarily difficult to pass. In this way, African-Americans may demand different types of responsiveness, such as floor speeches which bring attention to issues they care about. While some African-Americans may view such talk as being cheap, similar to other issues like climate change that have little chance of making it onto the legislative agenda, talk is perhaps the best they can ask for given the political environment. For these reasons, we should expect that legislators should provide speech that attends to the issues important to African-Americans in our case, civil rights when African- American constituents provide an electoral incentive to do so. Although we expect the number of African-Americans in the district will have some affect on black representatives, we expect white representatives will generally be more responsive to these demographic changes. While we cannot say definitively why representatives choose to speak to civil rights, previous literature has suggested white representatives are more extrinsically motivated. We expect to find the same relationship when it comes to speech-making. This leads us to our second hypothesis: Hypothesis 2: (White) legislators with a higher percent of African-Americans in their district will be more likely to talk about civil rights. Not all speeches on civil rights will convey equal amounts of meaning and commitment to constituents. A legislator might discuss civil rights in the abstract, but in a way that is divorced 9

10 from the struggle for civil rights in the African American community. Another legislator might speak on the same topic, but invoke well-known symbols of the Civil Rights Movement that convey to her constituents that she shares their experiences and concerns for civil rights. In this way, we do not expect all legislative speech on civil rights will produce the same amount of symbolic representation. To confer symbolic meaning and thus promote symbolic representation, legislators can rely on well-known symbolic cues to signal their understanding of and commitment to the struggle for civil rights. Symbolic references to the importance of certain civil rights icons, for example, can be used to contextualize the many obstacles African Americans faced in fighting for civil rights. We expect white MCs are more reluctant to invoke these symbols. After all, it is thought that the shared experience of black representatives is what allows them to speak...with a voice carrying the authority of experience (Mansbridge, 1999, 644). Constituents can rely on body language, choice of words, accent, and other external signals to predict the likelihood of a large body of experience shared with them and other African-Americans (Mansbridge, 1999, 645). This can clearly be seen in a speech from John Conyers (D-MI) on July 14, 2005: The United States is one of the richest nations in the world so I ask: Why aren t our schools thriving? Why are so many African-Americans unemployed? Why are health care, housing, and college education so difficult to afford? Civil rights pioneer Rosa Parks displayed a defiant act of courage 50 years ago so that these questions would not be raised today. I know that this Congress and the people of this nation can work to further the ideals of Ms. Parks and the Civil Rights Movement. In this speech Conyers, a black Democrat, invokes the memory of Rosa Parks in order to remind Congress that there is more work to be done on the struggle for civil rights. His word choice signals to his black constituents that he has just such a large body of experience shared with them, and is a dedicated champion of their issues. This speech demonstrates Conyers s ability to speak eloquently about Rosa Parks as a symbol of civil rights and link her 10

11 to contemporary issues of importance to the black community. Doing so requires a cultural sensitivity and knowledge of the importance of Rosa Parks that might make representatives who do not share an identity as an African-American more reluctant to invoke this symbol. This is the crux of why it is critical to understand not only whether legislators discuss civil rights, but whether they use symbolic language when they do so. When speaking on civil rights, the choice to invoke recognizable symbols can help promote a sense of shared experience with black constituents. We view this word choice as important it elevates civil rights discourse above simple substantive concerns to also conveying a sense that a representative has shared identity and interests with black constituents. Although every MC could theoretically invoke symbols of the civil rights struggle, some may not be comfortable doing so. In particular, a shared lived experience as an African-American should make it much easier to know which symbols carry meaning and how to invoke them. For this reason, we expect that MCs who descriptively represent African-Americans should be more likely to invoke such symbols. Hypothesis 3: African-American legislators will be more likely than white legislators to invoke symbolism in their discussion of civil rights. Immediately after the Voting Rights Act was passed, many white representatives had to walk a fine line. On the one hand, the increased number of African-American voters meant they had strong incentives to speak to descriptive issues like civil rights. On the other hand, they did not want to alienate white voters in their districts who may have animosity towards African- Americans and the issues they support. Ultimately, Hutchings, McClerking and Charles (2004) argue this meant white representatives had to make tough choices about when and where they could support the interests of the black community. Although we have made tremendous progress in the representation of African Americans, we still expect that white representatives have to walk a similar line when using symbolism. In many ways, white representatives likely feel as though it is inappropriate for them to use symbols like Rosa Parks or Martin Luther King, Jr. especially in the context of civil rights. Not only did those symbols represent a movement that is an essential part of the identity of 11

12 many African Americans, but the movement still represents an ongoing struggle against racial inequality. We expect this reluctance is more likely to fade when white representatives are elected by a large number of African American voters. Conversely, we expect black MCs will use symbolism when discussing civil rights, regardless of the number of African Americans in their district. Unlike their white counterparts, these representatives are more likely to be intrinsically motivated, meaning they show solidarity with the black community because they personally identify with that group. They too may be influenced by electoral forces, but their use of symbolism is less likely to be dependent on whether they have a large African-American constituency. Just as black MCs are likely to talk about African-Americans across a wide number of issues, these same representatives likely reference symbols like Rosa Parks and Martin Luther King even when it is not in their interest to do so. This leads us to our fourth hypothesis: Hypothesis 4: (White) legislators with higher percent of African-Americans in their district will be more likely to invoke symbolism in their discussion of civil rights. Finally, it is important to consider the effect such speeches have on constituents. It might be the case that representation in Congress is primarily about feelings of empowerment, rather than an actual influence over the policy-making process. As citizens who were unrepresented or underrepresented since the founding of American democracy, African Americans might care more for having a voice in the chamber or a seat at the table than actually receiving policy concessions. This would comport with the large literature on empowerment effects, which suggests that being descriptively represented might confer benefits in terms of electoral participation, trust, and efficacy (Bobo and Gilliam Jr., 1990; Gay, 2001, 2002). Regardless of whether a representative is white or black, when a legislator speaks on issues of importance to the black community, and invokes symbols that convey a shared understanding of the issue, they should promote positive evaluations among their African American constituents. Although we have no strong expectations regarding whether African Americans will respond more positively to the use of symbolism by white representatives, previous literature 12

13 suggests white representatives are also likely to be rewarded. Indeed, in order for white representatives to be extrinsically motivated, they must get some electoral benefit when they are responsive to the black voters in their district. If responsiveness also includes speech-making, then white representatives should be evaluated more positively by African Americans when they invoke symbols when discussing civil rights. For these reasons, we expect both black and white representatives will be evaluated more positively by African Americans when they invoke symbolism. Hypothesis 5: The use of symbolic language on civil rights should boost evaluations among African American constituents. In Sections 5 and 6 we find evidence consistent with many of our theoretical expectations. Not only do we show that black MCs are more likely to speak about civil rights, but when they do they are more likely to invoke symbolism. We also find white MCs seem to only discuss civil rights when they have strong electoral incentives. The same can be said for their use of symbolic references. Our experimental results found in Section 7 also show white MCs are rewarded by African-Americans when they use symbolism. In the next section, we take up the difficult task of defining and measuring symbolism as it relates to the African American struggle for civil rights. 4 Description of Coding Scheme and Floor Speech Data For this study we collected data on the content of legislative floor speeches using the now retired Capitol Words project. 1 We restrict our analyses to the U.S. House, and our data include all floor speeches given from Once speeches were collected, we identified those delivered by African-American, Latino, and Asian-American MCs. African-Americans and The Capitol Words database only went back to 1996 and we began our data collection in the Fall of Thus, the dates have no substantive importance. 13

14 Figure 1: Coding Symbolic References Found in Civil Rights Speeches Does the speech contain more than 150 words? no 1.00 yes Done... at least two symbolic references? no 0.94 yes symbolic non-symbolic Percent Agreement = 0.97 Note: This shows our coding procedure. The overall agreement is shown in the bottom right. Below each fork is the percent agreement for that specific question. Additional details can be found in the Supplemental Information. Latino were identified using the House Archives. 3 Asian-American MCs were identified using the most recent Congressional Research Service report. 4 If a member of Congress did not fit into any of these racial categories the representative was said to be white. Similar to Schickler, Pearson and Feinstein (2010) we used the phrase civil rights in order to determine the degree to which a MC talked about civil rights. 5 For white MCs this yielded 731,283 speeches of which 2,450 were about civil rights. For African-American MCs we collected 59,281 with 3,095 being about civil rights We fully acknowledge there are likely some speeches that talk about civil rights without using the phrase civil rights. This is a known limitation of using keyword searches. In the Supplemental Information we include several different model specifications all of which yield similar results. Undoubtedly, we are missing some relevant speeches by focusing on only those that contain the phrase civil rights, but we also think there is some value to mentioning civil rights, especially to African-Americans. This is one of the reasons why we focus on speeches containing this specific phrase. For example, it is difficult to imagine a speech talking about the struggle for civil rights without using the phrase civil rights. This is because references to civil rights carry considerable weight. Most know what civil rights means to the African-American community, meaning when it is uttered on the House floor it is done so intentionally. We argue this value outweighs any unsystematic error we introduce by using a simple keyword search. Schickler, Pearson and Feinstein (2010) use essentially the same keyword search to understand how the Democratic and Republican view on civil rights changed over time. The present study uses their technique to identify the degree to which black and white MCs talk about civil rights. With that said, the main contribution of the present study is introducing the importance of symbolism to civil rights discussions which is why we hand coded this variable. 14

15 To determine whether black and white MCs talk about civil rights differently, we used a modified version of the codebook developed by Hill and Hurley (2002). More specifically, two independent coders determine whether each floor speech contained symbolism. To do so we used the decision tree outlined in Figure 1. The first question asked was whether the speech contained 150 words. We found some speeches were a sentence or two. Ultimately, we concluded that these speeches had little representative value, so we created a 150 word threshold to prevent these speeches from being considered. Once this question was answered, we determined whether the speeches had at least two symbolic references. These symbols were drawn from people (e.g., Martin Luther King, Jr.), events (e.g., march on Washington), acts (e.g., Brown versus Board of Education), and concepts (e.g., slavery). If a speech had two symbolic references it was considered to be a symbolic speech. We encourage readers to review the full list of the symbolic references provided in the Supplemental Information. 6 To assess intercoder reliability, we randomly sampled ten percent of the speeches (n = 558). Using two independent coders, we achieved a percent agreement rate. Krippendorff s alpha (0.801), Cohen s Kappa (0.801), and Kendall s coefficient of concordance (0.902) were all found to be within acceptable ranges (Neuendorf, 2002). For the latter two measures, we calculated a z (18.60) and χ 2 statistic (960), respectively. In both instances, p approximated zero, meaning there was significant agreement between our two coders. The overall agreement rate is shown in the bottom right hand corner of Figure 1. At each branch of the decision tree we report the percent agreement on that particular question. We achieved a percent agreement on all coding decisions, meaning we can code the use of symbolism significantly above chance (χ 2 = , df = 1, p < 0.05). In addition to race, we also collected the proportion of African-Americans in each speaker s 6 Internally, we have debated over whether the Voting Rights Act and Civil Rights Act should be considered as symbols. In some instances, they are used as symbols. For example, one speech referred to the Voting Rights Act as landmark legislation that underlined the struggle for equality. Similarly, another speech referred to the Civil Rights Act as signaling the end of segregation. We have tried our best to differentiate these sorts of speeches from wonky policy extensions of either the Voting Rights or Civil Rights Acts. In total, less than one percent of the symbolic speeches uses these pieces of legislation as their only symbolic references, meaning a very small fraction of our data is affected by this coding decision. 15

16 district. This was obtained from the U.S. Census Bureau. For all years after 2000, we had a measure for each year. Prior to 2000, we used a combination of the 1990 and 2000 censuses. While we acknowledge the importance of other control variables, below we only include controls for ideology (using DW-NOMINATE scores), seniority, party identification, gender, and whether the speech was given by a representative from the South. 7 Both seniority and ideology were downloaded from the Voteview website. 8 We also included a dummy variable for whether the speech was delivered during an election year. In order to determine whether black and white MCs talk about civil rights differently, we answer four broad questions. First, when do white MCs talk about civil rights? Second, when do white MCs use symbols, like Rosa Parks and Martin Luther King? Third, when white MCs use symbolism, what symbols do they use? Finally, when white MCs engage in symbolic representation how do African-Americans respond? The next section will answer the first three questions. The final question will be answered using an experimental design in the following section. The details of our experiment will be explained at that time. 5 When Do Members of Congress Talk About Civil Rights? When do white MCs talk about civil rights? Figure 2 provides an initial answer to this question. Here, we compare black and white mentions of civil rights, climate change, free trade, gun control, health care, national defense, public schools, and social security. In every year since 1996, black legislators have uttered the phrase civil rights significantly more often than white legislators. For example, in 2013 black legislators were 8.85 times more likely to use the phrase civil rights than white legislators. The smallest difference between black and white legislators on civil rights was in 2002 when white MCs said civil rights in 7 A representative from any one of the following states was considered to be from the South. These staters are AL, AR, GA, FL, LA, MS, SC, and TX

17 Figure 2: The Proportion of Speeches Mentioning Civil Rights Civil Rights 0.06 Health Care µblack =.046 µwhite =.008 Public Schools µblack =.068 µwhite =.053 µblack =.022 µwhite =.009 Proportion of Black Speeches - Proportion of White Speeches Social Security 0.06 Gun Control µblack =.025 µwhite =.020 Free Trade µblack =.002 µwhite =.001 µblack =.003 µwhite = Climate Change National Defense µblack =.000 µwhite =.003 µblack =.004 µwhite = Year Note: For a given phrase, this figure shows the proportion of speeches delivered by black MCs (µblack ) including that phrase minus the proportion of speeches delivered by white MCs (µwhite ). The bars indicate the difference for a single year. The overall proportions can be found in the top left percent of their speeches. Even at this high water mark, black legislators still mentioned civil rights four times as often as white legislators. Moreover, this does not appear to be simply an issue of black members of the House being overwhelmingly Democratic and simply discussing liberal issues more frequently. The phrase climate change is actually mentioned more by white than by black MCs, despite the fact that in the House the term climate change is almost exclusively uttered by Democrats. We do see black MCs discussing the issue of public schools (arguably an issue of particular importance to the black community) more frequently than white MCs, but across the rest of the phrases we examined we do not see the same striking racial differences as on civil rights. To put this into perspective, we considered, on average, how often did black legislators give civil rights speeches as compared to white legislators. Here, we found that black legislators 17

18 gave 6.08 times as many speeches. By comparison, when the same is done for the other issues, in every instance the difference between black and white legislators does not even come close to equaling this amount. In fact, there are only two instances when black legislators outpace white legislators by more than 2 to 1: public schools and civil rights. However, public schools is only used 2.72 times more by black as compared to white legislators. This suggests the rate at which civil rights is referenced by black MCs is over two times higher than its next closest competitor. Similar to Figure 2, Table 1 shows that black and white legislators fundamentally differ in the degree to which they discuss civil rights. In these models, the unit of analysis is an MC in a given term. For example, Representative Bobby Rush (D-IL) spoke in every year of our data ( ), meaning he appears in each model nineteen times. Since the dependent variable is the number of civil rights speeches delivered in a given term either a multilevel Poisson or negative binomial regression should be used. In the Supplemental Information, we report the results from a series of likelihood ratio tests all of which suggest a negative binomial regression is the most appropriate given the distribution of civil rights speeches. To account for the likely clustering of errors around each representative, we estimated a randomly varying intercept for each MC. We also included an offset for the number of speeches to control for the fact that some MCs generally talk more than others. In all four models, white MCs are significantly less likely to talk about civil rights. Models 1 and 2 consider the direct effect of race on the number of civil rights speeches delivered. In the Civil Rights panel of Figure 3 we plot the predicted number of speeches from Model 2, holding all other variables at their mean and modal values including the offset number of speeches which was set to the non-zero mode (28). Thus, the counts in Figure 3 are the predicted number of speeches on civil rights for every 28 floor speeches delivered. With this in mind, when all other variables are set to their mean and modal values, African-American representatives are predicted to give 0.85 floor speeches on civil rights out of every 28 speeches they deliver in a given legislative term. Undoubtedly, this is an extraordinarily small number, but it is over seven times higher than white MCs who are only predicted to deliver 0.12 speeches on 18

19 Table 1: White Members of Congress Deliver Fewer Civil Rights Speeches Dependent variable: Number of Civil Rights Speeches (1) (2) (3) (4) Fixed Effects Constant (0.231) (0.289) (0.264) (0.330) White MC (0.245) (0.206) (0.275) (0.274) Percent Black (0.322) (0.449) (0.447) Democrat (0.303) (0.301) DW-Nominate (0.300) (0.298) Seniority (0.008) (0.008) Male (0.141) (0.140) South (0.134) (0.134) Election Year (0.038) (0.038) White MC Percent Black (0.620) (0.618) Random Effects MC (1.704) (0.987) (1.101) (0.972) N 1 7,915 6,651 6,659 6,651 N θ (0.911) (0.949) (0.965) (0.987) Log Likelihood 6, , , , AIC 12, , , , Note: Unit of analysis is a MC in a given term. Coefficients are from a multilevel negative binomial regression with a random intercept for each MC. Log-likelihood tests reported in the Supplemental Information suggest this is the most appropriate model given the distribution of dependent variable (# of civil rights speeches). Positive values imply more civil rights speeches were delivered. Offset included for the total number of speeches. Levels of significance are reported as follows: p <.1; p <.05; p <.01. The standard errors are reported in the parentheses. 19

20 Figure 3: Predicted Number of Civil Rights Speeches and Symbolic References Note: Predicted values were calculated using the coefficients from the second column in Tables 1 (Model 2) and 2 (Model 6). The former are reported in the Civil Rights panel. The latter are reported in the Symbolism panel. and imply White MC was set to 0 and 1, respectively. All other variables are held constant at their mean and modal values. Offsets are included, meaning in the Civil Rights panel each bar represents the number of speeches mentioning civil rights for every 28 speeches delivered. In the Symbolism panel, the offset is 1. This means bars should be interpreted as the number of speeches making a symbolic reference for every civil rights speech. Error bars represent 95 confidence intervals from 1,000 bootstrapped samples. 20

21 civil rights for every 28 speeches delivered. However, Models 3 and 4 show white MCs become more similar to their black colleagues as the proportion of African-Americans increases. To help interpret this interaction we plotted the predicted number of civil rights speeches as the proportion of African-Americans in the district ranged from the minimum (0) to maximum (0.79), holding all other variables constant at their mean and modal values including the speech offset which was again set to 28 speeches. The results are reported in Figure 5. Here, we subtracted the predicted number of speeches delivered by a white member of Congress from the predicted number of speeches delivered by a black member of Congress. When the proportion of African-Americans is set to the minimum, black MCs deliver 1.43 more speeches about civil rights for every 28 speeches they deliver. This difference declines substantially as the proportion of African-Americans increases. For example, when 20 percent of the district is African-American (or the first quartile), then black MCs only deliver 1.04 more speeches on civil rights for every 28 speeches they deliver. When the proportion increases to 60 percent (or the third quartile) that number is cut nearly in half to 0.45 more civil rights speeches for every 28 delivered. 6 How Do Members of Congress Talk About Civil Rights? From these results it is readily apparent that civil rights is rarely discussed on the House floor. Indeed, of the 790,564 speeches in our data, less than 1 percent (5,545) talk about civil rights. To us, this is not entirely surprising. For white MCs, there is little incentive to talk about civil rights since the topic often speaks to a small subset of their constituency. For black MCs, civil rights may not be as useful as symbolic references which can be applied to a large number of issues. For example, we found 60,631 speeches used at least one of the symbols outlined in Figure 4. Even though such results are far from exhaustive, they suggest that black MCs may use symbols in other contexts to symbolically represent African-Americans. With that said, when do white MCs make symbolic references? We found that 3,206 of the 5,545 speeches that talked about civil rights used some form of symbolism. Even though 21

22 black representatives are significantly more likely to use symbolism, the difference is only and 2.97 percentage points when talking about civil rights and other issues, respectively. Larger differences were found when we considered specific types of symbols. This is shown in Figure 4. As you can see, black MCs used nearly every symbolic category more than white MCs when talking about civil rights. For example, 975 speeches referencing civil rights also talked about Martin Luther King. Of these, percent of these speeches were delivered by black MCs. Starker differences are found when it comes to categories that deal with discrimination, like racism and segregation. 306 and 714 civil rights speeches mentioned racism and segregation, respectively. Of these, percent were delivered by black MCs. Conversely, white MCs actually delivered 9.62 percentage points more civil rights speeches referencing Abraham Lincoln than black MCs, suggesting white MCs may be more comfortable using some symbols as compared to others. Even though black MCs are significantly more likely to use symbolism, white MCs still use symbolism, especially when they represent predominantly African-American districts. This is shown in Table 2. In these models the dependent variable is the number of speeches that each MC delivered in a given legislative term. Since we only coded civil rights speeches, these models are restricted to MCs who delivered at least one civil rights speech in a given legislative term. Other than that, the models are identical to the ones used above. As you can see, the results are nearly identical. In general, when white MCs talk about civil rights they use less symbolism than black MCs. However, as the proportion of African-Americans in the district increases the difference between black and white MCs decreases. This is captured in Symbolism panel of Figure 5. As before, we allowed the proportion of African-Americans in the district to range from the minimum (0) to the maximum (0.79) holding all else constant at the mean and modal value including the offset number of civil rights speeches which again is the non-zero mode (1). When the proportion of African-Americans is set to the minimum, black representatives deliver 0.64 more symbolic speeches for every civil rights speech they deliver in a given leg- 22

23 Figure 4: The Proportion of Speeches Using Various Symbolic References Note: For a given phrase, this figure shows the proportion of speeches delivered by black MCs including the indicated symbolic reference (see x-axis) minus the proportion of speeches delivered by white MCs. In the top panel, we only include speeches which reference civil rights. In the bottom panel, we only include speeches which do not reference civil rights. 23

24 Table 2: White Members of Congress Use Less Symbolism When Talking About Civil Rights Dependent variable: Number of Symbolic Speeches (5) (6) (7) (8) Fixed Effects Constant (0.106) (0.336) (0.233) (0.357) White MC (0.128) (0.196) (0.255) (0.267) Percent Black (0.369) (0.446) (0.454) Democrat (0.434) (0.426) DW-Nominate (0.411) (0.403) Seniority (0.011) (0.011) Male (0.145) (0.142) South (0.145) (0.139) Election Year (0.061) (0.061) White MC Percent Black (0.674) (0.662) Random Effects MC (0.623) (0.579) (0.571) (0.543) N 1 2,064 2,063 2,064 2,063 N Log Likelihood 1, , , , AIC 3, , , , Note: Unit of analysis is a MC in a given term. Coefficients are from a multilevel Poisson regression with a random intercept for each MC. Loglikelihood tests reported in the Supplemental Information suggest this is the most appropriate model given the distribution of the dependent variable (# of symbolic speeches). Positive values imply more symbolic speeches were delivered. MCs had to deliver at least 1 civil rights speech to be included. Offset included for the number of civil rights speeches. Levels of significance are reported as follows: p <.1; p <.05; p <.01. The standard errors are reported in the parentheses. 24

25 Figure 5: Predicted Number of Civil Rights Speeches and Symbolic References Note: Predicted values were calculated using the coefficients from the last column in Tables 1 (Model 4) and 2 (Model 8). The former are reported in the Civil Rights panel. The latter are reported in the Symbolism panel. In the x-axis, the proportion of African-Americans in the district is allowed to vary from the minimum (0) to maximum (0.79). In the y-axis, the predicted number of black speeches (White MC = 0) is subtracted from the predicted number of white speeches (White MC = 1). Positive values indicate white MCs were predicted to deliver more speeches. All other variables are held constant at their mean and modal values. Offsets are included in both models. Error bars represent 95 confidence intervals from 1,000 bootstrapped samples. 25

26 islative term. As the proportion of African-Americans increases, the inverse becomes true. For example, when the proportion of African-Americans is set to 20 percent (or the first quartile), black MCs only deliver 0.46 more speeches. When the proportion of African-Americans is increased to 60 percent (or the third quartile), white MCs actually deliver 0.23 more symbolic speeches for every civil rights speech they deliver in a given legislative term. This result is particularly interesting for two reasons. First, when white MCs represent majority-minority districts, they do not seem to respond by speaking to issues like civil rights. Rather, they respond by using symbolism. Indeed, when the proportion of African-Americans increases from the minimum to the maximum, white MCs are 9.30 and 3.01 times more likely to use symbolism and talk about civil rights, respectively. Undoubtedly, the models used for these estimates are fundamentally different, but our results suggest when white MCs represent African-Americans on the House floor they do so using symbols, not issues. Second, regardless of whether one talks about civil rights or symbolism, white MCs are more responsive to changes in the proportion of African-Americans. In terms of the former, black MCs are less likely to talk about civil rights as the number of African-Americans increases. In terms of the latter, the inverse is true. For us, the absolute change is more interesting than the direction. Across both sets of models, white MCs change on average percent as the proportion of African-Americans is allowed to vary from the minimum to maximum. Conversely, black MCs only change on average percent, suggesting black MCs likely talk about civil rights and use symbolism regardless of the proportion of African-Americans in the district. These findings suggest that white MCs may have different incentives for talking about issues like civil rights and using symbols like Rosa Parks and Martin Luther King. One possibility centers on re-election. If African-American voters reward white MCs when they talk about issues like civil rights, then it makes sense for white MCs to respond when their district characteristics change. Similarly, if African-Americans view white MCs more favorably when they use symbols like Rosa Parks and Martin Luther King, then it should not be too surprising when white MCs use more symbols when they represent districts with a large number of 26

27 African-Americans. We also find it interesting that African-American legislators seem unresponsive to changes in the racial composition of their district. There are three plausible explanations for this finding. First, the vast majority of black MCs represent majority-minority districts. On average, their districts are 49.3% African-American. As a result, there is much less meaningful variation in the types of districts represented by black MCs, and so it is difficult to reach strong conclusions about how black MCs would respond with smaller black populations in their district. Second, the handful of black Republicans in our sample all hail from majority-white districts. There might be reason to expect these legislators to be more active in talking about civil rights issues in a symbolic way. Since floor time is partially constrained by party leadership, when civil rights issues need to be discussed by the GOP, leadership might be more likely to give the floor to black Republicans, making it easier for them to speak on these issues. And to the extent that they seek to find common ground with other African-Americans, they should generally use the same symbolic language as black Democrats. Third, it is possible that the use of meaningful rhetoric on civil rights is constrained less by desire to provide symbolic representation for African-American constituents, and more by ability. Mansbridge (1999) suggests most black representatives should have an advantage when it comes to using symbols related to the Civil Rights Movement because they have shared experiences and a deeper knowledge of the struggle for civil rights in the United States. As a result, even when districts are majority-white, black MCs are still aware of the key symbols of black struggle and can make use of them freely. White MCs may require substantial interaction with black constituents in their district before they become confident enough to use these symbols, giving white MCs who represent a large number of African-Americans a distinct advantage. In sum, our findings indicate that black and white members of the U.S. House of Representatives differ in both the amount of time they spend discussing an issue of importance to the black community civil rights and also in their use of symbols of the civil rights struggle. That is, black MCs are more likely to speak on civil rights and, on average, more likely to 27

28 invoke symbolism. Though white MCs speak less on civil rights and invoke symbolism less often, they are responsive to demographic changes in their district. In fact, white representatives with very high black populations in their district tend to invoke symbolism more than black representatives from similar districts. We turn now to investigating whether these racial differences in rhetoric on civil rights have meaning for constituents. Does invoking symbols of the civil rights struggle provide better symbolic representation for black constituents? If so, is there backlash against such rhetoric from white constituents? Answering these questions is crucial for our understanding of the consequences of racial rhetoric in Congress. 7 How Do Constituents Respond to Legislative Speech? C-SPAN gives MCs the ability to be beamed live into the homes of millions of Americans (Maltzman and Sigelman, 1996, 820). Not only are C-SPAN viewers politically attentive, but almost all vote. Appearing on news broadcasts or going viral on Twitter certainly carry their own benefits, but C-SPAN gives members open access to a large audience of interested viewers (Maltzman and Sigelman, 1996, 820). Indeed, scholars have advanced this argument so much that it has almost become synonymous with the study of floor speeches (e.g., see Morris, 2001; Osborn and Mendez, 2010; Pearson and Dancey, 2011a). Using images altered to look like C-SPAN videos, our experiment attempts to simulate the effect of watching black and white MCs deliver floor speeches on C-SPAN. Respondents were recruited on Amazon s Mechanical Turk (MTurk) service. We fully acknowledge this is neither a nationally representative sample, nor are MTurk respondents tuning into C-SPAN on a regular basis. With that said, past research has demonstrated MTurk to be a relatively reliable source of respondents (Berinsky, Huber and Lenz, 2012). We use the answers from these respondents to supplement, rather than replace our previous results, especially with regard to white representatives. The proportion of African-Americans in the district seems to influence the likelihood that white MCs speak to civil rights and use symbolism when doing so, suggest- 28

29 ing they have a strong electoral incentive to symbolically represent African-Americans. Our experiment is meant to give us additional causal leverage when making this claim. Subjects viewed a still image purportedly from a floor speech broadcast on C-SPAN. Below that image was a transcript from a section of the speech. We embedded the transcript in a video player, giving the treatments the appearance of being simple screenshots from the C-SPAN Video Library. 9 We then randomly assigned subjects to one of two types of floor speeches. The first floor speech discussed civil rights using symbolic language. This speech was originally delivered by John Conyers (D-MI) and invokes the symbol of Rosa Parks to call for more action on civil rights and equality. The second floor speech also discussed civil rights, but used non-symbolic language. It was originally delivered by John Yarmouth (D-KY), and lauded 50 years of work by his state s human rights commission on the issue of civil rights. Subjects were also randomly assigned either a white or a black representative (see Figure 6). A total of 223 subjects received the black representative/symbolic speech treatment, 210 received the white representative/symbolic speech treatment, 235 received the black representative/non-symbolic speech treatment, and 238 received the white representative/non-symbolic speech treatment. All subjects were from the United States. In the Supplement Information, we provide more details about the experiment, including how we selected these speeches. Figure 7 shows the effect of using a symbolic versus a non-symbolic speech on feeling thermometer ratings of the legislator. The results are particularly interesting for 106 black respondents who participated in our experiment. When their legislator shares their racial background, the type of rhetoric used seems to have no effect on evaluations. But when the legislator is white, the use of symbolic language has a large and statistically significant effect on their feeling thermometer ratings. White representatives giving non-symbolic speeches on civil rights are evaluated more negatively than black legislators. But when that white legislator invokes the types of symbols common in black rhetoric on civil rights, they are actually evaluated more positively than black legislators. This strongly suggests that voice matters. Our results demonstrate that were white legis

30 Figure 6: Examples of Symbolic and Non-Symbolic Experimental Treatments Note: These are examples of our experimental treatments. On the left, we include an image of a white representative we obtained from the New York State Senate. He is delivering a symbolic civil rights speech referencing Rosa Parks. On the right, we include an image of a black representative from the same legislative body. He is delivering a non-symbolic speech referencing the creation of the Commission on Human Rights. All four treatments can be found in the Supplemental Information. lators able to speak on civil rights in the way that black legislators do, they would be able to produce the symbolic representation that we have traditionally ascribed to racial resemblance. To the extent that legislators are able to do this to demonstrate that they get it we might have little concern about the levels of descriptive representation. 10 But remember that our speeches come from the content of actual floor speeches on civil rights. Even if it is hypothetically possible for white legislators to address black constituents concerns as vociferously as black legislators, they appear not to do so in reality. Another interesting finding from this experiment is that black legislators seem to receive the same evaluations regardless of if they give symbolic or non-symbolic speeches. This might suggest that resemblance is sufficient (but not necessary) for creating feelings of warmth and symbolic representation. This would reinforce the importance of descriptive representation, 10 Indeed, white representatives who represent majority-black districts do appear to speak and behave differently from other white representatives. One need look no further than Steve Cohen s (D-TN) twitter feed to see this in action. 30

31 Figure 7: Black Respondents Evaluate White Representatives More Positively When They Use Symbolism in Civil Rights Speeches 0.8 Black White Feeling Thermometer Rating Black Representative White Representative 0.4 Non Symbolic Symbolic Non Symbolic Symbolic Note: In the left panel, we report the feeling thermometer rating for the black (top row) and white (bottom row) representative, subsetting our data to only include black respondents (N = 106). We obtained the feeling thermometers using standard ANES questions. Specifically, we asked Please rate the representative you just saw, with a 100 representing warm and favorable and 0 implying cold and not favorable. They could choose any whole number between 0 and 100, with the default option being 50. To make the coefficients easier to interpret, we re-scaled the feeling thermometers to range between 0 and 1. In the right panel, we report the same feeling thermometers for white respondents (N = 1101). Within each quadrant, the left tick on the x-axis (labeled Non-Symbolic ) is the rating when the indicated representative is delivering a non-symbolic speech. The text of this speech talked about the formation of a Commission on Human Rights and can be found under the image of the black representative in Figure 6 (see right panel). The right tick on the x-axis (labeled Symbolic ) is the rating when the same representative is delivering a symbolic speech. The text of this speech talked about Rosa Parks and can be found under the image of the white representative in Figure 6 (see left panel). Vertical lines are 95 percent confidence intervals. 31

32 but also be susceptible to the same concerns Bianco (1994) had about legislators garnering trust absent action. A few caveats are in order. First, due to the limited number of African Americans in our sample, it is very challenging to identify statistically significant effects for black respondents. The fact that results do emerge despite this limitation give us confidence that our findings are both substantively meaningful and not just a statistical artifact. Second, our treatment did not allow for the introduction of a race-neutral control. We contemplated including a treatment where respondents did not receive a photo, but we decided that we would be unable to differentiate the effect of being race-neutral from simply the presence/absence of a picture. As such, we cannot determine definitively if black subjects are giving black legislators some boost in their feeling thermometers over and above what we would expect from a race neutral condition. With those caveats, our experimental results remain clear. Giving a speech that is typical of the symbolism used by black legislators appears to significantly boost the evaluations of white legislators. This suggests that a large part of why descriptive representation matters is due to voice, over and above the effects of racial resemblance. The experimental results also give us more confidence in the conclusions we reached using our observational data. It seems as though white MCs have strong electoral incentives to symbolically represent African- Americans in their district, especially when they represent a large proportion of their district. 8 Discussion Do black and white MCs talk about civil rights differently? Using the most comprehensive collection of civil rights speeches to date, we can answer with a definitive yes. Not only do black MCs make more symbolic references, but they are less influenced by changing electoral circumstances. When it comes to floor speeches, scholars generally suggest most are used to achieve electoral ends. We are not so sure this applies to black MCs and the discussion of civil rights. We show across several models their behavior is more consistent with an intrinsic mo- 32

33 tivation to advance the interests of the black community. Black MCs are over twice as likely to reference symbols like Martin Luther King and Rosa Parks when discussing civil rights. Moreover, these references are essentially unaffected as the proportion of African-Americans changes in their district. Both findings suggest black MCs are intrinsically motivated to symbolically represent African-Americans. Do these differences in the use of symbolic references matter? Again, we can answer with a definitive yes. Among black citizens, black legislators do not appear to benefit from the use of symbolic references. However, white legislators receive significantly warmer evaluations when they employ symbolism compared to when they give non-symbolic speeches. Moreover, the use of such symbolism does not appear to harm legislators of either race among either black or white citizens. Given that both black and white representatives can elicit more positive responses from black constituents by invoking symbolism, two lingering questions remain. First, why do white MCs not invoke such symbolism more regularly? And second, is the invocation of symbols of the African-American struggle for civil rights simply cheap talk that allows representatives to make claims that they are standing for the black community s interests while wholly ignoring their concerns when it comes to substantive representation? On the first question, we suspect that white representatives invoke symbolism only when they are extrinsically motivated to do so. There are two results that support this conclusion. Generally speaking, symbolic speeches are more likely in election years. This suggests that symbols, such as Martin Luther King and Rosa Parks, contain some value to certain constituent groups, namely African-Americans. General theories of substantive representation would support such a conclusion. Indeed, both black and white MCs can speak to African-American issues, like civil rights, using symbolism that is important to the African-American community, such as Martin Luther King. However, this is not what we find in our data. Instead, we find that symbols are only used by white representatives when strong electoral incentives exist, namely a high proportion of African-Americans in their district. For us, this underlines the value of descriptive representation. Unlike white representatives, black representatives use 33

34 symbolism because it appears to be a key component of their conception of black struggle for civil rights. Leaders like Martin Luther King and events like the march on Washington are not simply words on a page, they are core components of how African-Americans conceptualize civil rights; indeed, such symbols may even be a core part of African-American representatives identity. Whether it is an election year or not, they use symbols when discussing civil rights. In districts with both a high and low proportion of African-Americans they use these same symbols. Regardless of the circumstance, their behavior is constant. Such consistency suggests an intrinsic motivation to represent the African-American community. Thus, instead of black representatives standing for their African-American constituents, they are a fundamental part of that community. The same can not be said for their white counterparts. On the second question, we offer a more sanguine interpretation for four reasons. First, although symbolic speech may not directly relate to substantive representation, we believe it is important in and of itself. Any improvement in traditionally underrepresented groups satisfaction with the quality of their representation should be seen as normatively worthwhile, even if it is divorced from policy outcomes. Second, to the extent that African-Americans are aware that they are largely a captured group in the Democratic Party (Frymer, 1999), there is simply no way to expect black MCs to enact substantive policies favored by black constituents to the extent that they do not align with the interests of white Democrats. Third, it is not clear that voters would reliably be able to attribute policy success to their MC. As Grimmer, Westwood and Messing (2014) find, MCs are particularly skilled at claiming credit for legislative outcomes, regardless of whether they significantly influenced the process. As such, speaking on issues in ways that resonate with black constituents provides a clear and unambiguous (even if less meaningful) signal of a representative s commitment than credit claiming for substantive outcomes. Fourth, we believe that black MCs rhetoric falls in line with broader patterns of black legislators working hard to provide representation outside the confines of national-level policy-making. Our study adds to findings such as those by Grose (2011) that suggest that black legislators are more dedicated to both service and allocation responsiveness. Similarly, we argue such responsiveness should be taken as a source of optimism, regardless of whether 34

35 some may view speech-making as being somewhat cheap. Representatives speak to issues that are important to African-Americans for a number of reasons. Some are intrinsically motivated because they personally identify with the African- American community. For these representatives, speaking about civil rights follows naturally from a deeply held group identity which also influences the way they talk about high priority issues. For those who are extrinsically motivated, references to civil rights become more of a political currency that can be traded for constituent support. To these representatives, symbols of the civil rights movement, like Rosa Parks and Martin Luther King, are undoubtedly important, but they do not necessarily resonate with them personally. We show African-American voters do not seem to punish these representatives, ultimately emphasizing the importance of symbolic representation. Civil rights is not talked about nearly enough. Symbols like Rosa Parks and Martin Luther King are rarely invoked. In order to change the status quo, all representatives, regardless of race, need to speak to these concerns. It is not enough to simply talk about civil rights. Representatives need to do so in a way that demonstrates they get it. This does not necessarily mean one has to don a hoodie, but it does mean representatives need to speak in a way that connects with African-Americans. This is the only way to truly give voice to their concerns. 35

36 References Atkeson, Lonna Rae Not all cues are created equal: The conditional impact of female candidates on political engagement. Journal of Politics 65(4): Barnes, Tiffany D and Stephanie M Burchard Engendering Politics The Impact of Descriptive Representation on Women s Political Engagement in Sub-Saharan Africa. Comparative Political Studies p Berinsky, A.J., G.A. Huber and G.S. Lenz Evaluating Online Labor Markets for Experimental Research: Amazon.com s Mechanical Turk. Political Analysis 20(3): Bianco, William T Trust: Representatives and constituents. University of Michigan Press. Bobo, Lawrence and Franklin D. Gilliam Jr Race, sociopolitical participation, and black empowerment. The American Political Science Review pp Bratton, Kathleen A and Kerry L Haynie Agenda setting and legislative success in state legislatures: The effects of gender and race. The Journal of Politics 61(03): Broockman, David E Black Politicians are More Intrinsically Motivated to Advance Blacks Interests: A Field Experiment Manipulating Political Incentives. American Journal of Political Science 57(3): Bullock, Charles S., III The Impact of Changing the Racial Composition of Congressional Districts on Legislators Roll Call Behavior. American Politics Quarterly 23(2): Cameron, Charles, David Epstein and Sharyn O Halloran Do Majority-Minority Districts Maximize Substantive Black Representation in Congress? American Political Science Review pp Fleisher, Richard Explaining the Change in Roll-Call Voting Behavior of Southern Democrats. The Journal of Politics 55(2): Frymer, Paul Uneasy Alliances: Race and Party in America. Princeton, NJ: Princeton University Press. Gay, Claudine The effect of black congressional representation on political participation. American Political Science Review pp

37 Gay, Claudine Spirals of trust? The effect of descriptive representation on the relationship between citizens and their government. American Journal of Political Science pp Gerrity, Jessica C, Tracy Osborn and Jeanette Morehouse Mendez Women and Representation: A Different View of the District? Politics & Gender 3(02): Gillion, Daniel Q Governing with Words: The Political Dialogue on Race, Public Policy, and Inequality in America. Cambridge University Press. Griffin, J.D. and P. Flavin How Citizens and Their Legislators Prioritize Spheres of Representation. Political Research Quarterly 64(3): Grimmer, Justin, Sean J Westwood and Solomon Messing The impression of influence: legislator communication, representation, and democratic accountability. Princeton University Press. Grose, Christian R Congress in Black and White. Cambridge: Cambridge University Press. Hansen, Susan B Talking about politics: Gender and contextual effects on political proselytizing. The Journal of Politics 59(01): Harden, Jeffrey J Multidimensional Democracy: A Supply and Demand Theory of Representation in American Legislatures. Cambridge University Press. Hayes, Matthew Race and the Demand for Legislative Responsiveness. In Annual Meeting of the Southern Political Science Association. Hayes, Matthew and Matthew V Hibbing The Symbolic Benefits of Descriptive and Substantive Representation. Political Behavior pp Hill, Kim Quaile and Patricia A Hurley Symbolic speeches in the US Senate and their representational implications. The Journal of Politics 64(01): Hutchings, Vincent L Issue Salience and Support for Civil Rights Legislation Among Southern Democrats. Legislative Studies Quarterly 4: Hutchings, Vincent L., Harwood K. McClerking and Guy-Uriel Charles Congressional Representation of Black Interests: Recognizing the Importance of Stability. The Journal of Politics 66(2):

38 Hutchings, Vincent L. and Nicholas A. Valentino The Centrality of Race in American Politics. Annual Review of Political Science 7: Kathlene, Lyn Alternative views of crime: Legislative policymaking in gendered terms. the Journal of Politics 57(03): Lawless, Jennifer L Politics of presence? Congresswomen and symbolic representation. Political research quarterly 57(1): Levy, Dena, Charles Tien and Rachelle Aved Do differences matter? Women members of Congress and the Hyde Amendment. Women & Politics 23(1-2): Maltzman, Forrest and Lee Sigelman The politics of talk: Unconstrained floor time in the US House of Representatives. The Journal of Politics 58(03): Mansbridge, Jane Should blacks represent blacks and women represent women? A contingent yes. The Journal of politics 61(03): Miller, W.E. and D.E. Stokes Constituency influence in Congress. The American Political Science Review 57(1): Minta, M.D. and V. Sinclair-Chapman Diversity in Political Institutions and Congressional Responsiveness to Minority Interests. Political Research Quarterly. Morris, Jonathan S Reexamining the politics of talk: Partisan rhetoric in the 104th House. Legislative Studies Quarterly pp Neuendorf, Kimberly A The content analysis guidebook. Sage. Osborn, Tracy and Jeanette Morehouse Mendez Speaking as women: Women and floor speeches in the Senate. Journal of Women, Politics & Policy 31(1):1 21. Overby, L. Marvin and Kenneth M. Cosgrove Unintended Consequences? Racial Redistricting and the Representation of Minority Interests. The Journal of Politics 58(2): Pearson, Kathryn and Logan Dancey. 2011a. Elevating Women s Voices in Congress Speech Participation in the House of Representatives. Political Research Quarterly 64(4): Pearson, Kathryn and Logan Dancey. 2011b. Speaking for the Underrepresented in the House of Representatives: Voicing Women s Interests in a Partisan Era. Politics & Gender 7(04):

39 Pitkin, Hanna F The concept of representation. Univ of California Pr. Rocca, Michael S and Gabriel R Sanchez The effect of race and ethnicity on bill sponsorship and cosponsorship in congress. American Politics Research. Schickler, Eric, Kathryn Pearson and Brian D. Feinstein Congressional Parties and Civil Rights Politics from 1933 to The Journal of Politics 72(3): Schwindt-Bayer, Leslie A. and William Mishler representation. Journal of Politics 67(2): An integrated model of women s Swers, M.L Connecting descriptive and substantive representation: An analysis of sex differences in cosponsorship activity. Legislative Studies Quarterly 30(3): Tate, Katherine Black faces in the mirror: African Americans and their representatives in the US Congress. Princeton University Press. Whitby, Kenny J Bill sponsorship and intraracial voting among African American representatives. American Politics Research 30(1):

40 Supplemental Information This Supplemental Information includes additional analyses for the following article: Bryce J. Dietrich, Hayes, Matthew, Saumil Dharia, and Stella-Leonie Wancke. Racial Rhetoric in Black and White: Symbolic and Substantive References in U.S. House Speeches. 40

41 S1 Coding Symbolism In order to code the symbolic references in each civil rights speech we created a survey using Google Forms. The survey would display each of the following questions one at a time: 1. Who is coding this speech? [Coder 1/Coder2] 2. What is the file name? [Fill in Blank] 3. Is this speech at least 150 words? [Yes/No] 4. Do you see two symbols from the above list separated by at least one sentence? [Yes/No] Immediately below the final question, we included the following statement: The legacy of Dr. Martin Luther King (*symbol*) can be seen in this legislation. This bill does a lot of really good things for our country. Like the Voting Rights Act (*symbol*), this legislation advances civil rights. Coders were also encouraged to reference hyperlinked examples we provided as many times as they needed. We read several batches of sample speeches to come up with our list of symbols. This process involved everyone reading a sample of 100 speeches. Once everyone was done, we had a group meeting where each team member would recommend a symbol be either included or excluded from the final list. Ultimately, we settled on a final list of four main categories and several subcategories that fell under each: People Martin Luther King Coretta Scott King Rosa Parks Thurgood Marshall Caesar Chavez Medgar Evers 41

42 Malcolm X Abraham Lincoln Ghandi Events Birmingham sit in Selma sit in Selma March March on Washington Freedom Rides Policy Acts Voting Rights Act Civil Rights Act Brown v. Board of Education Plessy v Ferguson Jim Crow Concepts (at least 5 percent) Racism Segregation Lynching Slavery Equality Diversity Freedom For example, under the People category there are nine subcategories, meaning references to Martin Luther King could include a variety of things, such as MLK, Dr. King, etc. This 42

43 gave the coders guidance while giving them flexibility to interpret references to Martin Luther King we did not foresee in our group meetings. Under this category, there was considerable debate about whether to include Caesar Chavez and Ghandi. These references were made in a handful of speeches and they were always in conjunction with another symbolic reference which is why they were included in our final list. When they are excluded, they have essentially no effect on our substantive results. For the Concepts category, we asked the coders to visually inspect each document. For example, when they saw references to racism, they would do a search for other references in the document. This would highlight several terms they thought were relevant. They would then assess whether those terms represented more than 5 percent a number we settled on after a couple of training batches. Here is an example of one of the highlighted documents: Figure S1: An Example of the Visual Coding Scheme We Used To Code Symbolic Concepts The visual coding scheme was actually developed by our coders. We liked their idea because it limited the number of false positives. By looking at the distribution of each concept, our coders ensured the symbolic references were widely distributed across the entire speech, rather than simply being made in passing. This ultimately produced results that were very similar to what we were expecting when we created the Concepts category. To learn more about our coding scheme, please download the speeches we used in this study from the following URL: 43

44 The speeches are posted on one of the author s website. URL removed for blind review. Once downloaded, use the following to pair the text of each speech with our coding scheme: CSV file is posted on one of the author s website. URL removed for blind review. If there are any irregularities, please let us know. No coding scheme is perfect, but the one we developed for this study seemed to reliably differentiate between symbolic and nonsymbolic civil rights speeches. Indeed, Figure 1 shows we obtained a 97 percent agreement rate, meaning we could reliably code speeches using the scheme we developed. S2 Creating Our Experimental Treatments To find the speeches for our experimental treatments we identified the archetypal civil rights speech for black and white MCs. This was done by comparing all civil rights speeches delivered by black MCs to one another using the Jaccard distance. We did the same for white MCs. Prior to calculating the Jaccarding distance, we did the following pre-processing steps: 1. Converted each speech to lowercase letters. 2. Removed all punctuation. 3. Removed all numbers. 4. Removed all English stop words using the dictionary provided in the tm library. For white MCs, seventeen speeches ended up having the minimum Jaccarding distance. To break this tie, we used a more refined measure: the Levenshtein distance. With this measure, we re-compared those 17 speeches to every other civil rights speech delivered by white MCs. Once this was done, a single speech had the minimum Levenshtein distance. The speech was delivered by John Yarmouth (D-KY) in 2010: 44

45 Madam Speaker, I rise to pay tribute to the Kentucky Commission on Human Rights on the occasion of the organization s fiftieth year. In 1960, while the civil rights debate was raging across the country over the extension of Civil Rights to all Americans, Kentucky became the first state south of the Mason-Dixon line to establish a human rights commission. Since that time, the pioneering members of the Commission have been an ally of all Kentuckians fighting for equality, fairness, and the equal application of justice. The Kentucky Commission on Human Rights is charged with enforcing State and Federal Civil Rights law. It is also tasked with a powerful mandate, remarkable in its scope and comprehensive in its aims; To safeguard all individuals within the state from discrimination because of familial status, race, color, religion, national origin, sex, age 40 and over, or because of the person s status as a qualified individual with a disability, reads the mandate. Thereby to protect their interest in personal dignity and freedom from humiliation, to make available to the state their full productive capacities, to secure the state against domestic strife and unrest which would menace its democratic institutions, to preserve the public safety, health, and general welfare, and to further the interest, rights, and privileges of individuals within the state. Those who have worked with the Commission throughout the last five decades have committed themselves to meeting the goals of that mandate. And that dedication is reflected not just in the law books of the Commonwealth, but in communities throughout Kentucky. It was Kentucky that was the first southern state to pass a civil rights act and the first in the Nation to enact a fair housing law. And each and every day, the Kentucky Commission on Human Rights continues to fight to educate our citizens about how far we have come and how far we still have to go. Every citizen of our Commonwealth can be proud of the legacy of the Kentucky Commission on Human Rights. I urge my colleagues to join me in congratulating the Commission for fifty years of dedication to the expansion of equal rights across our Commonwealth and honor their vision of a people united against discrimination and united for equality. 45

46 The same process was used for black MCs. Using the Jaccard distance, we identified the 72 most similar speeches. Once these speeches were isolated, we compared each to the rest of the speeches delivered by black MCs using the Levenshtein distance. Ultimately, this gave us the following speech from John Conyers (D-MI) delivered in 2005: Mr. Speaker, today I rise to introduce a resolution commemorating Rosa Parks and her decision to stand up to injustice 50 years ago. On December 1, 1955, Ms. Parks refused to obey a bus driver s order that she give up her bus seat in the black section to a white man because the white section was full. It is the courage, dignity, and determination that Ms. Parks exemplified on this day that allows most historians to credit her with beginning the modern day civil rights movement. Ms. Parks s actions on December 1, 1955 led to the desegregation of American society and enabled all of this nation s citizens to realize freedom and equality. The arrest of Ms. Parks led African-Americans and sympathizers of other races to boycott the Montgomery city bus line until the buses in Montgomery were desegregated. The 381 day Montgomery Bus Boycott encouraged other courageous people across the United States to organize in protest and demand equal rights for all. The fearless acts of civil disobedience displayed by Rosa Parks and others resulted in the United States Supreme Court, on November 13, 1956, affirming a district court decision that held that Montgomery segregation codes deny and deprive African- Americans of the equal protection of the laws. This decision would lead to other landmark Supreme Court decisions in which the Court would rule in the interest of justice and equality. In the years following the Montgomery Bus Boycott, Ms. Parks moved to Detroit, Michigan in 1957 and continued her civil rights work by working in my District Office. Ms. Parks was with the office from 1965 until In the more than twenty years that Ms. Parks was in the office, she worked with a tireless spirit for the people of Detroit and other Americans. In 1987, she started the Rosa and Raymond Parks Institute for Self Development in Detroit, a nonprofit 46

47 organization which motivates youth to reach their highest potential. So it is with great pleasure and honor that I stand today to recognize not only a civil rights pioneer, but a member of my staff, a constituent, and a friend. It is in this recognition of the 50th Anniversary of Ms. Parks s refusal to give up her seat on the bus, that I ask the Congress and the great people of this nation to work with the same courage, dignity, and determination exemplified by her to address modern day inequalities and injustice. The American dream is out of reach for millions. The United States is one of the richest nations in the world so I ask: Why aren t our schools thriving? Why are so many African-Americans unemployed? Why are health care, housing, and college education so difficult to afford? Civil rights pioneer Rosa Parks displayed a defiant act of courage 50 years ago so that these questions would not be raised today. I know that this Congress and the people of this nation can work to further the ideals of Ms. Parks and the Civil Rights Movement. These archetypal speeches are representative of the language black and white MCs used when discussing civil rights. If one were to randomly sample all the civil rights speeches delivered by white MCs, then one would likely find a speech similar to John Yarmouth s speech about the Kentucky Commission on Human Rights. Similarly, if one were to randomly sample all the civil rights speeches delivered by black MCS, then one would likely find a speech similar to John Conyers speech about Rosa Parks. We ultimately used these archetypal speeches for the text of our experimental treatments precisely because of their statistical similarity to the rest of the civil rights speeches in our data. Using a simple 2 2 design, we randomly assigned respondents to either a black or white representative delivering a civil rights speech which used either symbolic or non-symbolic language. The Rosa Parks speech was our symbolic treatment. The Commission on Human Rights speeches was our non-symbolic treatment. Even though it is readily apparent the former uses symbolism, whereas the latter does not, these speeches were actually independently coded this way. The final treatments can be found below. 47

48 Figure S2: Black Representative Giving Non-Symbolic Civil Rights Speech Note: The image of the representative was obtained from a video of a floor session from the New York State Senate. The text is an anonymized and shortened version of the white archetypal speech delivered by John Yarmouth (D-KY) in This speech was independently coded as being non-symbolic. Please refer to Section S1 for more details about our coding scheme. 48

49 Figure S3: Black Representative Giving Symbolic Civil Rights Speech Note: The image of the representative was obtained from a video of a floor session from the New York State Senate. The text is an anonymized and shortened version of the black archetypal speech delivered by John Conyers (D-MI) in This speech was independently coded as being symbolic. Please refer to Section S1 for more details about our coding scheme. 49

50 Figure S4: White Representative Giving Non-Symbolic Civil Rights Speech Note: The image of the representative was obtained from a video of a floor session from the New York State Senate. The text is an anonymized and shortened version of the white archetypal speech delivered by John Yarmouth (D-KY) in This speech was independently coded as being non-symbolic. Please refer to Section S1 for more details about our coding scheme. 50