As Good as it Gets?
Public Approval of Congress in a Partisan Climate
David C. Kimball
University of Missouri-St. Louis
November 2001
Abstract
Congress has been the scene of increasingly partisan and ideologically polarized conflict in recent years. I examine the extent to which the national political climate mutes or amplifies the effect of partisanship on evaluations of Congress. Consistent with a priming hypothesis, public evaluations of Congress divide most sharply along party lines when elite-level discourse is most partisan (as during an election season or a highly charged partisan debate in Congress). In addition, the most knowledgeable citizens are most likely to be primed by the partisan political climate in Washington. In contrast, less attentive citizens tend to rely on non-partisan cues when evaluating Congress. I discuss the implications of these findings for public opinion and improving the public standing of Congress in an increasingly partisan climate.
1
Congress has been the focus of increasingly partisan and ideologically polarized conflict in the last two decades (Rohde 1991; Cox and McCubbins 1993; Poole and Rosenthal 1997; Bond and Fleisher 2000).[1] How does public opinion respond to a more partisan political environment? The scholarly consensus is that the public largely disapproves of party conflict and partisan rhetoric in Congress (Hibbing and Theiss-Morse 1995; Mayer and Canon 1999; Davidson et al. 1966; Cooper 1999).
I examine another way in which the national political environment can shape public evaluations of Congress. The political environment is a source of cues, making particular considerations more or less salient when citizens rate congressional performance. National politics is sometimes characterized by high levels of partisan conflict (as during a presidential election, a major policy debate between parties, or Senate consideration of a controversial judicial nomination) and sometimes it is the scene of low levels of party conflict (as during a war or a major policy debate that cuts across party lines).
I argue that increased partisan conflict in national politics primes citizens to alter the ingredients that form their evaluations of government, causing them to rely more on party attachments and less on other factors. By portraying higher or lower levels of partisan conflict, elite discourse and national events can mute or amplify the effects of partisanship on evaluations of Congress. Finally, while partisan cues are often readily available to citizens who lack other information, I find that people with high levels of political knowledge are most likely to respond to the partisan national climate in this fashion.
Public Evaluations of Congress
Congress is often regarded as “the people’s branch” of government, and public approval of Congress is often viewed as an important indicator of the legitimacy of the American governmental system (Easton 1965; Dennis 1981; Weatherford 1992). As it turns out, public evaluations of Congress, as measured by the standard job approval question, typically are not very positive (see Figure 1). Recent approval levels approaching 50% are quite unusual when compared to other public opinion readings taken over the last thirty years, and the typically low public ratings of Congress are often interpreted as a sign of trouble for American democracy (Cooper 1999; Mann and Ornstein 1992). However, to the extent that public evaluations of Congress reflect party attachments that can be primed by the political climate in Congress, then the conventional public approval measure may not be a good indicator of the institution’s legitimacy.
Nevertheless, public evaluations of Congress do have more practical consequences. An unpopular Congress may harm the reelection chances of incumbents (Born 1990), especially those of the majority party (Box-Steffensmeier et al. 2000). Furthermore, low public regard for Congress is cited as a reason why potential candidates decide against running for a seat in Congress (Theriault 1998; Fowler and McClure 1989). This has stimulated a lot of research on the ingredients of public evaluations of Congress (for recent examples, see Hibbing and Theiss-Morse 1995; Asher and Barr 1994; Cooper 1999; Mayer and Canon 1999). I take a closer look at one of these ingredients – partisanship.
[Figure 1 here]
It is no surprise that there is a partisan element in public evaluations of Congress and other legislative bodies (Patterson, Boynton, and Hedlund 1969; Patterson, Ripley and Quinlan 1992; Hibbing and Theiss-Morse 1995; Asher and Barr 1994; Kimball and Patterson 1997). Party identifiers are more approving when their party controls Congress and more disapproving when the opposition controls Congress. In addition, there has been a longstanding debate regarding the sources and consequences of citizen trust and support for government. Miller (1974) argues that public distrust is a result of fundamental discontent with government policies, while Citrin (1974) argues that trust in government is colored by more partisan attitudes, such as one's attachment to the president or the governing party. While there is evidence supporting both arguments (for example, see Lipset and Schneider 1983), this paper examines a parallel form of Citrin's argument in exploring partisan-based evaluations of Congress.
I use a “priming” hypothesis to examine whether the inclination for citizens to assess national institutions in terms of their attachment to the governing party should ebb and flow in response to the phases of the electoral cycle: partisan-based evaluations of institutions should be strongest during an electoral campaign, when partisan cues are most conspicuous (Citrin 1974, p. 974). I test two hypotheses in applying this theory to congressional approval. First, only the most knowledgeable citizens will see Congress through partisan-colored lenses. In order to evaluate Congress in partisan terms, one must know, at a minimum, which party controls Congress. Therefore, less knowledgeable citizens must rely on other information (such as presidential performance and the nation’s economy) in a non-partisan fashion to evaluate Congress. Second, by portraying higher or lower levels of partisan conflict, elite discourse and national events can mute or amplify the effect of partisanship on evaluations of Congress. More generally, this suggests that the political environment can prime citizens to alter the ingredients that form their evaluations of the government.
I use survey data collected by the National Election Studies from 1990 through 1998 and content analysis of newspaper coverage of national politics to examine the extent to which national politics can prime partisanship in assessments of Congress. The following section lays out the theoretical development of these hypotheses in more detail. I then provide evidence to test both hypotheses, which receive some empirical support. Finally, I conclude by discussing the implications of these findings. In particular, this paper sheds some light on how fixed (and inherently negative) opinions of Congress really are. To the extent that partisan evaluations of Congress can be primed, it provides some reassurance that public disapproval of Congress is based on good, old-fashioned party competition rather than simply anti-government or anti-democratic beliefs.
Theory
Several studies have examined how public opinion responds to elite-level discourse. For example, Key's (1966) description of the public as an "echo chamber," mirroring what goes on at the elite level, is one commonly used metaphor. This thread can be found in the debate over whether levels of issue voting and party polarization in public opinion increased in the 1960s as new issues entered the political marketplace (for example, Nie, Verba and Petrocik 1976; Kessel 1974; Pomper 1972; Sullivan et al., 1979). Similarly, public opinion has bifurcated on racial issues (with Democrats becoming more liberal and Republicans becoming more conservative) in response to elite positioning on these issues (Carmines and Stimson 1989). More recently, research on media priming effects (Iyengar and Kinder 1987; Krosnick and Brannon 1993) and attitude change (Zaller 1992) has examined ways in which the public responds to the world of politics.
Several of elements of the national political environment shape public approval of Congress. For example, scandals (Durr et al. 1997) and increasing levels of media coverage of Congress (Durr et al. 1997; Patterson and Caldeira 1990; Prinz 1994) tend to reduce public approval of Congress. Conflict between Congress and the president, measured by divided government, congressional support of the president's agenda (Patterson and Caldeira 1990), or veto activity (Durr et al. 1997), is accompanied by a drop in congressional approval. In contrast, "rally events" (such as a war) tend to boost public support for Congress (Parker 1977, 1981). Also, when economic conditions improve, approval of Congress and other government institutions rises (Durr et al. 1997; Lipset and Schneider 1983; Parker 1977; Patterson and Caldeira 1990). Some evidence suggests that presidential popularity influences aggregate-level support for Congress (Davidson et al. 1966; Patterson and Caldeira; Dennis 1981; Davidson and Parker 1972; but also see Parker 1977, 1981). Finally, major legislative activity tends to drive public approval of Congress downward (Durr et al. 1997), suggesting that the public may not be so fond of the lawmaking process and other congressional procedures (Hibbing and Theiss-Morse 1995).
The national political environment may also shape public opinion by making particular cues more or less salient when citizens evaluate Congress. This argument relies on a theory of priming, which holds that information receiving extensive media coverage comes to mind more readily when citizens are asked for their opinion on other related topics. For example, issues covered widely in the news have a stronger effect on presidential evaluations than issues receiving little or no coverage (Iyengar and Kinder 1987; Krosnick and Kinder 1990; Krosnick and Brannon 1993).
I apply the priming hypothesis to examine the effects of partisanship on congressional approval. There are several reasons for doing so. With recent increases in party voting in Congress (Rohde 1991; Cox and McCubbins 1993), and a more aggressive brand of partisanship practiced in national government in the 1990s (Bond and Fleisher 2000), we should be interested in whether and how the public is responding. In addition, the news media help convey the partisan nature of politics because competing parties often supply different points of view needed for an "objective" news story (Ansolabehere, Behr and Iyengar 1993).[2]
There is evidence of partisan priming in the voting literature. The presence or absence of party cues on the ballot influence whether voters cast a split ticket or even complete the ballot (Campbell et al. 1960; Burden and Kimball 1998; Walker 1966). In addition, the presence of party labels and images in campaigns can influence the extent to which voters rely on party stereotypes(Rahn 1993). Assessments of Congress offer another opportunity to test the priming hypothesis. To my knowledge, the priming hypothesis has not been applied to evaluations of Congress, for two likely reasons. First, there is little media coverage of Congress in comparison to other institutions, such as the presidency (Graber 1989). Second, there is little variation in the tone of media coverage accorded to Congress -- most of it is negative (Mann and Ornstein 1994). However, the media has little control over the level of partisan conflict in national politics. While the media theoretically could choose whether to frame a news story in partisan versus nonpartisan terms, partisan conflict is often the result of strategic decisions made by the parties and their leaders inside and outside of government. Thus, although media coverage of Congress may not change much over time, the level of party friction in national politics can vary a great deal.
I address another aspect of opinion priming by examining the moderating effect of political knowledge.[3] This will help identify those citizens who are most responsive to changes in the partisan tone of national politics. There is evidence that people with high levels of political knowledge are most vulnerable to priming effects because they are most efficient at incorporating new information into a summary judgement or political evaluation (Krosnick and Brannon 1993; Miller and Krosnick 2000; but see Iyendar and Kinder 1987). In addition, one's level of knowledge or exposure to political information determines the likelihood one will receive and accept information transmitted through the media (McGuire 1968; Zaller 1992). Furthermore, to the extent that people are exposed to political information in the media, they tend to emulate the arguments made by elites sharing their partisan persuasion. Thus, Zaller (1992, 1994) suggests that those with the highest levels of political knowledge will be most responsive to changes in the partisan character of national politics.[4]
There is a common sense reason why partisan evaluations of Congress are likely found only among the most knowledgeable citizens. For starters, evaluating Congress on partisan grounds requires knowledge about the party controlling the House and Senate. In addition, it requires a belief that party labels matter, which requires some knowledge about the policy positions and groups associated with each party.
If the most knowledgeable citizens respond to partisan cues when evaluating Congress, what about the less knowledgeable? It has been argued that less knowledgeable people fail to distinguish between different branches of government, even (perhaps paradoxically) when party control of government is divided (Delli Carpini and Keeter 1996). In the absence of other information about partisan politics, feelings about the president or the country's economic performance can serve as easy decision-making cues when evaluating Congress. Thus, for less knowledgeable citizens, assessments of Congress should be more closely linked to global evaluations of the president and the national economy.
There is plenty of evidence that support for Congress is linked to support for the president (Davidson et al. 1966; Davidson and Oleszek 1990; Durr et al. 1997; Kimball and Patterson 1997) and assessments of the national economy (Parker 1977; Patterson and Caldeira 1990; Mayer and Canon 1999; but see Patterson, Ripley and Quinlan 1992). I expect that these findings are driven by less knowledgeable citizens, and probably do not apply to the most knowledgeable segment of the public, who tend to view the world in more partisan terms and regularly make distinctions between different branches of government.
To summarize, I propose the following three hypotheses:
H1:Evaluations of Congress divide more sharply along party lines when national politics is characterized by relatively high levels of party conflict. Party attachments have less influence on evaluations of Congress when the national political climate exhibits low levels of party conflict.
H2:The most knowledgeable citizens are most likely to be primed by the partisan political climate when evaluating the performance of Congress.
H3:For the least knowledgeable citizens, party attachments will have little influence on evaluations of Congress, regardless of the level of party conflict in national politics. Instead, ratings of Congress will be more closely linked to global evaluations of the president and the national economy.
Data
The main source of data is the series of surveys conducted by the National Elections Studies from 1990 through 1998.[5] These data provide an excellent opportunity to test hypotheses about priming effects because they sample American opinion under a variety of conditions. Between 1990 and 1998, party control of the White House changed (after the 1992 election) and party control of Congress changed (after the 1994 election). In addition, some NES surveys conducted in even-numbered years took place under the relatively partisan conditions of national election campaigns, while in odd-numbered years surveys are collected during less partisan periods. For example, the 1991 NES survey took place just a few months after the Gulf War hostilities ended, and the 1993 survey was in the field during the NAFTA debate in Congress, an issue which cut across party lines.
In addition, I conducted a content analysis of headlines from front-page articles in the New York Times to get a better sense of the extent of elite-level partisan conflict during each period of this study. Each content analysis covered a two-and-a-half month period that roughly overlaps with the interview component of an NES survey. For each two-and-a-half month period, I counted the number of articles that referred to party conflict as a percentage of all stories that mentioned Congress or the president in the text. Thus, each article was coded as partisan or nonpartisan based on the information in the headline, and then the percentage of partisan news stories was calculated for each time period. The content analysis method and coding scheme is described in more detail in the appendix.
[Table 1 here]
The first row of Table 1 provides the results of the media content analysis. As expected, the even-numbered years (especially 1992) are characterized by a higher proportion of articles detailing partisan politics. As a validity check on the content analysis measure, Table 1 includes the correlation between the two party thermometer questions that are routinely included in NES surveys. On the assumption that the survey responses are a reflection of partisan discord in national politics, I expect that the correlation between the party thermometers (the second row of Table 1) will be stronger in periods when national politics is more partisan, as measured by the news articles. With the exception of 1990, the party thermometer correlations track the content analysis measure fairly closely.[6] In any case, the survey-based and media-based measures both indicate that, at least in the 1990s, national political discourse contained a stronger dose of partisanship in even-numbered years.