When to Attack

Abstract

While the study of the use of negative campaign tactics has been a prolific topic of scholarship, previous studies have been limited in their ability to test the dynamics of negativity over the course of the campaign because their analysis of content from political campaigns is either static in nature or limited in its scope. Although we know that the closeness of the race, the status of the candidate as an incumbent or a challenger, and other characteristics affect the volume of negativity over the entire campaign, we know little about how these factors affect when candidates choose to go negative. Using a unique data set of over 1,400 emails sent out by campaigns from a random sample of congressional districts in 2012, we examine negativity over the course of the campaign. We find that the dynamics of negativity vary by whether there is an incumbent in the race.

Keywords

congressional campaigns campaign negativity

In theory—if not necessarily in practice—Americans are overwhelmingly opposed to negative campaigning (Hartman, 2012), the presence of which may reduce participation in the democratic processes practiced in the United States (Doherty & Adler, 2014; Mattes & Redlawsk, 2015). Yet, negative campaigning has increased in recent years, both in the actual quantity (Fowler & Ridout, 2013) and in the perceptions of the general public (Marist Poll, 2012; Ridout & Fowler, 2010). What is less clear, however, is the role that negativity plays in candidates’ election strategies, specifically when in the electoral cycle campaigns choose to deploy it, and what electoral factors motivate that decision.

While a small number of studies have attempted to understand the timing behind candidates’ decisions to go negative, studies of campaign negativity in congressional races have relied on surveys of candidates, campaign staff, or consultants (Herrnson, 2012; Theilmann & Wilhite, 1998). The problem with using a measure based on recall is that it is prone to error (Goldstein & Freedman, 2002; Price & Zaller, 1993), and more importantly, like the other static measures, provides a limited view of the dynamics of campaign negativity. Likewise, measures based on stated intentions gathered from surveys about hypothetical situations may be fundamentally different from actual behavior (LaPiere, 1934; Schuman & Johnson, 1976; Schuman & Presser, 1980; Stout & Kline, 2008).

This article does two things to further the understanding of negativity in congressional campaigns. First, we use a dynamic, rather than a static, measure of campaign rhetoric to understand the timing of negativity in campaigns. Second, this analysis examines the strategic decision of congressional candidates to go negative over the course of the campaign using actual campaign communications. Rather than examining a campaign’s strategic decisions through news mediated message statements or costly advertising which limits and biases the sample to only those campaigns in competitive elections or with sufficient funds, we examine the message choices of candidates using the email updates these candidates send to supporters and journalists over the course of the campaign. Using more than 1,400 emails from 200 randomly selected campaigns, we show that there are fundamental differences in the strategic decisions of incumbents, challengers, and open-seat candidates about when in the course of the campaign to go negative.

We find that the trends of negativity among open-seat congressional candidates are distinct from the trajectories of incumbents and challengers. Candidates for open seats follow the traditional narrative of beginning with positive messaging, followed by a more negative and contrasting tone, and concluding on a positive note. We argue that, in contrast to open-seat candidates, incumbents and challengers are dealing with a scenario in which one candidate and that candidate’s political record are well known to the public. As such, these candidates are more likely to start the campaign with higher levels of negativity and, as a result, less likely to vary the content of their message over the course of the campaign.

Who Goes Negative

Although there has been little investigation of the dynamics of campaign communication over the course of an election cycle, scholars have identified the types of campaigns that are more likely to be negative on the whole. For the most part, competitiveness seems to drive tone. More competitive House and Senate races are generally characterized by more negativity (Goldstein & Freedman, 2002; Lau & Pomper, 2004). In competitive races, candidates are more likely to attack their opponents in an attempt to draw clear distinctions between them and their opponent and to encourage voters to defect from their opponent’s camp (Lau & Pomper, 2001; Skaperdas & Grofman, 1995; Theilmann & Wilhite, 1998).

Previous findings also suggest that challengers are more likely to attack an opponent than are incumbents or even open-seat candidates. Because negative rhetoric encourages information-seeking behavior (Brader, 2006), those who stand to benefit from it most are those about whom little is known and who need voters to reconsider their choices (Kahn & Kenney, 1999). At the same time, the choice to engage in negative campaigning also involves additional risks (Fridkin & Kenney, 2011). Because incumbents are more risk averse, they are more likely to eschew negative rhetoric compared with challengers (Druckman, Kifer, & Parkin, 2009).

The Dynamics of Campaign Rhetoric

In spite of a voluminous literature on what drives negativity in congressional campaigns, scholars have largely ignored the timing of vitriol. We can find no in-depth accounting of how rhetoric changes over the course of the congressional campaign. Although research has asked campaign consultants about hypothetical situations in an attempt to test dynamic electoral strategy (Theilmann & Wilhite, 1998), this information has not been corroborated within the context of real campaigns. This lone study also did not include any other contextual campaign information besides relative poll standing.

Studies of congressional campaign rhetoric have almost entirely utilized aggregate data compiled over the entire course of the campaign. As such, these studies have only examined negativity using static measures of political communication rather than looking at the dynamics of negativity (Druckman et al., 2009; Flowers, Haynes, & Crespin, 2003; Lau & Pomper, 2001). Likewise, previous work that has modeled the message strategy choices of candidates has also relied on models and analyses that are static (Sigelman & Shiraev, 2002; Skaperdas & Grofman, 1995).

The limited information available about the dynamics of real world campaign rhetoric comes from a single presidential campaign. In their work, Goldstein and Freedman (2002) note that the 2000 presidential campaign became more negative over time and that the race ended with high levels of negativity in its television advertising. However, the study does not analyze congressional campaigns and bases its conclusions on a single presidential race with no sitting president running for re-election. All of this leaves campaign dynamics a virtually unexplored topic.

With limited scholarship, summaries of campaign tactics have been forced to rely almost entirely on anecdotes. Anecdotal evidence argues that congressional campaigns follow a set trajectory in which candidates begin with positive rhetoric aimed at introducing themselves to the general public. This is followed by a campaign of negativity aimed at lowering the favorability of the opposing candidate. As the campaign draws to a close, however, the conventional wisdom suggests that campaigns back away from negative rhetoric. These anecdotal accounts indicate that campaigns conclude on a positive note to help reduce the negative opinion of their campaign accumulated from weeks of negative campaigning and to encourage voters to better relate with the candidate (Jackson, 2010; Neff, 2002).

There is good reason to believe, however, that this anecdotal evidence is an overgeneralized distortion of campaigns based predominantly on a subset of campaigns that get attention in the media. Campaigns vary significantly. Previous work has shown that challengers, incumbents, and open-seat candidates differ in the amounts of risk that they are willing to take in a campaign (Druckman et al., 2009). Likewise, the amount of information available to the public about the political stances of candidates varies drastically (Jacobson, 2009). In races with an incumbent, the political battleground has already been well defined. In races without an incumbent, both candidates must introduce themselves to an audience relatively unfamiliar with the options presented to them. In addition, because there is a lack of information about the candidates in open-seat races, it is less likely that either candidate has a clear understanding of the opponent’s policy positions and the grounds on which to base issue attacks.

Hypotheses

The availability of information about candidates and their policy positions, or rather the lack thereof in open-seat races, leads us to our first hypothesis:

Hypothesis 1: Open-seat candidates will engage in less negative campaigning earlier in the campaign.

We argue that this is the case for two reasons. First, in open-seat races, both candidates are less familiar to the public, forcing them to spend more time introducing themselves. Second, a relative dearth of information about the policy positions of candidates in open-seat races and the lack of experience in congressional battles make it more difficult for candidates to pin down an opponent’s position that is attackable.

The lower the likelihood of clear established policy positions for either candidate means that both candidates will have less issue material to use in negative attacks in comparison with races where there is an incumbent. Because personal attacks, however, are not based on policy positions formed in Congress or on the campaign trail, there is, theoretically, no less information about the personal lives of candidates in open-seat races when compared with challengers and incumbents. This leads us to our second hypothesis:

Hypothesis 2: Open-seat candidates will engage in much less policy negativity but in roughly the same amount of personal negativity as candidates in races with an incumbent.

Without either candidate having an established political record, the battle lines around issues are not clearly drawn. However, the availability of information about a candidate’s personal characteristics is not as dependent on incumbency. Thus, the difference in negativity early in a campaign should be driven by a lack of policy negativity, not personal negativity.

Using Campaign Email Communication to Study Campaign Rhetorical Strategy

Before we can test these hypotheses, we must find data sufficient to permit the analysis of campaign rhetoric over time. A core problem in the study of campaign rhetoric is the collection of appropriate data. This becomes especially difficult when examining the dynamics of the campaign. Previous studies of campaign rhetoric have relied primarily on either express television advertising captured through the Campaign Media Analysis Group (CMAG) (Fowler & Ridout, 2013; Freedman & Goldstein, 1999; Goldstein & Freedman, 2002), or by analyzing news coverage of candidate statements (Lau & Pomper, 2001, 2004).¹

The problem with using advertising to measure campaign dynamics is that many House candidates do not produce television advertisements because of the high cost in some media markets, the inefficient media coverage of the congressional district, or the lack of sufficient funds to pay for advertising (Druckman et al., 2009). As such, using television (or even other forms of paid) advertising to measure campaign rhetoric oversamples campaigns in certain districts and presents a skewed picture of campaign messaging. Research drawing from advertising samples oversamples races with high funding and high amounts of news coverage.

Using media coverage of campaigns to analyze campaign rhetoric also suffers from the same problem of selection bias, as news coverage also disproportionately favors competitive campaigns (Druckman et al., 2009). In addition, media coverage may not necessarily reflect the rhetorical choices of the campaign, as the representation of the rhetoric of the campaign may be incomplete due to the mediated nature of campaign reporting.

The use of campaign email solves many of these problems. First, unlike news coverage of campaign rhetoric, email to supporters and interested parties is sent by campaign staff and, in many cases, by the candidate (Herrnson, 2012). Similar to campaign advertising, rhetoric contained in email is unmediated. Whereas news coverage of a campaign reflects the interpretation of the journalist writing the news story, which campaigns constantly try to massage and correct, the message contained in email communications comes directly from the campaign.

Second, email communication is more comprehensive than other dynamic measures of campaign rhetoric. While Druckman and his colleagues (2009) report approximately 16% of campaigns in competitive races had insufficient news coverage to gauge campaign rhetoric (previous studies have required more than 15 news articles to classify campaign rhetoric), that number is buoyed by the presence of Senate campaigns which newspapers are much more likely to follow. As displayed in Table 1, when the point of analysis is House races, the percentage of campaigns without sufficient news coverage to allow analysis grows to 63%. It is even worse for non-competitive campaigns, where 96% of candidates receive insufficient news coverage to classify campaign rhetoric. While emails still provide a limited picture, the number of campaigns without coverage is greatly reduced. Studying campaign rhetoric through email communication thus provides both a dynamic and comprehensive way to understand the message choices of campaigns.

Table 1.

Campaigns in Sample Using Email Communication.

Race type	% with emails	% with >15 news articles	% with TV ads
Toss-up	80.0	37.5	100
Likely or leaning	69.0	35.7	83.1
Solid	40.0	4.8	23.3
Incumbents	45.0	9.4	34.9
Challengers	48.0	8.1	25.7
Open-seat candidates	56.7	16.7	53.4

New Campaign Technology and Campaign Rhetoric

One additional concern that some individuals may raise is whether the rhetoric and methods of campaigning online are representative of campaign communication and strategy on the whole. The use of campaign websites themselves began a mere 20 years ago, and such websites present opportunities to engage with voters in new and unique ways. As the use of Internet and social media have expanded, studies of campaign activity have extended to include campaign websites, blogs, Twitter, Facebook, web ads, and YouTube videos (Broockman & Green, 2014; Evans, Cordova, & Sipole, 2014; Herrnson, Stokes-Brown, & Hindman, 2007; Krupnikov & Easter, 2013; Puopolo, 2001).

Although some scholars early in the process conjectured that new media campaigning would be different than previous methods (Gibson, Margolis, Resnick, & Ward, 2003), that has not proven to be the case. In spite of opportunities for innovation, over time, online campaigning has come to mirror that of the offline world (Druckman et al., 2009; Krupnikov & Easter, 2013). Campaigning through email is similar to previously studied forms of electronic campaigning that have been shown to largely mirror campaign activity offline. Candidates use it to spread news about events, discuss issues, raise money, and convey their message and preferred image to voters (Hassell, 2011; Trammell & Williams, 2004). Previous studies comparing the content of online material with that of traditional broadcast media have found them almost identical in their messages and appeals (Druckman et al., 2009; Druckman, Kifer, & Parkin, 2010; Krupnikov & Easter, 2013).

We do recognize, however, that email may have a more personal component than other forms of campaign communication. Yet, although campaign email first functioned as a mechanism for voters to ask questions of candidates, its use today is different. It is a one-way, direct, intentional communication with supporters and journalists (Herrnson et al., 2007). This communication occurs frequently and is a more personal message from a campaign than paid advertisements. Campaign email, which requires little advance work compared with print or broadcast advertising, can also be more responsive to changes throughout the race, and many of the emails in the data set were instant reactions to local or national events. As such, the use of email as a measure representative of overall campaign strategy may be one of the more accurate methods available for study.

To validate our claims about the use of email as an accurate depiction of campaign communication strategy over the course of the campaign, we conducted a survey of campaign staff involved in the creation and distribution of campaign email during the 2014 election (N = 170). We provide details about the survey methodology and some additional results in the appendix, but we present here the most relevant results which are largely consistent with previous surveys of campaign staff examining electronic campaign communications (Druckman et al., 2009).

As part of the survey, we asked campaign staff to rate how well campaign email messages were representative of campaign strategy on a 7-point scale with higher numbers signifying more representativeness. The results, shown in Figure 1, indicate that email rates above the two other common instruments used to measure dynamic campaign rhetoric, television advertisements and media coverage. Only campaign websites, campaign mail, and speeches were considered slightly more representative of campaign strategy. Although we were initially concerned, campaign email did not represent overall campaign strategy due to the perceived audience of these emails, those concerns do not seem to be drawn out. Campaign staffers indicated that campaign emails were as representative or more representative of campaign message and tone as other commonly used measures of campaign activity. In the appendix, we provide additional details about the survey.

Figure 1.

Campaign staff indication of representativeness of campaign strategy.

Data and Methodology

To test our hypotheses about the dynamics of congressional campaign rhetoric, we rely on a sample of more than 1,400 emails collected from congressional campaigns in 2012 from a random sample of 100 congressional districts. To collect the emails, we visited each campaign’s website prior to September 1, 2012, and signed up with an email address to receive campaign updates from each campaign.² We received emails from the campaigns beginning from September 1, 2012, until Election Day, November 6, 2012. We began collecting data on campaign activity on September 1 because Labor Day is the traditional start of the general election campaign season (Druckman et al., 2009), and as some states that do not hold their primary elections until the first week of September, we wanted to avoid including any leftover primary rhetoric.

After collecting the emails, we coded basic information about the candidates—candidate name, gender, party—and the email they sent—the date the email was sent, the attributed author, and information about the number of links for donations and requests for volunteers. Beyond this, we also coded issue content and mentions of party affiliation.

In addition to this basic non-subjective information, we also coded the tone of the email and any response that the email attempted to make to both local and national political events. Emails could be coded as positive, contrasting, or negative consistent with previous research on campaign tone (Goldstein & Freedman, 2002; Lau & Pomper, 2001). Emails were characterized as positive if they discussed only the candidate’s own traits or stances on issues. If, in addition to mentioning the traits or stance of the candidate, the email was critical of or challenged the opposing candidate’s views, it was coded as contrast. A negative email only contained negative or critical information about the opposition. Because emails coded as contrast made negative mention of the opposing candidate, for purposes of some of the analysis that follows, we combined contrast and negative emails into a single category. Negative and contrasting emails were also coded as to whether the negativity was directed toward personal characteristics of the opposing candidate (e.g., “It is pathetic that [my opponent] continuously uses his millions to lie about, demean, and humiliate people in his quest for power”) or toward policy positions of the opposing candidate (e.g., “We need to hold [my opponent] accountable for wanting to privatize Social Security and Medicare and turn them into voucher programs”). Negative emails could be coded as having personal negativity, issue negativity, or both. This process provides an accurate perspective of negativity that can change over the course of a congressional campaign.

Results

Campaign Rhetoric on the Whole

We first look at the overall volume of rhetoric during the campaign. As is shown in Figure 2, campaigns increased the volume of email messages sent as the campaign drew to a close. The cyclical appearance comes from the fact that campaigns were less active in sending emails out on the weekend than they were during the week. As the election draws nearer, campaigns put more frantic energy into communicating their message.

Figure 2.

Total number of email updates sent by day.

More important to our analysis, however, is the tone these messages use. Consistent with previous analyses of negativity among incumbents and challengers, we find significant differences between each of the categories. While incumbents’ emails were 68% positive, only 48% of challengers and 59% of open-seat candidates were positive (all significantly different from each other at the p < .05 level).³ Figure 3 shows the percentage of campaign communications that were classified as positive, contrasting, or negative communications separating by the competitiveness of the race when the email was sent. To measure the competitiveness of the race, we use the commonly utilized Cook Political Report’s (CPR) weekly analysis of competitive House races (Campbell, 2010). For simplicity purposes, we have combined both the likely and leaning categories into a larger category and classified campaigns in that category as either “lean win” or “lean lose.”

Figure 3.

Campaign rhetoric by likelihood of electoral victory.

Contrary to previous reports, however, we do not find that campaigns more likely to lose are more likely to engage in more negative campaigning. Figure 3 shows that campaigns are more likely to go negative if they are not sure winners. It also shows, however, that sure losers are more likely to be positive than candidates in races that CPR classified as toss-ups, leaning, or likely to be won by a particular party. Nor is there any significant difference in the tone of the campaign’s message between campaigns that CPR identified as leaning or likely to win and campaigns that CPR identified as leaning or likely to lose.⁴ In essence, we find that no matter whether a candidate is favored or viewed as behind, there is little difference in campaign strategies in races that are competitive.⁵

Dynamic Campaign Rhetoric

We now turn to examine the dynamics of rhetorical choices over the course of the campaign. Figure 4 shows the percentage of communications that contained negative content in each week leading up to the general election for each type of candidate.

Figure 4.

Negative campaign rhetoric trend by week.

As is evident from Figure 4, there appear to be very different trends of negative campaign rhetoric from incumbents, challengers, and open-seat candidates. Consistent with our findings above, challengers are consistently more likely than any other type of candidate to engage in negative rhetoric in every point of the campaign except for the final week.

In addition, contrary to previous findings about the use of negativity in presidential campaigns, we find that the use of negative rhetoric remains relatively constant over the course of the campaign. From 9 weeks before Election Day until 2 weeks before Election Day, the percentage of emails containing some negative rhetoric vacillates between 29% and 49% for incumbents and 48% and 64% for challengers. Consistent with anecdotal commentary, campaigns lower their levels of negativity significantly the week prior to Election Day. Contrary to campaign folklore, however, the only types of campaigns that appear to vary the amount of negativity significantly over the whole course of the campaign are open-seat candidates. Incumbents and challengers appear just as likely to use negative rhetoric 2 months before the election as they are 2 weeks before the election. Unlike incumbents and challengers, however, open-seat candidates begin the general election cycle relying on comparatively very little negative rhetoric. The absence of negative rhetoric does not last long, as open-seat candidates quickly increase negativity to levels above that of incumbents and rival those of challengers.

Consistent with Hypothesis 2, we find that the differences in negativity appear to be driven by issue negativity. Figure 5 shows the trends of the percentages of negative issue content and negative personal content for each type of candidate over the course of the campaign. We find only a small difference in the amounts of negative rhetoric focused on the personal characteristics of opposing candidates by open-seat candidates, challengers, and incumbents. There are, however, large differences in the amount of negative content focused on issues at the beginning of the campaign.

Figure 5.

Types of negative campaign rhetoric trends by week.

While the clearly delimited battle lines provided by the political records of established incumbents spawn the fodder for the negativity used by both challengers and incumbents, open-seat candidates do not have as much information early in the election cycle about what the defining issues of the campaign will be or where their opponent will be vulnerable. Because political attacks gone awry can be damaging to the attacking campaign, open-seat candidates appear to be more cautious about attacking opponents over issues early in the campaign.

Modeling the Dynamics of Campaign Rhetoric

To measure negative campaign rhetoric over the course of a campaign, we built a model controlling for a litany of other factors that also contribute to a campaign’s decision to engage in negative campaigning, such as party affiliation, candidate gender, and race competitiveness (Lau & Pomper, 2001). To account for shared dependence that arises from the fact that two campaigns are competing in a single district, we use random-effects logits which allow the intercepts to vary by district. We implement a random-effects logit because of concerns that negativity might be a result of unmeasured covariates that are the dependent variables from other campaigns in the district. More simply, the decision to go negative may be the result of negativity by the opponent’s campaign in the district. The correlation of these biases with the explanatory variables in the model violates logit model assumptions. A random-effects model deals with this problem by allowing the intercepts to vary by district and treats these district intercepts as if they were randomly selected from a normally distributed set of intercepts.⁶ This allows us to deal with the shared dependence that comes from a campaign operating in a specific district environment.

Table 2 contains a series of random-effect logit models predicting the likelihood that an email sent by the campaign will contain negative rhetoric. In addition to looking at the effect over time, we include controls for CPR’s weekly evaluation of the likelihood of the candidate winning the election, candidate demographics, and the type of requests (volunteer, donation, or get out the vote) contained in the message.⁷

Table 2.

Hierarchical Random-Effects Logit Predicting Negative Rhetoric in Campaign Email Communication.

	(1)	(2)	(3)
	Negative content	Negative content	Negative content
Dynamic characteristics
Days before Election Day	0.017** (−0.005)	0.017** (−0.005)	0.010* (−0.005)
Open seat	0.327 (−0.509)	0.204 (−0.504)	−0.140 (−0.478)
Incumbent	−0.555* (−0.280)	−0.300 (−0.391)	0.027 (−0.402)
Open seat × Days before ED	−0.028** (−0.008)	−0.025** (−0.008)	−0.027** (−0.009)
Incumbent × Days before ED	−0.002 (−0.008)	−0.001 (−0.008)	−0.005 (−0.008)
Campaign characteristics
Likely winner		1.987** (−0.414)	1.778** (−0.396)
Toss-up		1.813** (−0.580)	1.814** (−0.569)
Likely loser		1.623** (−0.436)	1.322** (−0.417)
Sure loser		1.336** (−0.334)	1.161** (−0.339)
Candidate characteristics
Democrat		−0.097 (−0.189)	−0.278 (−0.195)
Female candidate		−0.683** (−0.245)	−0.798** (−0.245)
Quality candidate		0.504 (−0.274)	0.235 (−0.270)
Communication content
Contains volunteer requests			−0.420** (−0.161)
Contains donation requests			1.468** (−0.214)
Contains GOTV reminder			−0.979** (−0.259)
External factors
State with senate race			0.12 (−0.365)
Battle ground state			0.458 (−0.345)
Constant	−0.724** (−0.248)	−1.894** (−0.401)	−2.471** (−0.520)
Rho	0.315 (−0.060)	0.287 (−0.062)	0.231 (−0.060)
Sigma_u	1.232 (−0.171)	1.151 (−0.176)	0.997 (−0.168)
Wald χ²	33.2**	65.8**	126.5**
Observations	1,446	1,446	1,446
Log-likelihood	−890.5	−869.6	−826.6

Note. Excluded category for campaign characteristics is sure winner. Random effect logit coefficients with standard errors in parentheses. ED = Election Day; GOTV = Get-Out-the-Vote.

p < .05. **p < .01.

As we have shown before, open-seat candidates’ choice of rhetoric over the course of the campaign varies significantly from that of incumbents and challengers. Consistent with the previous figures, we find that the dynamics of campaign negativity is different for races without an incumbent. In the original model without controls, we find that a non-open-seat candidate engages in less negativity as Election Day approaches. Candidates for open seats, however, do not engage in less negativity as Election Day nears. Instead, the models, especially the third model, suggest that they actually engage in more negativity over the course of the campaign even after controlling for the purposes of the message which they are communicating. Dealing with a campaign where battle lines may not be clearly defined, open-seat candidates must wait for more information about an opponents’ policy positions, which provide clear lines of attack before engaging in negative campaign rhetoric.

Figure 6 plots the marginal effects of running for an open seat on the likelihood that a candidate will go negative. Candidates for open seats are almost 25 percentage points less likely to engage in negative campaigning 60 days out from an election compared with incumbents and challengers. That effect, however, disappears in about 20 days, when the marginal effect becomes insignificant.

Figure 6.

Marginal effects of open-seat candidacy on negativity.

Compared with the marginal effects of other predictors of negativity shown in Figure 7, we find that being an open-seat candidate in the early stages of the campaign reduces the likelihood more than changing the gender of the candidate from male to female (a decrease in the likelihood of negativity by 15%).

Figure 7.

Marginal effects of predicted candidate negativity.

We also find that once we control for the likelihood of victory or the competitiveness of the race they are in, incumbents become no less likely than challengers to use negative campaign tactics. Likewise, the decline in negativity among incumbents and challengers as Election Day approaches appears to be driven by the purpose of the messages communicated. Consistent with theories of affect in politics, campaigns are much less likely to use anxiety-prompting negative rhetoric when they want to encourage recipients of the communication to turnout to vote and to volunteer for the campaign. As Election Day approaches, campaigns are more likely to use email communications to remind individuals to vote or to recruit volunteers to participate in get-out-the-vote (GOTV) campaigns. Volunteer and GOTV requests are disproportionately sent out in the final stages of the campaign: Almost 30% of volunteer requests and more than 60% of GOTV messages come in the final week compared with only 20% of all emails. Including a volunteer request in an email reduces the likelihood of using negativity by 8%. The use of a GOTV message reduces the likelihood of a candidate going negative on an opponent by 19%. Declines in negativity as the campaign draws to a close, then, appear not to be a conscientious decision to reduce negativity in a campaign’s rhetorical message, but rather a shift in the purposes of the messages incumbent and challenger campaigns send at that time.

Discussion

While others have speculated or drawn conclusions from limited evidence, we provide the first comprehensive examination of the dynamics of negative campaigning in congressional campaigns. For the first time, our data on congressional campaign emails allow us to track the rhetoric of congressional candidates over the course of the campaign. Contrary to anecdotal accounts, we find that incumbent and challenger campaigns do not begin the general election campaign with positive campaign rhetoric before switching to negative campaigning and finally ending with a return to positive rhetoric. Instead, incumbents and challengers alike do not drastically vary their levels of negativity over the course of the campaign. Although there is a decline in negative rhetoric as the campaign draws to a close, this decline is attributable to the increased emphasis on get-out-the-vote efforts and the volunteer recruitment that accompanies those efforts.

The dynamics of negativity exhibited by candidates in open-seat races, however, differs significantly from challengers and incumbents. The anecdotal accounts of an early positive campaign followed by an increase in negativity largely match the rhetorical patterns of congressional campaigns in open-seat races. Indeed, previous anecdotal evidence appears to have suffered from focusing on open-seat races which are generally more salient in the media. Without a candidate with a congressional record in the race, candidate rhetoric is more likely to be positive in the early stages of the race. This positive bent is driven by the relative absence of policy negativity. While we find no significant difference in the amount of personal negativity, there is a significant and substantial difference between the use of policy negativity by candidates in open-seat races compared with those in races with an incumbent. It is only after a few weeks into the general election campaign that negative campaigning in open-seat races begins to escalate before dying down again as campaigns turn to focus on get-out-the-vote efforts.

While previous research has emphasized the differences that exist between the risk-taking and rhetorical choices of incumbent and challenger campaigns, we show that the rhetorical dynamics of these campaigns are remarkably similar. Challengers and incumbents occupy the same campaign ground where at least one candidate in the race has established policy positions on which to defend or to attack. As a result, the dynamics of their rhetorical strategies, while different in the volume of negativity that they use, are similar. In contrast, the campaign ground on which open-seat candidates wage their electoral competition is distinct from that of incumbents and challengers. In these situations, candidates are less likely to have a well-established position. As a result, the dynamics of the rhetoric during the campaign is different. In conclusion, in considering the tactical decisions of electoral campaigns, it is important not only to focus on the characteristics of the candidate running for office but also on the ground on which the electoral campaign is waged.

Footnotes

Appendix

In the summer of 2014, we conducted a survey of campaign staff involved in the creation of campaign emails. We identified respondents through access to websites of U.S. House and Senate primary candidates in early 2014. We contacted the 987 campaigns that provided a workable email address or an online inquiry form in May and June of 2014. We worked to ensure that our requests were either sent out more than 4 weeks prior to the primary election or on the day after the primary election to ensure that our requests arrived at a time when campaign staff would be less busy and involved in fewer campaign activities. We asked that a campaign staff member on the campaign (or the candidate) fill out a brief and confidential survey online through Qualtrics. We contacted each campaign up to 3 times with each request coming 1 week after the last contact, with the third contact by phone. We received 176 responses (a 17.6% response rate which mirrors response rates to other nearly identical studies; Druckman, Kifer, & Parkin, 2009) and falls within a standard range of response rates (see Couper, 2008; Dillman, Smyth, & Christian, 2009). The sample reflects the population of congressional campaigns fairly well in terms of party, as 54% of responses came from Democrats, and campaign status (21% came from incumbents, 18% came from open seats, and 61% from challengers) and mirrors closely the results from previous surveys using the same methodology (Druckman et al., 2009). In addition, 13% of responses came from Senate campaigns. Removing these observations from the data has no effect on the substantive effects.

We asked respondents to indicate how well informed they were about how the content of emails from the campaign was determined, with higher scores indicating more knowledge. More than 70% indicated that they were very or extremely informed. An additional 20% indicated that they were somewhat informed, leaving less than 10% who indicated they were less informed about the process. The average rating was 5.8 on a 7-point scale.

In addition to the items reported in the text, we asked respondents to indicate how important various goals of email were on a 7-point scale. As should be expected, respondents rated fundraising (6.0) as the highest priority, followed by recruiting volunteers (5.8) and encouraging people to vote (5.7). Information dissemination on the candidate’s background (5.3) and the issues (5.5) was less important.

We also asked about the targeted recipients of campaign email and the frequency which campaign staff thought each group read. Potential donors were the highest ranked target of campaign emails (6.3 on a 7-point scale), followed by Supportive Activists (6.1) and Engaged Voters (6.0). However, it was Engaged Voters (5.6 on a 7-point scale) who campaign staff viewed as most likely to open and read the emails the campaign sent. Journalists (4.8) were also ranked high on the list as well just below Supportive Activists (5.6) and Donors (4.9), suggesting that campaigns recognize that the content they send out reaches more than just their base of supporters.

Acknowledgements

We are grateful and acknowledge the support of two Cornell College Student-Faculty Summer Research Grants which provided resources that enabled us to undertake the data coding and campaign surveys necessary to complete this research. We thank Julie Pacheco, Johanna Schuster-Craig, and reviewers and editor of this journal for their helpful and insightful comments and Nick Marn for his research assistance.

Declaration of Conflicting Interests

The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The author(s) received no financial support for the research, authorship, and/or publication of this article.

Supplementary Material

Supplementary materials are available on the American Politics Research website at .

Notes

Author Biographies

Hans J. G. Hassell is an assistant professor of Politics at Cornell College. His research focuses on the actions of parties, campaigns, candidates, and interest groups and how those actions affect the decisions of individuals to become involved in the political process.

Kelly R. Oeltjenbruns recently graduated from Cornell College. She is currently a law student at Washington University in Saint Louis.

References

Brader

(2006). Campaigning for hearts and minds: How emotional appeals in political ads work. Chicago, IL: University of Chicago Press.

Broockman

D. E.

Green

D. P.

(2014). Do online advertisements increase political candidates’ name recognition or favorability? Evidence from randomized field experiments. Political Behavior, 36, 263-289.

Campbell

J. E.

(2010). The seats in trouble forecast of the 2010 elections to the U.S. House. PS: Political Science & Politics, 43, 627-630.

Couper

M. P.

(2008). Designing effective Web surveys. New York, NY: Cambridge University Press.

Dillman

D. A.

Smyth

J. D.

Christian

L. M.

(2009). Internet, mail, and mixed-mode surveys: The tailored design method. Hoboken, NJ: Wiley & Sons, Inc.

Doherty

Adler

E. S.

(2014). Comparing the effects of positive and negative campaign mailers: Findings from a field experiment. Political Research Quarterly, 67, 562-573.

Druckman

J. N.

Kifer

M. J.

Parkin

(2009). Campaign communications in U.S. congressional elections. American Political Science Review, 103, 343-366.

Druckman

J. N.

Kifer

M. J.

Parkin

(2010). Timeless strategy meets new medium: Going negative on congressional campaign web sites, 2002-2006. Political Communication, 27, 88-103.

Evans

H. K.

Cordova

Sipole

(2014). Twitter-style: An analysis of how house candidates used Twitter in their 2012 campaigns. PS: Political Science & Politics, 47, 454-462.

10.

Flowers

J. F.

Haynes

A. A.

Crespin

M. H.

(2003). The media, the campaign, and the message. American Journal of Political Science, 47, 259-273.

11.

Fowler

E. F.

Ridout

T. N.

(2013). Negative, angry, and ubiquitous: Political advertising in 2012. The Forum, 10(4), 51-61.

12.

Freedman

Goldstein

K. M.

(1999). Measuring media exposure and the effects of negative campaign ads. American Journal of Political Science, 43, 1189-1208.

13.

Fridkin

K. L.

Kenney

P. J.

(2011). Variability in citizens’ reactions to different types of negative campaigns. American Journal of Political Science, 55, 307-325.

14.

Gibson

R. K.

Margolis

Resnick

Ward

S. J.

(2003). Election campaigning on the WWW in the USA and UK: A comparative analysis. Party Politics, 9, 47-75.

15.

Goldstein

K. M.

Freedman

(2002). Lessons learned: Campaign advertising in the 2000 elections. Political Communication, 19, 5-28.

16.

Hartman

R. R.

(2012, July 24). 78 percent of Americans frustrated by negative campaigns, poll finds. Yahoo! News. Retreived from http://news.yahoo.com/blogs/the-ticket/78-percent-americans-frustrated-negative-campaigns-poll-finds-151125834.html

17.

Hassell

H. J. G.

(2011). Looking beyond the voting constituency: A study of campaign donation solicitations in the 2008 presidential primary and general election. Journal of Political Marketing, 10, 27-42.

18.

Herrnson

P. S.

(2012). Congressional elections: Campaigning at home and in Washington (6th ed.). Washington, DC: CQ Press.

19.

Herrnson

P. S.

Stokes-Brown

A. K.

Hindman

(2007). Campaign politics and the digital divide: Constituency characteristics, strategic considerations, and candidate internet use in state legislative elections. Political Research Quarterly, 60, 31-42.

20.

Jackson

H. C.

(2010, October 28). Candidates use children to make final pitch. Seattle Times. Retreived from http://www.seattletimes.com/seattle-news/politics/candidates-use-children-to-make-final-pitch/

21.

Jacobson

G. C.

(2009). The politics of congressional elections (7th ed.). New York, NY: Pearson Longman.

22.

Kahn

K. F.

Kenney

P. J.

(1999). The spectacle of U.S. Senate campaigns. Princeton, NJ: Princeton University Press.

23.

Kennedy

(2003). A guide to econometrics (5th ed.). Cambridge, MA: The MIT Press.

24.

Krupnikov

Easter

B. C.

(2013). Negative campaigns: Are they good for American democracy. In Ridout

T. N.

(Ed.), New directions in media and politics (pp. 100-112). New York, NY: Routledge.

25.

LaPiere

R. T.

(1934). Attitudes vs. actions. Social Forces, 13, 230-237.

26.

Lau

R. R.

Pomper

G. M.

(2001). Negative campaigning by US Senate candidates. Party Politics, 7, 69-87.

27.

Lau

R. R.

Pomper

G. M.

(2004). Negative campaigning: An analysis of U.S. Senate elections. New York, NY: Rowman & Littlefield.

28.

Marist Poll. (2012, July). Campaign 2012: Civility lost. Retrieved from http://www.kofc.org/en/resources/communications/kofc_civility_lost_july_2012.pdf

29.

Mattes

Redlawsk

D. P.

(2015). The positive case for negative campaigning. Chicago, IL: University of Chicago Press.

30.

Neff

(2002, October 25). Campaign ads are accentuating the negative. Las Vegas Sun. Retreived from http://lasvegassun.com/news/2002/oct/25/campaign-ads-are-accentuating-the-negative/

31.

Price

Zaller

(1993). Who gets the news? Alternative measures of news reception and their implications for research. Public Opinion Quarterly, 57, 133-164.

32.

Puopolo

(2001). The web and U.S. senatorial campaigns 2000. American Behavioral Scientist, 44, 2030-2047.

33.

Ridout

T. N.

Fowler

E. F.

(2010). Explaining perceptions of advertising tone. Political Research Quarterly, 65, 62-75.

34.

Schuman

Johnson

M. P.

(1976). Attitudes and behavior. Annual Review of Sociology, 2, 161-207.

35.

Schuman

Presser

(1980). Public opinion and public ignorance: The fine line between attitudes and nonattitudes. American Journal of Sociology, 85, 1214-1225.

36.

Sigelman

Shiraev

(2002). The rational attacker in Russia? Negative campaigning in Russian presidential elections. Journal of Politics, 64, 45-62.

37.

Skaperdas

Grofman

(1995). Modeling negative campaigning. American Political Science Review, 89, 419-461.

38.

Stout

Kline

(2008). Ashamed not to vote for an African-American; ashamed to vote for a woman: An analysis of the Bradley effect from 1982-2006 (Working paper). Irvine: Center for the Study of Democracy, University of California, Irvine.

39.

Theilmann

Wilhite

(1998). Campaign tactics and the decision to attack. Journal of Politics, 60, 1050-1062.

40.

Trammell

K. D.

Williams

A. P.

(2004). Evaluating candidate e-mail messages in the 2002 Florida gubernatorial campaign. Journal of E-Government, 1, 105-120.

Supplementary Material

Please find the following supplemental material available below.

For Open Access articles published under a Creative Commons License, all supplemental material carries the same license as the article it is associated with.

For non-Open Access articles published, all supplemental material carries a non-exclusive license, and permission requests for re-use of supplemental material or any part of supplemental material shall be sent directly to the copyright owner as specified in the copyright notice associated with the article.

0.00 MB

0.14 MB