Political Behavior

, Volume 38, Issue 3, pp 747–766 | Cite as

Negative Advertising and the Dynamics of Candidate Support

Original Paper

Abstract

Scholars have spent a great deal of effort examining the effects of negative advertising on citizens’ perceptions of candidates. Much of this work has used experimental designs and has produced mixed findings supporting one of two competing theories. First, negative ads may harm candidates who sponsor them because citizens tend to dislike negativity. Second, negativity may drive down citizens’ support for the targeted candidate because the attacks give people reasons to reject the target. We argue that the mixed findings produced by prior research may be driven by a disregard for campaign dynamics. We present a critical test of these two theories using data drawn from 80 statewide elections—37 gubernatorial and 43 U.S. Senate contests—from three election years and public opinion polling collected during the last 12 weeks of each campaign. We find that a candidate’s support declines as her advertising strategy includes a higher proportion of negative ads relative to her opponent and that this process unfolds slowly over the course of the campaign.

Keywords

Negative campaigning Attack advertising Political advertising Campaigns Candidate support 

Introduction

Candidates for political office exert a great deal of effort attempting to select campaign strategies that will maximize their chances of winning elections. Those contesting major statewide elections largely seek to do so through the use of political television advertisements. One of the most vexing puzzles relating to advertising that social scientists have repeatedly attempted to tackle centers on the political relevance of advertising tone, especially in regards to negative advertisements, i.e. those in which ad sponsors criticize their opponents. Whereas the conventional wisdom among campaign professionals (Thurber and Nelson 2000) and political pundits is that negative advertisements are effective campaign tools, the evidence of the electoral utility of negative campaign tactics is quite mixed (Lau et al. 2007). Thus there remains more to be learned about the impact of negativity on political outcomes. We focus specifically on how negativity drives candidate support attitudes over time in order to adjudicate among these mixed fixings.

Citizens’ expressed levels of candidate support over the course of a campaign should be influenced by the levels of negativity expressed by both candidates’ campaigns. This is because negative campaigning tends to be unpopular (see for example Hitchon and Chang 1995), memorable (Brians and Wattenberg 1996) in ways that encourages citizens to think about politics in terms of avoiding costs rather than seeking benefits Lau (1982, 1985), and more informative (Freedman et al. 2004) than positive campaigning. These findings lead to several expectation, which we simplify into two competing schools of thought, both of which have some support in the extant literature. First, candidates who make greater use of negative campaign tactics may lose support (e.g. Ansolabehere et al. 1994) because citizens do not like negativity and are apt to punish candidates who run negative campaigns. Candidates may further gain support relative to their opponent when their opponents’ campaigns are more negative, as some voters either withdraw their support from the latter or switch to the former. Alternatively, we may observe the opposite phenomena: candidates who spend more time attacking their opponents may gain support (e.g. Kaid 1997) as citizens learn more about the weaknesses of the candidate’s opponent. They may also lose support as citizens learn more about their own weaknesses when their opponent’s campaign becomes more negative.1

We further make a more fundamental claim about the methodological approach to studying campaigns. We partially attribute the mixed findings in the extant literature to the lack of temporal considerations in the research designs employed in prior studies. Our analysis accounts for the dynamics of candidates’ actual strategies. These strategies may change over the course of campaigns in response to other temporal phenomena, which may in turn influence public opinion over time. We model these campaign dynamics and how strategic changes by a candidate in the use of negative advertisements affect the public’s support of candidates.

We present a critical test of these two competing theories using advertising data collected by the Wisconsin Advertising Project from 43 U.S. Senate and 37 gubernatorial general election campaigns occurring across three election cycles in 39 states. We also use public opinion data on the intended two-party vote choice reported by a large number of polling agencies collected over the last 12 weeks of these 80 statewide contests. Our results suggest that candidates’ standings in the polls improve in the short-term and suffer in the long-term when they run increasingly negative campaigns relative to their opponents. Furthermore, these long-term effects unfold slowly over the course of the campaign. This implies that candidates should take care when determining the degree to which their campaigns focus on negative messaging strategies.

Campaigns, Negativity, and Political Persuasion

While some of the early research on campaigns suggested that they were largely ineffective, more recent studies suggest that campaigns fundamentally affect the outcomes of elections and citizens’ perceptions of the political world (e.g. Wlezien and Erikson 2002; Hillygus and Shields 2008; Vavreck 2009). Campaigns tend to generate a sense of “enlightenment” about the state of the country among citizens (Gelman and King 1993) and stimulate citizens’ underlying political predispositions (Berelson et al. 1954). Campaigns may also act as priming mechanisms (Bartels 2006, but see Lenz 2009), alter citizens’ perceptions of the salience of various issues (Carsey 2000), and affect the degree to which citizens feel uncertain about political candidates (Franklin 1991; Alvarez 1997; Peterson 2004, 2009).

Although the exact mechanisms through which campaign messages influence citizens have been debated, few express skepticism that campaigns play a critical role in determining election outcomes (e.g. Holbrook 1996; Franz and Ridout 2007; Box-Steffensmeier et al. 2009). Television advertising is the primary way that candidates competing in salient contests for statewide offices communicate with citizens in modern campaigns, though the medium appears to lead only to short-term effects on public opinion (Gerber et al. 2011; Hill et al. 2013).2 Candidates attempt to persuade the electorate through the information contained within their advertisements (Goldstein and Ridout 2004). A handful of scholars, citing the normative benefits of political advertising, have argued that campaign advertising has enhanced democratic citizenship because it presents voters with informational and emotional content that contributes to a more informed, more engaged, and more participatory citizenry (e.g. Freedman et al. 2004; Mattes and Redlawsk 2015).

Negative campaigning represents a potentially useful tool for candidates. First, it explicitly centers on critiquing one’s opponent. Geer (2006), for example, argues that democracy requires some degree of negativity because without it, citizens would have no reason to consider supporting new candidates rather than incumbents. In other words, candidates may be able to use negativity to undermine support for their opponent (Skaperdas and Grofman 1995) because citizens use negative messages to inform their choices. Second, attacking may allow candidates to influence the issue emphases of their opponents because the targets of attacks may feel the need to respond in order to “set the record straight.” This kind of behavior may give candidates an opportunity to influence their opponent’s campaign messages (Damore 2002). Third, the news media tends to cover negative advertisements (Iyengar and Simon 2000), thus making it more likely that candidates will capture the attention of potential voters. Last, negativity appears to increase the likelihood that citizens will choose to support one candidate over another (Krupnikov 2012).

Much of the early work on negative campaigning suggests that increasing negativity depressed voter turnout (e.g. Ansolabehere et al. 1994; Ansolabehere and Iyengar 1995).3 More recent studies, however, suggest that increasingly negative campaigns either stimulate turnout (e.g. Finkel and Geer 1998; Goldstein and Freedman 2002; Kahn and Kenney 2004; Brader 2005; Geer 2006; Geer and Lau 2006; Jackson and Carsey 2007, but see Krupnikov 2011) or do not affect turnout (Clinton and Lapinski 2004).4 Krupnikov (2014) shows that negativity can drive turnout up or down and that these effects are heavily conditioned by context. Several explanations for the mobilization result have been offered by scholars (i.e. Finkel and Geer 1998; Jackson and Carsey 2007). For example, citizens may react more powerfully to negative campaigns than to more positive campaigns because the former make clear the differences between the two candidates (Carsey 2000) and decrease the uncertainty citizens feel towards candidates (Alvarez 1997). Citizens may also weigh negative information more heavily than positive information (Lau 1982, 1985; Garramone et al. 1990), and negative attacks may produce stronger emotional responses among voters (Marcus and MacKuen 1993; Finkel and Geer 1998). In particular, negative advertising is often meant to raise feelings of concern, anxiety, and even fear among citizens about an opposing candidate in the upcoming election.

Negative campaigning presents a potentially powerful avenue for achieving political gain. Exactly how candidates’ negative advertising strategies affect public opinion about themselves and their opponents is, however, far less clear.

Public Opinion and Campaign Negativity

Candidates attack their opponents in order to gain an electoral advantage. The means through which negativity generates this advantage may be by driving down citizens’ evaluations of the target of an attack. Citizens should be less inclined to view the target of a message favorably after being exposed to negative information about that candidate. This may not, however, be the only effect that negativity produces. Citizens may also view the candidate who sponsored the attack less favorably due to the “boomerang effect,” whereby voters punish candidates who engage in negative campaign tactics (Garramone 1984).

The research on citizens’ attitudes towards candidates who attack and who are attacked has thus far generated inconsistent findings. Some studies indicate that negative messages drive down citizens’ evaluations of the candidates targeted by negative messages (Merritt 1984; Kaid and Boydston 1987; Basil et al. 1991; Kaid 1997; Pinkleton 1998). Others show that the candidates who sponsor negative messages are viewed less favorably (Merritt 1984; Hill 1989; Martinez and Delegal 1990; Basil et al. 1991; Kaid et al. 1992; Lemert et al. 1991; Hitchon and Chang 1995; Haddock and Zanna 1997; Hitchon et al. 1997; Pinkleton 1998). Another set of findings suggest that citizens may feel sympathetic towards the targets of attacks and thus view them more favorably (Hill 1989; Martinez and Delegal 1990; Lemert et al. 1991; Haddock and Zanna 1997). In contrast, Kaid (1997) finds that candidates who attack their opponents are viewed more favorably. In addition, more recent research suggests that citizens’ evaluations of candidates in the presence of negative information can be conditioned by the characteristics of candidates (Fridkin and Kenney 2011), specifically by the candidates’ sex (Fridkin et al. 2009) and—in conjunction with individuals’ racial attitudes—race (Krupnikov 2015).

While the above research focuses on how negativity influences citizens’ evaluations of candidates, another stream in the literature centers on the degree to which negativity affects citizens’ intentions to vote for a given candidate on Election Day. The findings here are similarly mixed. Some research suggests that citizens support candidates who attack at lower levels (Shapiro and Rieger 1992; Weaver-Lariscy and Tinkham 1996; Matthew and Dietz-Uhler 1998; Lemert et al. 1999; Min 2004; Brader 2005) whereas other research implies just the opposite: that citizens are less supportive of candidates who use negative messages (Roddy and Garramone 1988; Ansolabehere and Iyengar 1995; Kaid 1997; Shen and Wu 2002; King and McConnell 2003).

The above research generates important insights about the effects of negative advertising on citizens’ views of candidates. The mixed findings uncovered by these studies may be driven in part by their lack of attention to campaign dynamics. We do not yet understand how the changing negative campaign strategies of candidates engaged in back and forth exchanges influence public opinion over time. We thus need to examine the dynamic nature of these campaigns by modeling candidates’ strategic behavior within the contextual environment of their electoral contests (e.g. Box-Steffensmeier et al. 2009; Carsey 2000; Carsey et al. 2011; Banda 2013; Windett 2014; Banda 2015; Banda and Carsey 2015). By accounting for the changes in candidate strategy inherent to campaigns, we will be able to better examine the dynamic nature of negative advertising and its impact on citizens’ evaluations of the candidates.

The inconsistencies in the literature summarized above lead to two competing hypotheses explaining how and why the balance of support for candidates competing in general elections should change over time in response to negative campaigning. The first comes out of much of the early research on negative campaigning and suggests that candidates may be punished by citizens for attacking their opponents because citizens do not like negativity (e.g. Ansolabehere et al. 1994). Accordingly, the balance of support should shift away from attacking candidates over time as citizens punish attackers for increasing their use of negative campaigning, which they may view as being uncivil. Thus the relative levels of support enjoyed by the two candidates should become more favorable for the candidate who was targeted by those attacks. Key to our argument is the notion that citizens respond to the environment created by the strategies of candidates seeking office. Citizens should be responsive to the degree to which the negativity in a campaign environment targets one candidate or the other, not just the degree to which each competing candidates’ individual campaigns are negative.5 Stated more formally:

Hypothesis 1

A candidate’s standing in the polls will decline as her campaign becomes increasingly negative relative to her opponent’s campaign.

The second theoretical framework rests on the notion that negative campaigning drives down citizens’ evaluations of the candidate who is attacked relative to the candidate who sponsors negative messages. Negative messages provide citizens with reasons to reject the candidate who is attacked. If the attacks are issue or position-based, then citizens whose views do not line up with the attacked candidate may be less supportive of that candidate after they learn about the incongruence between themselves and the attacked candidate. This may be because citizens wish to avoid the costs associated with experiencing displeasing policy outcomes. If the attacks are character-based, citizens may be more apt to reject the targets of negative messages due to the decline in feelings of trust citizens may experience after being exposed to such attacks. Negative messages, then, may provide citizens with reasons to vote against the target of an attack. Citizens should again respond to the negative campaign environments with which they are faced by altering their support for candidates. In other words:

Hypothesis 2

A candidate’s standing in the polls will improve as her opponent’s campaign becomes increasingly negative relative to her campaign.

An additional possibility is that citizens’s evaluations of both the target and the sponsors of negative messages will decrease in response to negative advertising. Citizens punish both candidates—the sponsor of the attack because citizens view negativity as uncivil and the target because of the criticisms contained within the attack. The tension underlying this theory centers on which candidate’s evaluations decrease the most. In other words, testing Hypotheses1 and 2 may allow us to determine which candidate suffers more in the polls in response to negative advertising: the target or the sponsor of negative messages?6

Research Design

We test our theory using advertising data from 43 U.S. Senate and 37 gubernatorial elections collected by the Wisconsin Advertising Project in 2000, 2002, and 2004.7 These data contain information on the date, time, and television station on which each political advertisement ran in the 100 largest U.S. media markets. We use all U.S. Senate and gubernatorial contests in which both major parties were represented by a candidate,8 both candidates ran general election television advertisements, and a substantial number of opinion polls were conducted over the course of the campaign. We eliminated contests in which at least one candidate aired fewer than 100 advertisements.9 Contests involving a sacrificial lamb are thus excluded from our analysis. We report each of the 80 contests that we analyze in Table 1.
Table 1

U.S. Senate and gubernatorial election contests included in these data

 

Senate

Gubernatorial

Both

2000

CA, FL, GA, MD, ME

IN, NC, NH

DE, MO

 

MI, MN, NE, NJ, NV

  
 

NY, PA, VA, WA

  

2002

MO, NC, NJ, SC

AZ, CA, CT, FL, HI

AL, AR, CO, IA, ME

  

IL, KS, MA, MD, MI

MN, NH, OK, OR, TN

  

NM, NY, PA, VT, WI

TX

2004

CO, FL, GA, IL, KY

IN, NH UT

MO, NC, WA

 

OK, PA, SC, WI

  

Each advertisement airing is coded for a large number of characteristics, the most important of which for this research is the tone of the ad. WiscAds codes the primary purpose of each advertisement as attacking a candidate, promoting a candidate, or contrasting the candidates. Attack advertisements contain information that is only relevant, at least explicitly so (but see Banda 2014), to the target of the ad. They are entirely negative in nature. Contrast ads, on the other hand, contain some information about the opposing candidate and positive information about the sponsor. They are thus only partially negative. Because these two types of advertisements provide citizens with fundamentally different kinds of information, citizens may be influenced by them in different ways. Thus we only treat ads coded as attacks as being negative.

We analyze 637,059 candidate-sponsored advertisement airings across these 80 general election contests.10 Democratic candidates sponsored 325,515 of these ads while the remaining 311,544 were sponsored by Republican candidates. Just under 22 % of Democratic ads were negative whereas approximately 27 % of Republican-sponsored ads were negative.11 The WiscAds data includes variables indicating whether or not advertisements focus on policy, personal characteristics, or both. Only about 5 and 4.5 % of Democratic and Republican-sponsored ads were coded as containing only personal attacks. The remaining negative ads utilized policy-based attacks or both policy and personal attacks.

Our measure of candidate support is drawn from an original data set composed of polling data collected from the National Journal’s Daily Hotline report. These data represent a collection of most of the public opinion polls that centered on candidate support conducted during a given election cycle. We collected additional polling data from the Polling Report.12 The data made available on these websites generates a nearly complete population of polling data that has been made publicity available. Our data are produced from about 1,560 polls fielded by around 300 different agencies. These data represent the Democratic candidate’s level of support over the course of each general election contest. High values of this variable indicate that the Democrat enjoys a high degree of support among potential voters while low levels indicate that the Republican candidate is advantaged.

To avoid the potential for the polling data to be skewed due to voter uncertainty in the early stages of campaigns and to account for missing data, we utilize Stimson’s (1999) WCalc polling algorithm to generate smoothed estimates of candidates’ polling positions. This algorithm accounts for present and past values of the series to calculate a more accurate estimate of the position of the candidates. Where there is little variation in polling position, the impact of the smoothing process is minimal. Where there is high variance, on the other hand, the smoothing effect is much greater.

We start with the two-party Democratic vote share as our polling indicator. This means we are smoothing our polls based only on responses for preference, excluding undecided and third party responses. We structure our data in a weekly format, utilizing the last 12 weeks of a campaign because this is the time period during which we have polling data. If a poll spans two weeks of a campaign, we identify the last day the poll was in the field as the week of the poll.13 The algorithm estimates an exponential smoothing model in which the output is a smoothed value based on the raw polling values for that week as well as the values observed at previous time points.14
Fig. 1

Examples of smoothed and raw polling results by week

The influence of the smoothing is dependent on the amount of data included in the series. Where there are large amounts of data that is highly patterned, the influence of the smoothing algorithm is minimal. Series that have less patterned data, or greater levels of between period fluctuations, will see a much large influence in the smoothing process. In Fig. 1, we show two examples of the influences of the smoothing algorithm. The left panel shows the smoothed and raw polling results in Colorado’s 2004 open seat Senate race between Democrat Ken Salazar and Republican Peter Coors, a contest in which Salazar won by a margin of 3.9 %. In this race, we observe 28 polls across 8 of the 12 weeks of the election. In this case, the polling data are relatively dense. The second panel shows both types of polling data for the 2000 Georgia Senate contest in which Incumbent Democrat Zell Miller defeated Republican challenger Mack Mattingly by over 20 %. The polling data in this contest is not as dense as in the previous example, likely due to the lower level of electoral competition.

Our analysis utilizes weekly measures of both campaign advertising negativity and candidate support as captured by the polling data discussed above. For the advertising data, this generates contest-level weekly time series capturing the negative advertising strategies of both major party candidates in each contest. Generating weekly data like this allows random fluctuations in both candidates’ advertising strategies and their levels of support as indicated by the polls to cancel out through the aggregation process.

Table 2 contains weekly-level summary statistics of our data. “Democratic polling advantage,” the dependent variable, ranges from roughly 15 to 80. Our independent variable of interest, the “Democratic attack advantage” of a campaign, captures the negative advertising strategies of competing Democratic and Republican candidates. We calculate it by subtracting the percentage of negative ads aired by Republicans from the percentage of Democratic-sponsored negative spots that aired in a given week and contest. This measure thus accounts for the balance of negative advertising in a given week between competing candidates. Higher values thus indicate that the Democratic candidate devoted a larger share of her weekly advertising agenda to negative ads than did her Republican opponent.
Table 2

Summary statistics

 

Mean

Standard deviation

Minimum

Maximum

Democratic polling advantage

48.71

8.90

15.60

80.34

Democratic attack advantage

−3.63

34.19

−100

100

Gubernatorial election

0.47

0.50

0

1

Democrat is an incumbent

0.20

0.40

0

1

Republican is an incumbent

0.31

0.46

0

1

Year: 2002

0.52

0.50

0

1

Year: 2004

0.22

0.42

0

1

We control for whether or not an election was for governor (46 %) and whether or not the Democratic (20 %) or Republican (31 %) candidate is an incumbent. About 26 % of the contests occurred in 2000 while 49 and 25 % of the contests took place in 2002 and 2004 respectively. In total, Democratic candidates aired 325,515 ads, 22 % of which were negative ads, and Republican candidates aired 311,544 advertisements, of which 27 % were negative.

Modeling Campaigns and Candidate Support Dynamically

We use an error correction model (ECM) when estimating our equations in order to model campaign behavior and public opinion dynamically.15 ECMs can be used to analyze data that is integrated or stationary (DeBoef and Keele 2008). The general ECM is algebraically equivalent to an auto-regressive distributed lag model (ADL). Both types of models allow researchers to observe both the contemporaneous and the long-run effects of a change in an independent variable on the dependent variable.

The dependent variable of an error correction model must be the first difference rather than the value of the variable at time t. In a single time series, differencing forces stationarity by extracting any first-order autocorrelation that may exist within the series. As (Wooldridge (2000) Chap. 14) notes, first differencing pooled time series data essentially removes the differences between the mean levels of the variables across the pooling unit, which in this case is across contests. First differencing of pooled time series data is identical to including unit fixed effects when there are only two time points. When there are more than two time points, first differencing is not identical to including unit fixed effects, but they are similar to one another. First differencing therefore essentially removes any unit effects that might exist across the campaigns we analyze. Any constant differences across campaigns like the presence of an incumbent, state characteristics, or pre-campaign expectations of competitiveness are effectively controlled for by this design.

This framework also requires the inclusion of a lagged dependent variable, the coefficient of which estimates the rate of error correction, and both first differences and lagged levels of the remaining time serial covariates. While our theory centers on the influence of candidate advertising strategies on candidate support, it is possible that the strategies employed by candidates are informed by their polling positions. We use a seemingly unrelated regression (SUR) approach to account for this possibility.16 A SUR is useful in this case because it allows us to observe and account for the degree to which candidate support and candidate strategy are driven by one another. We thus estimate a two equation SUR as follows:
$$\begin{aligned} \Delta DemSupport_{it} = \beta _0 + \beta _1DemSupport_{it-1} + \beta _2\Delta Neg_{it} + \beta _3Neg_{it-1} + \beta _qControl + \epsilon \\ \Delta Neg_{it} = \phi _0 + \phi _1Neg_{it-1} + \phi _2\Delta DemSupport_{it} + \phi _3DemSupport_{it-1} + \phi _qControl + \varepsilon \end{aligned}$$
\(DemSupport_{it}\)” refers to the level of support for the Democratic candidate expressed through public opinion polling in contest i and week j. “\(Neg_{it}\)” represents the Democratic attack advantage, again in contest i and week j. “Control” refers to a matrix of control variables including whether or not the contest is for a governorship, if the Democrat is an incumbent, and if the Republican is an incumbent.

The coefficients for the differenced independent variables estimate the average contemporaneous change in the dependent variable that results from a one unit increase in the covariate. Contemporaneous changes occur at time t. The coefficients of the lagged endogenous variables captures part of a second short-term effect,17 this time at time t  + 1. These effects may not be theoretically interesting on their own, but when the coefficients that estimate them are divided by the negative of the coefficient generated for the lagged dependent variable, they generate the long run multiplier (LRM), which represents the total contemporaneous and long-run effect of a one unit increase in an endogenous covariate on the dependent variable.18 For this reason, the LRMs are the most interesting quantity produced by the model given the hypotheses we test.

Expectations

We are interested in observing the effects of the negative advertising environment created by competing candidates’ television advertising strategies on support for Democratic candidates over the course of general election campaigns, so we focus primarily, but not exclusively, on the long run multipliers produced by our models. The two broad theories presented above produce divergent hypotheses about the effects of negative advertising on candidate support. Hypothesis 1 is predicated on the notion that candidates who attack more than their opponent will lose support from citizens relative to their opponents. Hypothesis 2 suggests the opposite: candidates who attack more than their opponents will receive higher levels of support relative to their opponents as expressed by public opinion polls. If the long run multiplier of the Democratic attack advantage measure is negative, this will provide evidence in support of the Hypothesis 1 because it will indicate that support for Democratic candidates declines as Democrats attack more than their Republican opponents. If the signs of the long run multiplier is instead positive, the evidence will favor Hypothesis 2 because it would indicate that support for Democrats increases as Republican candidates attack more relative to Democrats.

Results

Table 3 contains the results of a two equation SUR model in which one equation estimates the level of support for the Democratic candidate while the other estimates the advantage in negative advertising enjoyed by the Democrat relative to her Republican opponent.19 We first turn to observing the effects of Democratic attack advantage on Democratic polling advantage. We focus on two key results. First, the coefficient generated for the first difference of the Democratic attack advantage indicator is positive and differs significantly (p ≤ 0.05) from zero, indicating that a one percentage-point increase in the negative advertising environment measure on average leads to a 0.008 percentage-point increase in Democratic support at time t. This modest effect offers preliminary support for Hypothesis 2 because it indicates that candidates who attack more relative to their opponents enjoy a small contemporaneous boost in their levels of popular support.
Table 3

Negative Advertising and Candidate Support in Gubernatorial and Campaigns, 2000–2004

 

Democratic polling advantage

Democratic attack advantage

Long run multipliers

  

 Democratic attack advantage

−0.100* (0.003)

 

 Democratic polling advantage

 

−0.590* (0.146)

\(\Delta\)Democratic attack advantage

0.008* (0.003)

 

Democratic attack advantage\(_{t-1}\)

−0.003 (0.003)

−0.448* (0.034)

\(\Delta\)Democratic polling advantage

 

1.050* (0.439)

Democratic polling advantage\(_{t-1}\)

−0.030* (0.013)

−0.265 (0.147)

Gubernatorial election

−0.146 (0.196)

−2.328 (2.294)

Democratic incumbent

0.476 (0.255)

1.371 (2.999)

Republican incumbent

−0.337 (0.240)

−1.807 (2.818)

Intercept

1.501* (0.653)

13.033 (7.676)

BIC

10,322.12

 

N

714

 

Estimated ordinary least squares coefficients are reported along with standard errors in parentheses. Long-run multipliers and their standard errors were generated using the Bewley (1979) transformation. BIC Bayesian information criteria

* p ≤ 0.05 (two tailed),  p ≤ 0.1 (two tailed)

The second result, however, offers support for Hypothesis 1. The LRM for the Democratic attack advantage measure is negative and its effect attains a traditional level of statistical significance (p ≤ 0.05). This LRM suggests that a one unit increase in the attack environment—i.e. when a Democratic candidate devotes 1 % more of their advertisements to attacks than does their Republican opponent—leads to a combined average decrease of 0.1 percentage points in support for the Democratic candidate across contemporaneous and future time periods. In other words, as the Democrat attacks more relative to the Republican, the Democrat’s support as measured by public opinion polls declines over the course of the campaign.

We next turn to the results of the equation predicting the Democratic attack advantage measure. As indicated by the estimated coefficient for the first difference of Democratic polling advantage, a one percentage point increase in the Democrat’s standing in the polls on average leads to a 1.05 unit contemporaneous increase in the Democrat’s advertising advantage. The LRM of \(-0.59\) produced in this equation, on the other hand, shows that in the long run, this effect is negative.

While the LRMs capture the total short and long-term effects of interest, they do not allow us to directly observe how these effects unfold over specific time periods. The dynamics of this process should be of interest to scholars, perhaps more so given the differences between the contemporaneous and long-run effects uncovered by both equations described above. We plot distributed lag effects over the course of five weeks from both equations in Fig. 2.20 These effects are generated by a one standard deviation increase in the key dependent variables for both equations on the dependent variables in the week that the change occurs in addition to the four following weeks. A one standard deviation increase in the Democratic attack advantage measure leads on average to a total decrease in Democratic polling advantage of about 3.4 percentage-points. A one standard deviation increase in the Democratic polling advantage indicator, on the other hand, leads to a total decrease of 5.3 unit decrease in Democratic attack advantage.
Fig. 2

The distributed effects of (left panel) candidates’ negative advertising strategies on polling support for the Democratic candidate and (right panel) support for Democratic candidates on candidates’ negative advertising strategies. Note that (1) the scale of the Y axes differ greatly between panels and (2) the effects plotted here were generated by one standard deviation increases in either the negative advertising environment or Democratic support as appropriate

The left panel of Fig. 2 shows the effects of Democratic attack advantage on support for the Democratic candidate. The contemporaneous effects—the effect at time t or week 0 in Fig. 2—are positive whereas the effects over future time periods are all negative. This is congruent with the results described above. On average, a one standard deviation increase in Democratic attack advantage (i.e. an increase in the percentage of attack ads aired by Democrats relative to the percentage aired by Republicans), leads to a 0.26 percentage point increase in the support enjoyed by the Democratic candidate during the week in which the change in Democratic attack advantage occurs. Democratic support then declines in the following four weeks by 0.111, 0.108, 0.105, and 0.102 percentage points respectively. Across the five weeks, these changes sum to a total decline in Democratic support of 0.167 percentage points. This is only 5 % of the 3.4 percentage point decrease in Democratic support estimated by the LRM. The effects of a change in Democratic attack advantage on Democratic support thus unfold slowly over the course of many weeks. In the long run, the more candidates attack relative to one another, the worse their standing in the polls becomes. The short-term benefits enjoyed by candidates who attack more than their opponents appear to quickly disappear. Thus our results primarily offer support for Hypothesis 1.

To what extent can candidates take advantage of the short-term benefits of attacking given the long-term costs? Suppose that Democratic advertising advantage increased by one standard deviation not just in week 0 as above, but again in weeks 1 and 2. Thus through addition we can estimate the effects of these strategic decisions over the same five week period. The Democratic polling advantage would again increase by 0.26 units as above, but the change in Democratic support in week 1 would now be an increase of 0.149 percentage points (\(0.26 - 0.111\)). In week 2, the Democratic support would increase by 0.041 units (\(0.26 - 0.108 - 0.111\)). These effects during weeks 3 and 4 would also similarly cumulate such that the changes would equal \(-0.324 (-0.105 - 0.108 - 0.111\)) and \(-0.315\) (\(-0.102 - 0.105 - 0.108\)) respectively. These results thus suggest that continuous doses of disproportionate negativity can decrease the potential utility of the short-term benefits of negative advertising.

Figure 2’s right panel shows the effects of a one standard deviation increase in Democratic polling advantage on Democratic attack advantage, again over five weeks. This change leads to a contemporaneous increase in the Democratic attack advantage measure of 9.315 units—i.e. Democrats on average devote a larger share of their advertisements to attacks relative to their Republican opponents. But this initial positive effect becomes negative over the following weeks as indicated by the plotted decreases of 6.576, 3.626, 1.999, and 1.102 units in weeks one through four. The total change in Democratic attack advantage across these five weeks is \(-3.989\). This is about 75 % of the of the 5.3 unit decrease generated by the previously discussed LRM. These effects relatively quickly. On the whole, this suggests that as candidates attract higher levels of support, they devote a smaller portion of their total campaign advertising agenda to attacks relative to their opponents.

Conclusions

The results of our analysis suggest that candidates’ negative advertising strategies relative to one another influence the levels of electoral support they enjoy as captured by public opinion polls over the course of campaigns. The support enjoyed by candidates who attack more than their opponents declines over time. Thus candidates who sponsor more negative ads relative to their opponents appear to harm themselves overall. These effects further appear to unfold slowly over the course of the campaign. This suggests that citizens—at least at the aggregate level—may respond to changes in candidates’ campaign strategies weeks after the changes occur. Candidates, then, may benefit or suffer from the strategic decisions they make early in their campaigns long after they employ a given strategy. This may be one reason why candidates struggle to alter their images during campaigns.

Given this finding, why do candidates continue to attack one another during election campaigns? There are likely at least two reasons. First, our analysis suggests that candidates enjoy a small short-term boost to their levels of support during the week in which they attack more relative to their opponents. Candidates may weigh the short-term benefits against the long-terms costs of attacking and decide that the former outweighs the latter, a calculation that is likely easier towards the end of the campaign season because the long-term costs are not important after Election Day. Thus, our findings dovetail with previous research showing that candidates’ campaign strategies can have short-term effects on public opinion (Gerber et al. 2011; Hill et al. 2013; Doherty and Adler 2014), but they are also surprising because they show that, at least in this case, attack advertising also has an impact on opinion that plays out slowly over time.

Second, prior research shows that candidates respond to one another’s negative advertising strategies (Carsey et al. 2011). They may do so for a number of reasons: to counter criticisms, level their own critiques against their opponents, or to try to change the focus of the debate. Whatever their reasoning, candidates change their advertising strategies in response to those used by their opponents and thus when one attacks, the other may do so as well in response. Thus they may be induced to attack, even when doing so is not in their best interests.

The results of this research lead to two key implications. First, because candidates’ negative advertising strategies can influence support for their candidacies, candidates may have a strategic incentive to try to induce their opponents to spend more time attacking them than they spend attacking their opponent, especially early in their campaigns. Candidates may risk attacking their opponents and driving down citizens’ evaluations of them in the hopes that their opponent will respond disproportionately, thus driving down their own support to a greater extent. Future research should examine this possibility in greater depth. For example, do certain kinds of negative messages—perhaps personal attacks rather than attacks focused on policy—provoke greater responsiveness from a candidate’s opponent? Do personal attack produce changes in candidate support that are similar to those produced by policy-based attacks?

Second, our results further suggest that researchers should take care when analyzing dynamic phenomena using cross-sectional methods. Our findings are at odds with some prior research. This may be in part because our tests account for the dynamic nature of campaigns while those reported in the extant literature do not. We also tested our theory observationally using the actual strategies employed by candidates and real world public opinion polls rather than with a laboratory experiment. Experiments are powerful and useful tools for social scientists, but their levels of external validity are sometimes low. Many of these experimental studies used relatively small numbers of subjects selected from nonrepresentative convenience samples, factors that may contribute to the mixed findings reported in the literature. These limitations, especially the small sample sizes, may call into question the generalizeability of the extant literatures’ findings (but see Druckman and Kam 2011 for a defense of student samples). Our research does not suffer from these limitations.

Our results are limited in three ways. First, we only examined candidate behavior in state-wide elections that were competitive enough for both candidates to mount substantial television advertising campaigns. It is possible that negative campaign strategies influence public opinion in unique ways that we did not observe in different—i.e. in less competitive or in less salient—contexts. Citizens may respond to negativity differently in campaigns that are not very competitive or in local elections in which candidates are more likely to meet with potential voters face to face. Second, campaign strategies may be different in less salient contests and in elections for different types of offices. Third, we focus on campaign behavior as expressed through television advertising, but there are many other forms of campaign communication including public speeches, media appearances, direct mail, e-mail, and door to door canvassing. Public opinion may be affected by negativity transmitted through these different communication mediums in different ways.

Taken as a whole, these results suggest that the attack advertising strategies employed by candidates matter. These strategies influence public opinion in the form of candidate support over the course of general election campaigns. Future research might examine how different kinds of candidates benefit or suffer from negative advertising over the course of campaigns and how other kinds of advertising strategies—such as issue emphasis or the use of contrast ads—affect public opinion, both in primary and in general elections. It might also examine the dynamics of political advertising that is sponsored by groups like political parties and super PACs.

Footnotes

  1. 1.

    It is also possible that negative messages fail to influence public opinion.

  2. 2.

    Doherty and Adler (2014) report evidence that campaign mailers produce similar short-term effects on public opinion.

  3. 3.

    Kahn and Kenney (1999) argue that campaigns should stimulate turnout as they become more negative, but only until the election environment becomes overly saturated with negative information. When this happens, they argue that turnout will decline.

  4. 4.

    That said, the results in this literature are decidedly mixed. See Lau et al. (2007) for a thorough meta-analysis on negativity’s effects on turnout and other topics.

  5. 5.

    The above argument centers on the support enjoyed by competing candidates relative to one another because this is the focus of our analysis. We could also consider the level of support enjoyed by individual candidates. Citizens, for example, may be less likely to support candidates who attack while their likelihood of supporting the targeted candidates may be unaffected. However the process unfolds, candidates who attack their opponents should suffer.

  6. 6.

    Because of the nature of our public opinion polling data—the Democratic candidate’s share of the two-party intended vote—that we discuss below, we are unable to test this theoretical possibility independently of the above hypotheses. Still, it provides an additional reason why we might observe higher or lower levels of support for one candidate relative to the other.

  7. 7.

    The data were obtained from a project of the University of Wisconsin Advertising Project includes media tracking data from TNSMI/Campaign Media Analysis Group in Washington, D.C. The University of Wisconsin Advertising Project was sponsored by a grant from The Pew Charitable Trusts. The opinions expressed in this article are those of the authors and do not necessarily reflect the views of the University of Wisconsin Advertising Project or The Pew Charitable Trusts.

  8. 8.

    We treated Paul Wellstone and Walter Mondale as a single candidate in Minnesota’s 2002 Senate race due to the former’s death late in the campaign. Excluding the race from our analysis does not alter the substantive findings we report in this research.

  9. 9.

    This eliminated four contests. The Democrat in the 2004 Senate election in Ohio aired 69 advertisements. In the other three contests, at least one of the candidates aired fewer than 10 ads. This choice did not alter the substance of the findings reported in this research.

  10. 10.

    We do not include advertisements sponsored by political parties or interest groups in our analyses because this research focuses only on the political impact of candidate strategy. We also do not include any advertisements aired by 501(c) groups, which did not become prominent parts of campaigns for federal offices until the 2004 election cycle.

  11. 11.

    An additional 28 and 22 % of Democratic and Republican-sponsored spots were coded as contrast ads. All but a very small number of the remaining advertisements were coded as positive.

  12. 12.

    National Journal data and Polling Report data are subscription based polling agencies. Website access can be found at www.nationaljournal.com and www.pollingreport.com.

  13. 13.

    WCalc can estimate daily, monthly, quarterly, annual, or multi-year series of polling figures.

  14. 14.

    See the Wcalc manual at http://stimson.web.unc.edu/files/2015/08/Wcalc6 for more detailed information. Stimson notes that WCalc “implements the Dyad Ratios algorithm for building a continuous regular time series from the scraps of dated survey results that are typically available for public opinion analysis. It emulates the logic of principal components analysis in most regards but, unlike principal components, it does not require that all variables have a complete set of cases or, indeed, anything close to it. Input is survey data expressed in a summary score, e.g., percent liberal responses, which is dated and includes a number of cases (although the program will accept 0 for N when it is unknown).”

  15. 15.

    Grant and Lebo (2016) argue that ECMs are often misunderstood in the literature and make several recommendations for practitioners. We address their concerns in an appendix.

  16. 16.

    Multiple equation VAR models are equivalent to SUR models as long as the models exhibit the same lag structure in each equation (Hamilton 1994, p. 314).

  17. 17.

    Many scholars refer to these coefficients as long-term effects because they can be transformed algebraically to produce estimated effects on the dependent variable across additional future time periods.

  18. 18.

    We calculate the standard errors of each long run multiplier using the Bewley (1979) transformation (see also DeBoef and Keele 2008).

  19. 19.

    Replication materials for all of the analyses presented in the manuscript and its appendix are available at https://dataverse.harvard.edu/dataset.xhtml?persistentId=doi:10.7910/DVN/U1GGGQ.

  20. 20.

    Error correction models are algebraically equivalent to auto distributed lag models. Because the latter allow researchers to easily calculate the effect of a change in the independent variable on the value of the dependent variable in individual time periods, this can be done in an error correction context as well with a bit of simple algebra (DeBoef and Keele 2008). Researchers need to use three coefficients from the output of an ECM to calculate these effects: those of lagged Y, lagged X, and the first difference of X. Using the results of the equation predicting Democratic support from Table 3 as an example, that means that we use the coefficients generated for lagged Democratic polling advantage (\(-0.03\)), lagged Democratic attack advantage (\(-0.003\)), and differenced Democratic attack advantage (0.008). Assume that we are interested in calculating the distributed effects of a one unit increase in the Democratic attack advantage on Democratic polling advantage. The contemporaneous effect is captured by the estimated coefficient of the differenced Democratic attack advantage indicator: 0.008 \(\times\) 1 = 0.008. Thus Democratic support increases by 0.008 units when the attack environment increases by one unit during the week in which the change occurred. We can also calculate the effects of this change that are distributed over future weeks. In the week following the change in the Democratic attack advantage, this effect can be calculated by subtracting the coefficient of differenced Democratic attack advantage indicator from the coefficient of lagged Democratic attack advantage measure and multiplying that quantity by the change that we are interested in, in this case an increase of one unit. We then add that quantity to one minus the absolute value of the coefficient of lagged Democratic polling advantage, which is multiplied by the contemporaneous change. Algebraically, we can express that as follows: (\(-0.003 - 0.008\)) \(\times 1 + (1 - |-0.03|) \times 0.008 = -0.003\). The formula to calculate the effect of a change in the Democratic attack advantage on Democratic polling advantage in all periods after time t + 1 is simpler: \((1 - |-0.03|) \times\)g, where g is the distributed effect calculated for the previous time period. Thus the effect during the second week after the change in the Democratic attack advantage is (1 − |−0.03|) \(\times\) −0.003 = −0.0029, in the third week it is (1 − |−0.03|) \(\times\) −0.0029 = −0.0028, and so on. The the effect of a one unit increase in the Democratic attack advantage leads to an immediate increase in Democratic polling advantage followed by several small decreases in Democratic polling advantage over future weeks. Note that when the estimated coefficient of the lagged dependent variable is small, the effect of a change in an independent variable on the value of the dependent variable decay slowly over time.

Notes

Acknowledgments

We would like to thank Tom Carsey, Peter Enns, Nate Kelly, Jamie Monogan, Jeff Harden, John Henderson, Alex Theodoridis, Susanne Martin, Clayton Cleveland, four anonymous reviewers, and Dave Peterson for their helpful comments and suggestions about earlier versions of this research.

Supplementary material

11109_2016_9336_MOESM1_ESM.pdf (157 kb)
Supplementary material 1 (PDF 157 kb)

References

  1. Alvarez, R. M. (1997). Information and elections. Ann Arbor: University of Michigan Press.CrossRefGoogle Scholar
  2. Ansolabehere, S., & Iyengar, S. (1995). Going negative: How political advertisements shrink and polarize the electorate. Free Press.Google Scholar
  3. Ansolabehere, S., Iyengar, S., Simon, A., & Valentino, N. (1994). Does attack advertising demobilize the electorate? American Political Science Review, 88(December), 829–838.CrossRefGoogle Scholar
  4. Banda, K. K. (2013). The dynamics of campaign issue agendas. State Politics and Policy Quarterly, 13(4), 446–470.CrossRefGoogle Scholar
  5. Banda, K. K. (2014). Issue-based negativity and candidate assessment. Public Opinion Quarterly, 78(3), 707–720.CrossRefGoogle Scholar
  6. Banda, K. K. (2015). Competition and the dynamics of issue convergence. American Politics Research, 43(5), 821–845.CrossRefGoogle Scholar
  7. Banda, K. K., & Carsey, T. M. (2015). Two-stage elections, strategic candidates, and campaign agendas. Electoral Studies, 40, 221–230.Google Scholar
  8. Bartels, L. (2006). Priming and persuasion in presidential campaigns. In H. E. Brady & R. Johnston (Eds.), Capturing campaign effects (pp. 78–112). Ann Arbor: University of Michigan Press.Google Scholar
  9. Basil, M., Schooler, C., & Reeves, B. (1991). Positive and negative political advertising: Effectiveness of ads and perceptions of candidates. In F. Biocca (Ed.) Television and poltical advertising (Vol. 1). Lawrence Erlbaum.Google Scholar
  10. Berelson, B. R., Lazarsfeld, P. F., & McPhee, W. N. (1954). Voting: A study of opinion formation in a presidential campaign. Chicago: University of Chicago Press.Google Scholar
  11. Bewley, R. A. (1979). The direct estimation of the equilibrium response in a linear dynamic model. Economics Letters, 3(4), 357–361.CrossRefGoogle Scholar
  12. Box-Steffensmeier, J. M., Darmofal, D., & Farrell, C. A. (2009). The aggregate dynamics of campaigns. Journal of Politics, 71(January), 309–323.CrossRefGoogle Scholar
  13. Brader, T. (2005). Striking a responsive chord: How campaign Ads motivate and persuade voters by appealing to emotions. American Journal of Politial Science, 49(March), 388–405.CrossRefGoogle Scholar
  14. Brians, C. L., & Wattenberg, M. P. (1996). Campaign issue knowledge and salience: Comparing reception from TV commercials, TV news, and newspapers. American Journal of Politial Science, 40(January), 172–193.CrossRefGoogle Scholar
  15. Carsey, T. M. (2000). Campaign dynamics: The race for governor. Ann Arbor: University of Michigan Press.CrossRefGoogle Scholar
  16. Carsey, T. M., Jackson, R. A., Stewart, M., & Nelson, J. P. (2011). Strategic candidates, campaign dynamics, and campaign advertising in gubernatorial races. State Politics and Policy Quarterly, 11(September), 269–298.CrossRefGoogle Scholar
  17. Clinton, J. D., & Lapinski, J. S. (2004). ’Targeted’ advertising and voter turnout: An experimental study of the 2000 presidential election. The Journal of Politics, 66(February), 69–96.CrossRefGoogle Scholar
  18. Damore, D. F. (2002). Candidate strategy and the decision to go negative. Political Reseach Quarterly, 55(September), 669–686.CrossRefGoogle Scholar
  19. DeBoef, S., & Keele, L. (2008). Taking time seriously. American Journal of Political Science, 52(1), 184–200.CrossRefGoogle Scholar
  20. Doherty, D., & Scott Adler, E. (2014). The persuasive effects of partisan campaign mailers. Political Reseach Quarterly, 67(3), 562–573.CrossRefGoogle Scholar
  21. Druckman, J. N., & Kam, C. D. (2011). Students as experimental participants: A defense of the “Narrow Data Base. In J. N. Druckman, D. P. Green, J. H. Kuklinski, & A. Lupia (Eds.), Cambridge handbook of experimental political science (Vol. 4, pp. 41–57). New York, NY: Cambridge University Press chapte.Google Scholar
  22. Finkel, S. E., & Geer, J. G. (1998). A spot check: Casting doubt on the demobilizing effect of attack advertising. American Journal of Political Science, 42(April), 573–595.CrossRefGoogle Scholar
  23. Franklin, C. H. (1991). Eschewing obfuscation? Campaigns and the perception of U.S. Senate incumbents. American Political Science Review, 858, 1193–1214.CrossRefGoogle Scholar
  24. Franz, M. M., & Ridout, T. N. (2007). Does political advertising persuade? Political Behavior, 29(December), 465–491.CrossRefGoogle Scholar
  25. Freedman, P., Franz, M., & Goldstein, K. (2004). Campaign advertising and democratic citizenship. American Journal of Political Science, 48(October), 723–741.CrossRefGoogle Scholar
  26. Fridkin, K. L., & Kenney, P. J. (2011). Variability in citizens’ reactions to different types of negative campaigns. American Journal of Political Science, 55(2), 307–325.CrossRefGoogle Scholar
  27. Fridkin, K. L., Kenney, P. J., & Woodall, G. S. (2009). Bad for men, better for women: The impact of stereotypes during political campaigns. Political Behavior, 31(1), 53–77.CrossRefGoogle Scholar
  28. Garramone, G. M. (1984). Voter responses to negative political ads. Journalism Quarterly, 61(Summer), 250–259.CrossRefGoogle Scholar
  29. Garramone, G. M., Atkin, C. K., Pinkleton, B. E., & Cole, R. T. (1990). Effects of negative political advertising on the political process. Journal of Broadcasting and Electronic Media, 34(Summer), 299–311.CrossRefGoogle Scholar
  30. Geer, J., & Lau, R. R. (2006). Filling in the blanks: A new method for estimating campaign effects. British Journal of Political Science, 36(April), 269–290.CrossRefGoogle Scholar
  31. Geer, J. G. (2006). In defense of negativity: Attack ads in presidential campaigns. Chicago, IL: The University of Chicago Press.Google Scholar
  32. Gelman, A., & King, G. (1993). Why are American presidential election campaign polls so variable when votes are so predictable? British Journal of Political Science, 23, 409–451.CrossRefGoogle Scholar
  33. Gerber, A. S., Gimpel, J. G., Green, D. P., & Shaw, D. R. (2011). How large and long-lasting are the persuasive effects of televised campaign ads? Results from a randomized field experiment. American Political Science Review, 105(1), 135–150.CrossRefGoogle Scholar
  34. Goldstein, K., & Freedman, P. (2002). Campaign advertising and voter turnout: New evidence for a stimulation effect. The Journal of Politics, 64(August), 721–740.CrossRefGoogle Scholar
  35. Goldstein, K., & Ridout, T. N. (2004). Measuring the effects of televised political advertising in the United States. Annual Review Political Science, 7, 205–226.CrossRefGoogle Scholar
  36. Grant, T., & Lebo, M. J. (2016). Error Correction models with political time series. Political Analysis, 24(1), 3–30.CrossRefGoogle Scholar
  37. Haddock, G., & Zanna, M. P. (1997). The impact of negative advertising on evaluations of political candidates: The 1993 Canadian federal election. Basic and Applied Social Psychology, 19(June), 204–223.Google Scholar
  38. Hamilton, J. D. (1994). Time series analysis. Princeton, NJ: Princeton University Press.Google Scholar
  39. Hill, R. P. (1989). An exploration of voter responses to political advertisements. Journal of Advertising, 18(Winter), 14–22.CrossRefGoogle Scholar
  40. Hill, S. J., Lo, J., Vavreck, L., & Zaller, J. (2013). How quickly we forget: The duration of persuasion effects from mass communication. Political Communication, 30(4), 521–547.CrossRefGoogle Scholar
  41. Hillygus, D. S., & Shields, T. G. (2008). The persuadable voter: Wedge issues in presidential campaigns. Princeton, NJ: Princeton University Press.CrossRefGoogle Scholar
  42. Hitchon, J. C., & Chang, C. (1995). Effects of gender schematic processing on the reception of political commercials for men and women candidates. Communication Research, 22(August), 430–458.CrossRefGoogle Scholar
  43. Hitchon, J. C., Chang, C., & Harris, R. (1997). Should women emote? Perceptual bias and opinion change in response to political ads for candidates of different genders. Political Communication, 14(January), 49–69.CrossRefGoogle Scholar
  44. Holbrook, T. M. (1996). Do campaigns matter?. Thousand Oaks, CA: Sage.Google Scholar
  45. Iyengar, S., & Simon, A. F. (2000). New perspectives and evidence on political communication and campaign effects. Annual Review of Psychology, 51, 149–169.CrossRefGoogle Scholar
  46. Jackson, R. A., & Carsey, T. M. (2007). U.S. Senate campaigns, negative advertising, and voter mobilization in the 1998 midterm election. Electoral Studies, 26(March), 180–195.CrossRefGoogle Scholar
  47. Kahn, K. F., & Kenney, P. J. (1999). The spectacle of U.S. Senate campaigns. Princeton, NJ: Princeton University Press.Google Scholar
  48. Kahn, K. F., & Kenney, P. J. (2004). No holds barred: Negativity in united states senate campaigns. Upper Saddle River, NJ: Pearson Prentice Hall.Google Scholar
  49. Kaid, L. L. (1997). Effects of television spots on images of Dole and Clinton. American Behavioral Scientist, 40(August), 1085–1094.CrossRefGoogle Scholar
  50. Kaid, L. L., & Boydston, J. (1987). An experimental study of the effectiveness of negative political advertisements. Communication Quarterly, 35(Spring), 193–201.CrossRefGoogle Scholar
  51. Kaid, L. L., Chanslor, M., & Hovind, M. (1992). The influence of program and commercial type on political advertising effectiveness. Journal of Broadcasting and Electronic Media, 36(Summer), 303–320.CrossRefGoogle Scholar
  52. King, J. D., & McConnell, J. B. (2003). The effect of negative campaign advertising tone on vote choice: The mediating influence of gender. Social Science Quarterly, 84(December), 843–857.CrossRefGoogle Scholar
  53. Krupnikov, Y. (2011). When does negativity demobilize? Tracing the conditional effect of negative campaigning on voter turnout. American Journal of Political Science, 55(October), 797–813.CrossRefGoogle Scholar
  54. Krupnikov, Y. (2012). Negative advertising and voter choice: The role of ads in candidate selection. Political Communication, 29(4).Google Scholar
  55. Krupnikov, Y. (2014). How negativity can increase and decrease voter turnout: The effect of timing. Political Communication, 31(3), 446–466.CrossRefGoogle Scholar
  56. Krupnikov, Y. (2015). Accentuating the negative: Candidate race and campaign strategy. Political Communication, 32(1), 152–173.CrossRefGoogle Scholar
  57. Lau, R., Sigelman, L., & Rovner, I. B. (2007). The effects of negative political advertisements: A meta-analytic reassessment. The Journal of Politics, 69(October), 1176–1209.CrossRefGoogle Scholar
  58. Lau, R. R. (1982). Negativity in political perception. Political Behavior, 4(4), 353–377.CrossRefGoogle Scholar
  59. Lau, R. R. (1985). Two explanations for negativity effects in political behavior. American Journal of Political Science, 29(February), 119–138.CrossRefGoogle Scholar
  60. Lemert, J. B., Wanta, W., & Lee, T.-T. (1999). Party identification and negative advertising in a U.S Senate election. Journal of Communication, 49(2), 123–134.CrossRefGoogle Scholar
  61. Lemert, J. B., Elliot, W. R., Bernstein, J. M., Rosenberg, W. L., & Nestvold, K. J. (1991). News verdicts, the debates, and presidential campaigns. New York: Praeger.Google Scholar
  62. Lenz, G. S. (2009). Learning and opinion change, not priming: Reconsidering the priming hypothesis. American Journal of Political Science, 53(October), 821–837.CrossRefGoogle Scholar
  63. Marcus, G. E., & MacKuen, M. B. (1993). Anxiety, enthusiasm and the vote: The emotional underpinnings of learning and involvement during presidential campaigns. American Political Science Review, 87(September), 672–685.CrossRefGoogle Scholar
  64. Martinez, M. D., & Delegal, T. (1990). The irrelevance of negative campaigns to political trust: Experimental and survey results. Political Communication and Persuasion, 7 (January/March), 25–40.Google Scholar
  65. Mattes, K., & Redlawsk, D. P. (2015). The positive case for negative campaigning. Chicago, IL: University of Chicago Press.CrossRefGoogle Scholar
  66. Matthew, D., & Dietz-Uhler, B. (1998). The black-sheep effect: How positive and negative advertisements affect voters’ perceptions of the sponsor of the advertisement. Journal of Applied Social Psychology, 28(20), 1903–1915.CrossRefGoogle Scholar
  67. Merritt, S. (1984). Negative political advertising: Some empirical findings. Journal of Advertising, 13(Fall), 27–38.CrossRefGoogle Scholar
  68. Min, Y. (2004). News coverage of negative political campaigns: An experiment of negative campaign effects on turnout and candidate preference. The International Journal of Press/Politics, 9(Fall), 95–111.CrossRefGoogle Scholar
  69. Peterson, D. A. M. (2004). Certainty or accessibility: Attitude strength in candidate evaluations. American Journal of Political Science, 48(3), 513–520.CrossRefGoogle Scholar
  70. Peterson, D. A. M. (2009). Campaign learning and vote determinants. American Journal of Political Science, 53(April), 445–460.CrossRefGoogle Scholar
  71. Pinkleton, B. E. (1998). Effects of print comparative political advertising on political decision-making and participation. Journal of Communication, 48(Autumn), 24–36.CrossRefGoogle Scholar
  72. Roddy, B. L., & Garramone, G. M. (1988). Appeals and strategies of negative political advertising. Journal of Broadcasting & Electronic Media, 32(May), 415–427.CrossRefGoogle Scholar
  73. Shapiro, M. A., & Rieger, R. H. (1992). Comparing positive and negative political advertising on radio. Journalism & Mass Communication Quarterly, 69(March), 135–145.CrossRefGoogle Scholar
  74. Shen, F., & Denis Wu, H. (2002). Effects of soft-money issue advertisements on candidate evaluation and voting preference: An exploration. Mass Communication and Society, 5(4), 395–410.CrossRefGoogle Scholar
  75. Skaperdas, S., & Grofman, B. (1995). Modeling negative campaigning. American Political Science Review, 89(1), 49–61.CrossRefGoogle Scholar
  76. Stimson, J. A. (1999). Public opinion in America: Moods, cycles, and swings (2nd ed.). Boulder, CO: Westview Press.Google Scholar
  77. Thurber, J. A., & Nelson, C. J. (2000). Campaign warriors: Political consultants in elections. Washington, DC: Brookings Institute.Google Scholar
  78. Vavreck, L. (2009). The message matters: The economy and presidential campaigns. Princeton and Oxford: Princeton University Press.CrossRefGoogle Scholar
  79. Weaver-Lariscy, R. A., & Tinkham, S. F. (1996). Advertising message strategies in U.S. congressional campaigns: 1982, 1990. Journal of Current Issues & Research in Advertising, 18(1), 53–66.CrossRefGoogle Scholar
  80. Windett, J. H. (2014). Gendered campaign strategies in U.S. elections. American Politics Research, 42(4), 628–655.CrossRefGoogle Scholar
  81. Wlezien, C., & Erikson, R. S. (2002). The timeline of presidential election campaigns. Journal of Politics, 64(4), 969–993.CrossRefGoogle Scholar
  82. Wooldridge, J. M. (2000). Introductory econometrics: A modern approach. South-Western College.Google Scholar

Copyright information

© Springer Science+Business Media New York 2016

Authors and Affiliations

  1. 1.University of Nevada, RenoRenoUSA
  2. 2.Saint Louis UniversitySt. LouisUSA

Personalised recommendations