theeffectsof negativepoliticalcampaigns: ameta...

The Effects of Negative Political Campaigns:A Meta-Analytic Reassessment

Richard R. Lau Rutgers UniversityLee Sigelman The George Washington UniversityIvy Brown Rovner Rutgers University

The conventional wisdom about negative political campaigning holds that it works, i.e., it has the consequences itspractitioners intend. Many observers also fear that negative campaigning has unintended but detrimental effects onthe political system itself. An earlier meta-analytic assessment of the relevant literature found no reliable evidence forthese claims, but since then the research literature has more than doubled in size and has greatly improved in quality.We reexamine this literature and find that the major conclusions from the earlier meta-analysis still hold. All told,the research literature does not bear out the idea that negative campaigning is an effective means of winning votes,even though it tends to be more memorable and stimulate knowledge about the campaign. Nor is there any reliableevidence that negative campaigning depresses voter turnout, though it does slightly lower feelings of political efficacy,trust in government, and possibly overall public mood.

Conventional wisdom among political consult-ants, candidates, and consultants during the1980s and 1990s held that election campaigns

had become increasingly mean-spirited and that thepervasive negativism of campaigns was exacting aheavy toll on American democracy, undermining citi-zens’ positive feelings about elections in particular andgovernment in general and thereby demobilizingpotential voters. Negative campaigning had come todominate American politics, it was believed, because itworks; that is, candidates who go on the attack usuallysee their ratings rise and reap greater support on Elec-tion Day than they would have gotten had they stayedpositive. The 1988 presidential campaign, whenGeorge H.W. Bush came from 10 points behind in thepolls to a comfortable victory after the (in)famousWillie Horton, Boston Harbor, and Dukakis-in-a-tankads began airing, has been offered as the poster child ofeffective attack politics.1

Lau et al.’s (1999) meta-analysis of social scienceresearch on the effects of negative campaigning foundlittle hard evidence for these claims. Even though Lauet al.’s findings were widely publicized, it would benaive to expect results reported in a scholarly journalto have an immediate or substantial impact on whatpolitical strategists recommend, what political candi-dates do, and what political commentators believe.Still, one might have hoped for a gradual but growingawareness of these new findings among those whosebusiness is running campaigns or covering them,a glimmer of skepticism toward previously unchal-lenged beliefs, or a revamping of standard operatingprocedures. Insofar as we can determine, though, littleof this has happened. Despite changes in campaignfinance laws designed to reduce the negativism ofcampaigns, the most recent federal elections are beingcalled the most negative—by far—on record.2 Forexample, it has been reported that whereas only 1%

1Geer (2006) has debunked the conventional view of the 1988 campaign, demonstrating that Dukakis attacked Bush as early and often asBush attacked him, and more importantly that Bush had overcome the challenger’s early lead in the polls before any of this negativecampaigning began.

2See, e.g., May (2006), or Tucker (2006). During the 2006 campaign alone, hundreds of newspaper articles focused on negative campaign-ing, many with similar “this is the worst ever” themes.

The Journal of Politics, Vol. 69, No. 4, November 2007, pp. 1176–1209

© 2007 Southern Political Science Association ISSN 0022-3816

1176

and 46%, respectively, of the ads sponsored by theDemocratic and Republican Congressional CampaignCommittees in 2004 were negative, in 2006 thosefigures skyrocketed to 83% and 89% (CQ Weekly,October 16, 2006).3

Far and away the commonest explanation for thiswidespread and apparently growing negativism ofcampaigns is the presumed effectiveness of attackpolitics, even as it is simultaneously decried as a cor-rosive influence on the American system of govern-ment. For one thing, negative ads are believed to drawattention:

“Voters don’t pay much attention to campaign ads,”claims Bob Stern from the Center for GovernmentStudies in California, “but when they’re negative theydo. . . . That’s why negative ads are busting out all over—they can cut through the flotsam of an election-year blitz;they tend to stick with us when less provocative ads fadeaway; and they often provide voters with usable informa-tion about candidates they know next to nothing about.”(May 2006)

For another, negative campaigning is believed to beadvantageous to the attacker:

Ugly, combative, negative advertising targeting a politicalopponent works. You can see your opponent’s favorablepolling numbers degrade while the negative ad runs.(Richard Romero, former Democratic president pro temof the New Mexico State Senate; quoted by Quigley 2006)

If positive advertisements moved things to the extent thatnegative ads move things, there would be more of them(Rep. Thomas Reynolds, chairman of the NationalRepublican Congressional Committee; quoted byNagourney 2006).

Warnings that these negative political advertisementsare undermining American democracy persist, too.Brooks (2006) conducted a systematic study of 186newspaper and magazine articles linking negativeadvertising and turnout from 2000 through 2005, andreports that 65% of the articles concluded thatnegative campaigning depresses turnout, while only

6% concluded that it might increase turnout. Forexample, a recent Washington Post columnist charac-terizes the research literature as “show[ing] that nega-tive ads can reduce turnout; Democrats hope aconstant drumbeat of scandal, Iraq and ‘stay thecourse’ will persuade conservatives to stay home ofNov. 7. . . . Republicans . . . are equally eager todepress Democratic turnout and fire up their conser-vative base” (Grunwald 2006). Similarly, political sci-entist Thomas Patterson claims that “numerousstudies show that misleading negative ads corrodetrust in democracy” (quoted by Christopher Shea inThe Boston Globe, May 21, 2006).

None of these conclusions was supported in ourearlier meta-analysis of research on the effects ofnegative campaigning (Lau et al. 1999). Why the dis-connect between the evidence in the social science lit-erature and the actual beliefs and practices ofcandidates, consultants, pundits, and even manypolitical scientists? One possibility, which we explorehere, is that the conclusions that we drew in ourearlier study were simply incorrect. After all, the poli-ticians who approve of negative ads and the consult-ants who recommend and produce them have toomuch at stake and are paid too much to be mistaken.Moreover, the research literature itself has changed intwo important ways. In less than eight years it morethan doubled in size, mushrooming from 52 studiescontaining 123 pertinent findings in late 1998 to 111studies containing 294 pertinent findings by mid-2006. Less obviously but no less importantly, thisrapid growth was accompanied by an equally markedincrease in methodological rigor, particularly inanalyses of the effects of actual political campaigns.The question then becomes whether the quantitativeand qualitative growth of the research literature hasinvalidated our earlier conclusions. Were our conclu-sions time-bound and premature and are they nolonger operative—if, indeed, they ever were operative?Is the persisting conventional wisdom about negativecampaigning correct after all? Fortunately, the tre-mendous growth and enhanced quality of research onnegative campaigning now enable us to launch a morecomprehensive and reliable assessment of the conven-tional wisdom than was feasible at the time of ourearlier study.

Method

To address these questions, we have conducted anew meta-analysis—a quantitative synthesis of

3These records go back only to 2004, when the Federal ElectionCommission began requiring all independent (that is,noncandidate-sponsored) groups to report whether the primarypurpose of their political expenditures was to support or oppose acandidate. Recent changes in campaign laws now require all can-didates for federal office to appear in their own ads and to statethat they approve of the message being presented. This require-ment may have reduced the negativism of candidate-sponsoredads while shifting attacks to ads sponsored by the parties and otherindependent groups. We know of no solid evidence on whether thetotal negativism of the 2006 campaign, counting both candidate-sponsored and supposedly independent party and PAC advertise-ments, was significantly greater than previous years.

the effects of negative political campaigns: a meta-analytic reassessment 1177

findings from independent studies—of research onthe effects of negative campaigning. Taking the find-ings themselves (rather than the raw data upon whichthey are based) as the basic data for analysis, we haveset out to estimate the consistency and magnitude offindings about the effects of negative campaigning andto account for variability among these findings. Foursteps defined this undertaking: (1) defining the typesof findings to be analyzed; (2) locating published andunpublished studies of the effects of negative cam-paigning; (3) translating the findings of these studiesinto a common metric to facilitate comparisons acrossstudies; and (4) synthesizing the assembled findings.

Defining the Pertinent Findings

Given our focus is on the effects or consequences ofnegative political campaigns, we honed in on researchon both actual and hypothetical political settings inwhich candidates or parties vied for electoral support.This stipulation led us to include studies of specificinstances of negative campaigning (e.g., a single tele-vision advertisement) as well as characterizations ofentire campaigns.4 It led us to exclude studies of nega-tive campaigns in nonpolitical settings (e.g., productadvertising) and studies of the effects of negativedescriptions of various stimuli (even political figures)in nonelectoral settings (e.g., news of wrong doing bya prominent political figure). Definitions of negativecampaigning vary from study to study, and we gener-ally relied on the definition employed in a given studyrather than trying to fit studies into a definition of ourown making. To qualify for inclusion here, an actual orhypothetical campaign setting had to feature variabil-ity in tone, so that every research finding consideredhere is based on comparison of negative ads or cam-paigns to positive, neutral, or at least less negative adsor campaigns.5

Research findings about the effects of negativecampaigning on the memorability of campaigns or

some aspect thereof and on the interest or knowledgethat such campaigns evoke constituted the initial focusof our meta-analysis. These are intermediate effectsthat could help explain the “bottom-line” conse-quences of negative campaigns, which constituted ourmain focus. The latter are of two types. Direct electoraleffects include affect toward the target of negative cam-paigning (that is, the opposing candidate), affect forthe candidate on whose behalf a negative campaign iswaged (that is, the attacker), relative affect for theattacker compared to the target, and the intention orprobability of voting for the attacker. Irrespective of itsimpact on the outcome of a particular election, nega-tive campaigning could have broader systemic conse-quences. Stemming from the common suppositionthat negative campaigning turns people off on politics,this category of effects includes, most prominently,actual or intended voter turnout, along with citizens’sense of political efficacy, trust in government, andoverall political mood.

Locating Studies

Our next task was to identify and access every perti-nent published or unpublished study within thedomain. To avoid biases induced by the “file drawer”problem (the tendency for nonsignificant results to berelegated to the investigator’s files; see, e.g., Rosenthal1979), this step must be comprehensive. We beganwith the studies that Lau et al. (1999) had analyzedand added papers and articles we had identified andcollected by mid-2005. Then, in June of 2005, we con-tacted approximately 275 scholars with interests evenremotely related to negative advertising or campaign-ing, describing our project, stressing the need for com-prehensiveness, requesting references to or copies ofnew studies they may have conducted, and asking forleads to other researchers who, unbeknownst to us,may have conducted potentially relevant research.These contacts yielded dozens of new studies. Finally,to all the studies we had gathered by these means weadded a new round of papers from professional meet-ings held during the following year. Guiding ourefforts was the goal of obtaining every relevant pieceof research completed by mid-2006. We believe thatwe achieved that goal within a very small margin oferror, yielding the aforementioned total of 111 studies.Thumbnail descriptions of these studies are presentedin Table 1A in the appendix.6

4In our earlier study, we stated that our focus was on the effects ofnegative political advertising, but our review included manystudies in which the focal variable was a characterization of anentire campaign rather than an advertisement or a series thereof.Thus we are not increasing the number of relevant studies here viadefinitional expansion; we are simply describing the criteria forinclusion more accurately.

5A few studies have manipulated the nature of attacks—forexample, contrasting issue-based attacks to character-basedattacks without including any positive ad conditions (e.g., Pfauand Burgoon 1989). We included such findings in subsidiaryanalyses of the relative effectiveness of issue- and character-basedads, but excluded them from analyses of the effects of negativeadvertisements per se.

6A few of the studies that we had in hand by the end of 2006 weresubsequently published in 2007. The table reflects their publica-tion dates in cases where we are aware of them.

1178 richard r. lau, lee sigelman, and ivy brown rovner

At least two of the three authors examined eachstudy to determine whether it met our criteria forinclusion. In several instances we replaced findingsthat Lau et al. (1999) had included because theresearchers had subsequently reanalyzed the samedata. For example, we replaced findings from Ansola-behere et al.’s (1994) analysis of data on the 1992Senate elections with findings from Ansolabehere,Iyengar, and Simon’s (1999) later analysis of thesedata. We did not “double-count” findings that aresearch team had presented in multiple venues (e.g.,a convention paper and a published article), but wedid include a few independent analyses of the samedata by different research teams; for example, wetreated as independent Brooks’s (2006) and Watten-berg and Brians’s (1999) reanalyses of the same 1992Senate elections that Ansolabehere et al. (1994) hadanalyzed. We will address the consequences of thesedecisions in the section on refinements and modera-tor effects.

Calculating (and Adjusting) Effect Sizes

Once we had accessed relevant studies, we had totranslate their findings into a common metric.Several decades ago, statisticians developed various“vote counting” methods of combining results fromindependent studies (e.g., Fisher 1932; Mosteller andBush 1954). These methods merely test the nullhypothesis that the research literature on somehypothesized effect does not contain even a singlesignificant finding (Becker 1994). Such tests obvi-ously are not very discriminating, but because theyprovide a rock-bottom indication of whether there isany evidence at all on behalf of a hypothesized effect,we applied them here.

More recently, techniques have been developedfor gauging and aggregating the magnitude of everyreported effect, enabling more discriminatinghypothesis tests. Two major types of effect sizes havecome into widespread use, d measures based onmean differences between groups and r measuresbased on correlations. In our earlier study weemployed d because it is a natural metric for experi-mental studies, which constituted the majority of thestudies then available. To maintain continuity, we didthe same here.

For a simple design with experimental and controlgroups, d represents the difference between the meansof the two groups, expressed in standard deviationunits. For more complex factorial designs, the calcu-lation of d becomes more complicated but the basic

idea remains the same.7 However, for many researchstudies no widely accepted effect size measure exists,and particularly vexing problems arise when, as here,findings from nonexperimental multivariate analysesare to be combined with results from randomizedexperiments. For data from multivariate analyses, Lauet al. (1999) used the formula d = 2t/√df, where t refersto the t statistic and df refers to the degrees of freedomassociated with this t-test (Stanley and Jarrell 1989).This formula, which Rosenthal and Rubin (2003) call“dequivalent” represents what the effect size would havebeen if t had been obtained from a comparison ofmeans in an analysis involving two groups and a nor-mally distributed dependent variable. We followedRosenthal and Rubin’s recommendation that dequivalent

be used when only probability or t values are reported,or when no widely accepted method exists for calcu-lating effect sizes for the analytic procedures used in astudy.

After effect sizes have been calculated, they mustbe combined. Analyzing raw effect sizes equates effectsizes based on different sample sizes. Given the widevariation in sample sizes across studies, we instituted a“barebones” adjustment for sampling error-standardpractice in meta-analyses. This adjustment weightseach finding by the number of cases on which afinding is based, divided by the total number of casesin studies with pertinent findings. Beyond that, weadjusted for measurement reliability in the dependentvariable (Hunter and Schmidt 1990).8 We reportbelow the unadjusted, sampling error-adjusted, andreliability-adjusted effect sizes.9

7Formulas for calculating d for different research designs and sta-tistical tests can be found in any text on meta-analysis—see forexample Cooper and Hedges (1994), Glass, McGaw, and Smith(1981), Hunter and Schmidt (1990), or Rosenthal (1984).

8Lau et al. (1999) describe these adjustments in greater detail.

9Adjusting for reliability of measurement is obviously problematicfor studies in which no reliability measure is reported, as was thecase for most of the studies considered here. When reliabilityinformation was missing but the dependent measure was a multi-item scale, we assumed the mean reliability reported for that typeof dependent variable (e.g., one average reliability for affect for theattacker, another for affect for the target, and so on). When reli-ability information was missing and the dependent variable was asingle item, we used two-thirds of the mean reliability for that typeof dependent variable instead. For official vote totals (relevant tofindings concerning vote outcomes and turnout), we set reliabilityat .98. Although these adjustments were somewhat arbitrary, weconsider them reasonable. Adjusting for measurement unreliabil-ity necessarily increases the absolute value of the estimated effectsize, but it also increases the standard error of the estimate.


Observing Central Tendencies andExplaining Variance in the Findings

Once relevant studies have been located and theirfindings translated into a common metric, they mustbe summarized statistically and tested for moderatoreffects—factors that could explain why an effect ispresent or large or positive in some studies but absentor small or negative in others. For example, we mightexpect the estimated effects of negative campaigningto be larger in experimental than field studies becausethe latter may involve so much noise that it becomesdifficult to detect any reliable effects; or we mightexpect attacking to be effective for challengers butcounterproductive for incumbents, in which caselumping all the effect sizes together would mislead-ingly produce support for the null hypothesis. Becausea few of the studies considered here had extremelylarge sample sizes, we also conducted a series ofrobustness analyses in which we minimized theimpact of these few studies. After determining thebasic finding for each dependent variable, we searchedfor such moderator effects whenever enough caseswere available to sustain subgroup analyses.

Results

Intermediate Effects

By mid-2006, the research literature contained 21findings on the memorability of negative ads or cam-paigns. In the vote-count test, the null hypothesis thatthe literature contains no statistically significantresults at all could easily be rejected (z = 8.5, p <. 001),indicating that the literature contains at least somesignificant evidence (see Table 1). More tellingly, theunadjusted effect size across the 21 findings averaged.28, in the moderate range but not reliably greater thanzero. Adjusting for sampling error left the estimatedmean effect untouched but greatly reduced the stan-dard error. Because both the largest positive and thelargest negative effects are from small-sample (pre-dominantly experimental) studies, this adjustmentgreatly reduced their influence on the estimatedsample variance and lifted the 95% confidence intervalabove zero. Adjusting for measurement unreliabilityleft the effect size in place but raised its standard errorback up to its unadjusted level. The overall message ofthe research literature concerning memorability, then,is that negative ads and campaigns are somewhateasier to remember than comparable positive ads andcampaigns, but because the statistical significance of

the effects varies according to various adjustments wemade, these differences are neither strong nor consis-tent. The effect sizes for memorability adjusted forsampling error and measurement unreliability areshown in the first panel of the top row of Figure 1.

Ten studies have examined the effects of negativecampaigning on campaign interest. Two of these (bothconducted by Pinkleton) reported moderately largeincreases in campaign interest, but four of the tenfound effects in the opposite direction, and overall wecould not reject the null hypothesis that none of thereported studies found any significant effects. Theeffects of negative campaigning on campaign-relatedknowledge are more consistent. Eleven of the fifteenpertinent studies reported positive effects—that is,negative campaigns increased campaign knowledge—and we could easily reject the null hypothesis thatnone of the 15 effects is significant. However, thiseffect is small, featuring unadjusted and adjustedmean effect sizes in the single digits; across the 15pertinent findings, the average boost in campaignknowledge associated with negative campaigningamounts to less than one-tenth of a standard devia-tion. The consistency and modest magnitude of theseeffects is easy to see in the rightmost entry in the firstrow of Figure 1. (We maintain the same scale in allpanels of Figure 1 to give a clear idea of the relativemagnitude of effects across dependent variables.)More broadly, although the findings that have accumu-lated in the research literature are consistent with theidea that several (but not all) of the mechanisms bywhich negative campaigning is presumed to shape affectfor candidates are indeed operative, the observed differ-ences are not strong.

Do Negative Campaigns Work?

Because the mechanisms through which negativecampaigning is supposed to work are operative, albeitto a modest extent, the next question is whether goingnegative itself works. Negative campaigns aredesigned, first and foremost, to diminish positiveaffect for their target, the opposing candidate.Although proponents of negative campaigning recog-nize that it may simultaneously produce lower affectfor the attacker (the so-called “backlash” effect: see forexample Roese and Sande 1993), they contend that thenet effect, as reflected in differential candidate affect,should work to the attacker’s advantage. For candidateaffect to matter, it must translate into the choicebetween candidates, and here again attacking isbelieved to favor the attacker. Lau et al. (1999) uncov-ered empirical support for the ideas that negative cam-


TA

BL

E1

Met

a-A

nal

ysis

Sum

mar

y

Eff

ect

Nu

mb

erof

stu

die

s

Sub

ject

s

Test

ofco

mb

ined

sign

ifica

nce

(Z)

Un

adju

sted

Cor

rect

edfo

rsa

mp

lin

ger

ror

Cor

rect

edfo

rm

easu

rem

ent

erro

r

Cu

rren

tL

auet

al.

Eff

ect

size

s.e.

Eff

ect

size

s.e.

Eff

ect

size

s.e.

Dir

ect

elec

tora

leff

ects

(cod

edsu

chth

atin

crea

sed

neg

ativ

ism

hel

psth

eat

tack

er)

1.M

emor

yfo

rad

21(1

4)9,

438

8.5*

**.2

8.1

7.2

8**

.10

.35*

*.1

22.

Cam

paig

nin

tere

st10

(0)

47,7

731.

5.0

7.0

5.0

0.0

1.0

0.0

13.

Cam

paig

nkn

owle

dge

15(3

)15

,947

2.5*

.05

.02

.06*

*.0

2.0

7**

.02

4.A

ffec

tfo

rth

eta

rget

31(1

6)18

,341

10.7

***

.29*

*.0

9.1

1.0

6.1

4*.0

75.

Aff

ect

for

the

atta

cker

40(2

5)20

,146

-15.

6***

-.37

**.0

9-.

16*

.06

-.21

*.0

86.

Net

affe

ct10

(6)

19,8

73-6

.7**

*-.

28.4

6-.

04.1

4-.

13.2

77.

Vot

ein

ten

tion

27(1

3)28

,904

-1.2

-.03

.14

-.01

.06

-.02

.08

8.A

ctu

alvo

tech

oice

16(5

)11

,898

-2.9

**-.

14*

.06

-.03

.02

-.04

.03

Syst

emic

effe

cts

(cod

edsu

chth

atin

crea

sed

neg

ativ

ism

isas

soci

ated

wit

hin

crea

sed

turn

out,

effi

cacy

,tru

st,a

nd

soon

)9.

Inte

nde

dtu

rnou

t21

(4)

41,8

26-3

.0**

-.08

.05

-.01

.01

-.02

.02

10.

Act

ual

turn

out

36(1

5)16

3,37

03.

0**

-.01

.05

.03*

.01

.03*

.01

11.

Polit

ical

effi

cacy

21(4

)22

,516

-2.8

**-.

05**

.02

-.03

***

.00

-.04

***

.01

12.

Tru

stin

gove

rnm

ent

11(2

)10

,134

-2.8

*-.

07*

.03

-.04

**.0

1-.

07**

*.0

213

.P

ubl

icm

ood/

Aff

ect

for

gove

rnm

ent

10(2

)5,

334

-3.2

**-.

15.1

1-.

03.0

5-.

04.0

6

*p<

.05;

**p

<.0

1;**

*p<

.001

.


paigning does drive affect for the target of attacksdown but also lessens affect for the attacker. Is thatconclusion borne out in the greatly expanded researchliterature considered here? If it is, which of these twoeffects takes precedence?

The literature now contains 31 findings concern-ing affect for targets of attacks. (See the first panel ofthe second row of Figure 1.) In 22 of these tests—morethan two out of every three—the expected decline inaffect for the target occurs. To be sure, not all of theseeffects are statistically significant, let alone powerful,but the vote-counting null hypothesis that none ofthem is significant could easily be rejected (z = 10.7,p < .001). The mean unadjusted effect size is .29—reliably greater than zero and moderately large.However, adjusting for sampling error reduced theeffect size to .11, with a 95% confidence interval nowextending slightly beyond zero. This reduction reflectsthe fact that many of the largest effect sizes again comefrom studies with relatively few subjects, whereas thesmaller or negative effect sizes tend to come from

studies with much larger samples. Adjusting for mea-surement unreliability produced an estimate of .14—just reaching the .05 significance level and indicative ofa relatively small effect. Overall, then, the picture ismixed, with the bulk of the evidence pointing to amodest tendency for negative campaigns to under-mine positive affect for the candidates they target.

The other side of the coin is backlash against theattacker, the subject of 40 reported findings. Contraryto what attackers would prefer, 33 of these 40 findingsare negative, indicating a decrease in affect for attack-ers. (See the middle panel of the second row ofFigure 1.) Again the null hypothesis that none of theseeffects is significant could easily be rejected, and themean unadjusted effect size (-.37) is farther from zerothan the corresponding decrease in affect for thetarget. Correcting for sampling error and measure-ment error reduces the mean effect size considerablybut it remains statistically significant. Because theresults for targets and attackers are not necessarilypaired from the same studies, they do not directly

FIGURE 1 Effect Sizes Adjusted for Sampling Error and Measurement Unreliability, by Type of DependentVariable

Memory Campaign Knowledge/Interest

Affect Toward the Attacker Affect Toward the Target

The Vote Choice Turnout Political Efficacy/Trust/Public Mood

Net Affect (Attacker–Target)


gauge the net effect of going negative on affect forattackers and their targets, but they do sound an initialnote of caution about the validity of claims about thebenefits of going negative.

Evidence bearing directly on the question ofwhether attacks undermine affect for their targetsmore than for the attackers themselves is in surpris-ingly short supply, having been reported in only 10studies. Three of the unadjusted effects are moder-ately large and positive, indicating a net differentialin favor of the attacker; one is very large and nega-tive, indicating a backfire on the attacker; and theremaining six all cluster closely around zero. (See thelast panel of the second row of Figure 1.) The nullhypothesis that the research literature contains nosignificant findings at all could again be easilyrejected, but this is due entirely to the one very largenegative finding, which of course runs counter to theprevailing conventional wisdom. Overall, the unad-justed (-.28) and adjusted (-.04 and -.13) meaneffects are negative, though the standard errors are sosizeable that they do not even approach statisticalsignificance. Accordingly, the findings reported in theresearch literature do not bear out the proposition thatattacking is an effective way to bolster one’s own imagerelative to that of one’s opponent. Having said that, wehasten to caution against reading the opposite impli-cation into this conclusion. Although the evidencepoints in the direction of a net backfire againstattackers, it does not do so decisively enough tosupport the conclusion that attacks exact a signifi-cantly greater toll on attackers than on their targets.Rather, the research literature provides no determi-native resolution one way or the other insofar asaffect for the competing candidates is concerned.

Ultimately, how much the voters like the candi-dates matters to the candidates only to the extent thatit helps or hurts their chances of being elected. Weuncovered 43 relevant findings in the literature, 27involving intended vote choices and 16 involvingreported vote choices or official vote totals. As can beseen in the bottom left-hand corner of Figure 1, only12 of these outcomes are positive from the attacker’sstandpoint. Only five of these 12 are at all appreciable,and they are counterbalanced by five negative effectsof similar magnitude. The remaining effects are allsmall. The overall unadjusted effect of negative cam-paigns on vote choice indicates a modest disadvantageto the attacker, but that effect vanished when weadjusted the effect sizes. It bears mentioning that thetwo largest negative effects and all five of the positiveones come from experimental studies in which thedependent variable is vote intention; that particular

design evidently produces more volatile outcomes.10

The broader message, though, is that the research lit-erature does not bear out the proposition that negativepolitical campaigns “work” in shifting votes toward thosewho wage them.

Overall, then, social science research providessome evidence that the mechanisms through whichnegative campaigning is supposed to work do in factoperate, but there is an overriding lack of evidencethat negative campaigning itself works as it is sup-posed to. Intriguingly, the conclusion that negativecampaigning is no more effective than positive cam-paigning holds even though negative campaignsappear to be somewhat more memorable and togenerate somewhat greater campaign-relevantknowledge.

Do Negative Campaigns Harm thePolitical System?

Aside from any immediate impact that it may or maynot have on the candidates and electoral outcomes,negative campaigning could have consequences—according to the conventional wisdom, dire ones—forthe political system itself. By far the best knownexample of this possibility is the demobilizationhypothesis, which holds that negative campaigningalienates many potential voters from politics ingeneral and from electoral politics in particular.Ansolabehere et al.’s (1994) estimate of a 5% drop-offin turnout due to negative campaigning generatedwidespread concern and sparked an explosion offollow-up research. We identified 57 different tests ofthe demobilization hypothesis in the last dozen years,two-thirds of which postdate our previous meta-analysis.

The effect sizes for voter turnout, coded to benegative if negative campaigning depresses turnout,vary widely. A few large positive effects can be seen inthe middle panel of the bottom row of Figure 1, alongwith a few large negative ones and many that hoveraround zero. The overall findings pertaining tointended turnout are negative while the overall find-ings for actual turnout are positive, and if we combinethese findings (as in Figure 1) we cannot even rejectthe null hypothesis that none of the effects is statisti-cally significant. Several of these studies do report sta-tistically significant results, but some are positive and

10We discuss in the text and illustrate in Figure 1 the results of all43 studies together, but to provide added perspective we give theresults for vote intention and actual vote choice separately inTable 1.


others negative. The null effect from the combinedsignificance test suggests that we are sampling from anoisy distribution centered at 0, from which the sig-nificant positive and negative effects we have observedamount to random draws. For intended turnout, theunadjusted and adjusted effect sizes are consistent indirection with the demobilization hypothesis butnever significant. For actual turnout, the adjustedmean effect sizes run counter to the demobilizationhypothesis but are too small to be of practical conse-quence. When all 57 findings are combined into asingle analysis, the mean unadjusted effect is -.07 (ns).Adjusting for sampling error and measurement reli-ability causes the mean to turn very slightly positive(.02), again not significantly different from 0. Itfollows that the research literature provides no generalsupport for the hypothesis that negative political cam-paigning depresses voter turnout. If anything, negativecampaigning more frequently appears to have a slightmobilizing effect.

This result bears out our earlier conclusion,though it is now based on three times as many studies.However, decreasing turnout is only one way thatnegative campaigning could adversely affect the politi-cal system. It could also undermine system-supportingattitudes, darken the public’s general mood (Rahn andHirshorn, 1995), or even diminish satisfaction withthe government itself. In our earlier study, we couldnot locate enough pertinent findings to test any ofthese hypotheses, but the expansion of the literature inrecent years makes it possible to do so now. Thebottom right-hand corner of Figure 1 displays theeffects of negative political campaigning on feelings ofpolitical efficacy (21 studies), trust in government (11studies), and public mood/satisfaction (10 studies),with positive effect sizes denoting favorable conse-quences for the political system. The effects are, asFigure 1 reveals, overwhelmingly negative—not largebut very consistent, and statistically significant forboth political efficacy and trust in government. Thus,the conclusion to be drawn from the literature is clear:Negative campaigning has the potential to do damage tothe political system itself, as it tends to reduce feelings ofpolitical efficacy, trust in government, and perhaps evensatisfaction with government itself.

Further Refinements and PotentialModerator Effects

The conclusions we have drawn to this point havebeen based on analyses of all relevant results reportedin the research literature, without distinctions amongstudies except for those involved in adjusting effect

sizes. Some readers may suspect that the results of ourmeta-analysis are artifacts of our “double counting” ofsome findings reported by separate analyses of thesame data or of the “swamp effect” induced by thepresence of a few studies with extremely large samplesizes. To determine whether this might be the case, weconducted a series of additional analyses. First, weaveraged the effect sizes reported by different investi-gators who were analyzing the same data, and reana-lyzed the effect sizes with each pair of results treated asa single result. The greatest potential effect of doublecounting would revolve around the data reported inAnsolabehere et al.’s (1994) article on the demobiliza-tion effect, which have subsequently been reanalyzedby several different research teams. However, when wetreated these several results as a single one, the unad-justed mean shifted only from .01 to .02 and theadjusted means remained the same to two decimalplaces.

To see whether the results from a few studies withvery large sample sizes were swamping the remainingresults, we sorted the data by sample size and thenreduced any reported sample size that was more thantwice as large as the next one in order. In mostinstances we achieved this reduction by dividing theactual sample size by 2 or 3, but in one case we reduceda sample size by a factor of 10. These adjustmentscould only reduce the impact of findings from studieswith extremely large samples. (They also increased theestimated standard errors by reducing the degrees offreedom upon which the standard errors were esti-mated.) Even so, the resulting mean effect size shiftedby more than �.03 only for differential affect, forwhich the sampling error-corrected effect size grewfrom -.04 to -.16; at the same time, the estimatedstandard error more than doubled, so the increasedeffect size was still not significantly different fromzero. Our basic findings, then, appear to be very robustto validity threats posed by double counting and theswamp effect.

Although the evidence we have examined so farpoints away from the conclusions that negative cam-paigning works and that it bears some responsibilityfor declining voter turnout, it could still be the casethat some factor or factors we have not yet consideredmight moderate the suspected effects and observednoneffects of negative campaigning. Some types ofnegative campaigning might be particularly effective(for example, issue-based attacks) and others (per-sonal attacks) less so; some types of voters might beinfluenced by attacks (the less politically sophisticated,or political independents) and others not; and sometypes of research designs might be better suited than


others for detecting the effects in question. Search-ing for such moderator effects involves conductingsubgroup analyses, for which a very sizeable researchliterature is a prerequisite. Despite the rapid growthof research on negative campaigning, the numberof findings bearing on some of its supposed effectsremains relatively small. Thus, we could not look formoderator effects for all the purported consequencesof negative campaigning, but we could do so for affectfor the target of attacks, affect for the attacker, votechoice, voter turnout, and broader systemic effects(the latter by combining findings for trust in govern-ment, political efficacy, and public mood).

In each of these instances, we tested for potentialmoderating effects of: (a) the particular stimulus con-trast (negative versus positive, negative versus neutral,or contrast versus positive) on which a finding wasbased; (b) the number of ads to which subjects wereexposed; (c) the type of research design (experimental,survey-based, or aggregate analysis); (d) the type ofsubjects (undergraduates, a sample of the generalpublic, or aggregate data); the use of real or simulated(e) candidates and (f) ads; (g) the medium throughwhich the ads were delivered (written, audio, orvideo); and even (h) the academic discipline of theresearchers (communications, political science, orpsychology). It would be pointless to detail the resultsof these analyses, for none of these potential modera-tors displayed any consistent effect on affect for thetarget or the attacker, vote choice, or turnout.11 Unsur-prisingly, the more negative ads people were exposedto, the more their efficacy, trust, and satisfactiondeclined, and—as we noted above—experimentsuncovered more negative effects than did surveys.

We also searched for moderator effects specific toparticular electoral or systemic outcomes, but againwithout much success. For example, we found no evi-dence that an attacker’s party affiliation had anybearing on the success of the attack. Nor, more sur-prisingly, was there any indication that going negativeworks better for challengers or open-seat candidatesthan it does for incumbents. We did find some supportfor the idea that whereas negative campaigns stimulatepartisans to get out and vote, they are more likely to

turn independents off on voting; however, too fewstudies (only nine) were involved in this hypothesistest to achieve the power necessary for conventionallevels of statistical significance (t = 1.6, p < .07,one-tailed).

Discussion

The idea that negative political campaigns work isgenerally taken as both a truism and a source of regret(but see Geer, 2006). The campaign ads that live onin memory—ranging from the Willie Horton ads of1988 through the Swift Boat Veterans for Truthads of 2004—are widely seen as having achievedtheir intended result of making the opposing candi-date seem duplicitous or even dangerous. Theseand other dramatic cases are routinely cited as proofpositive of the power of negative campaigning. As areporter recently summarized the prevailing viewamong political professionals and commentators:“The people who produce these ads and the consult-ants who hire them know that negative campaign-ing works. These people are paid way too muchto be mistaken about whether poison is effective”(Mansnerus 2005).

On the other hand, many counterexamples—instances in which going negative did not prevent, oreven contributed to, the loss of a campaign—couldalso be told, though they rarely are. For example, thesame consultants who had produced the “successful”Swift Boat Veterans ads also produced, in the 2005New Jersey gubernatorial election, an ad in whichthe Democratic candidate was criticized by hisex-wife—an attack that backfired on the Republicancandidate and helped turn a close election into arunaway victory for his Democratic opponent(Whelan and Margolin 2005).

The popularity of certain views, the salaries ofthose who hold them, and the availability of a fewdramatic examples do not constitute convincing evi-dence. To state the matter bluntly: There is no consis-tent evidence in the research literature that negativepolitical campaigning “works” in achieving the electoralresults that attackers desire. Although attacks probablydo undermine evaluations of the candidates theytarget, they usually bring evaluations of the attackersdown even more, and the net effect on vote choice isnil. Nor have we uncovered evidence that negativecampaigning tends to demobilize the electorate. A fewstudies have reported significant demobilizing effects,a few have reported significant mobilizing effects, andthe great majority have reported almost no effect one

11There was a significant effect of the realism of the candidates onthe magnitude of the effect sizes observed for affect toward theattacker and the target of negative campaigns, and on turnout, butonly for the unadjusted effect sizes. In the first two cases, signifi-cantly smaller (in absolute value) effects were observed for realcandidates than for mock candidates created specifically for anexperiment. For turnout, whereas studies with real candidateswere more likely to produce positive effects, studies with artificialcandidates were more likely to produce demobilizing effects.


way or the other; the overall mean effect is approxi-mately zero. Negative campaigning does, however,have some negative systemic consequences, includinglower trust in government, a lessened sense of politicalefficacy, and possibly a darker public mood. Althoughthe latter effects are not large, and may be due more tocoverage of negative ads in the media rather than theads themselves (see Geer 2006), in the long run theycould prove worrisome. We hope that more research-ers will explore the effects of negative campaigns onthese system-supporting attitudes, because the exist-ing evidence seems fairly promising and the long-runimplications seem fairly alarming.

Skeptics might worry that the negative ads andcampaigns that have been considered in the researchliterature, the contexts in which they have beenstudied, and the effects that have been documentedmay be too unusual, artificial, weak, or poorly instru-mented to have enabled the researchers on whosework our meta-analysis is based to detect conse-quences that under more auspicious circumstancesmight be much easier to detect. In response, we wouldemphasize that the findings synthesized here are notlimited to a few isolated studies or to unrealisticexperimental settings. The research literature includesmany studies of actual campaigns with careful mea-surement of their tone. It also bears reemphasis thatwe have considered a large array of possible moderatorvariables that plausibly could have influenced our con-clusions but did not.

In sum, our analysis of the greatly expandedresearch literature reinforces our earlier conclusionthat most of the conventional wisdom about negativecampaigning is not on sound empirical footing. Whatremains is to try to understand why negative cam-paigning is so pervasive and why the old saws about itseffectiveness for its practitioners and its destructive-ness to the political system continue to be repeated. Weoffer several speculations. First, every careful studywith which we are familiar that attempts to measurethe relative amounts of positive and negative cam-paigning has reported that attacks comprise 30% to40% of all campaign messages (Geer, 2006; Goldsteinand Strach, 2004; Lau and Pomper, 2004). Earlier wecited the CQ Weekly analysis of how negative the 2006congressional campaigns were. But that analysis per-tained only to ads sponsored by the political parties.Those ads were largely negative. In recent campaignyears, candidates have sponsored about two-thirdsof all the ads during a campaign, and if those adswere overwhelmingly positive, the overall 2006 cam-

paign would, like all of its predecessors, have beenpredominately positive. Attack ads are more memo-rable than the typical positive ad, and this bias inevi-tably distorts perceptions of the prevalence of negativecampaigning.

Why do consultants continue to urge candidatesto attack when there is little evidence that this strategyactually works, and when an attack almost inevitablyprovokes a counter attack (Lau and Pomper, 2004)? Ifthe 30% to 40% estimate cited in the preceding para-graph is accurate, in most instances consultants do notadvocate attacking. Sometimes, though, they find itmore feasible to craft high-visibility negative messagesthan equally high-visibility positive ones. Most candi-dates play up one or two general positive themes buttry many different attacks on the opponent in smaller,targetable subsets of the electorate. It is probably easierto fine-tune attacks than positive messages, and there-fore a focus on what is more controllable and new—the negative messages of a campaign—requiresconsultants to spend most of their time crafting thenegative messages of a campaign (even though mostcampaign dollars buy positive ads) and makes it morelikely that they will give those messages undue creditfor favorable outcomes.

The behavior of journalists and political punditsvis-à-vis social science research on negative campaign-ing seems easier to explain. Undoubtedly many ofthem are simply unfamiliar with this research. Beyondthat, just as local news programs typically lead withstories of murder and mayhem, political commenta-tors seem to relish writing about awful, false, mislead-ing, unfair, and mean-spirited political attack ads.Academic research sometimes provides a “hook” fortheir analyses of the causes and consequences of nega-tive campaigning. That there is little sound evidencefor most of the conventional wisdom about negativeads is not a story that most journalists are predisposedto tell or that most lay readers are predisposed tohear—especially because they are likely to remember afew especially vivid attack ads that seemed to work.

We might suggest a new hook for journalists to tryin the next election cycle: why do candidates continueto attack when there is so little evidence that attack-ing works? It seems inevitable that the conventionalwisdom will continue to be espoused by political com-mentators and acted upon by political professionalsand the candidates they advise, but we hope that atsome point even pundits and practitioners will beginto view more skeptically prevailing beliefs about thepurported potency of negative campaigning.


APPENDIX TABLE 1A Description of Negative Campaign Studies Used in the Meta-Analysis, Including Old(from Lau et al. 1999), Updated, and New Findings

Study Independent VariableSubjects and

DesignDependentVariables Results

1. Abbe et al. 2000 Ratings by candidates/consultants as towhether eithercampaign was negative

287 candidates andcampaign consultantsinvolved incompetitive Houseraces in 1998

Actual vote share Negative campaigning proved to bea slightly ineffective campaignstrategy, d = -.06.

2. Ansolabehereand Iyengar 1995

Positive or negative adfor actual candidatesinserted into regularcommercial break oflocal news program

Experiment with aconvenience sample of2,216 residents ofSouthern California

Intended turnout

Vote intention

Political efficacy

Negative ads depressed intendedturnout, d = -.10.Negative ads decreased intendedvote for the attacker during primaryelections, d = -.14, but increasedvote intention for the attackerduring general elections, d = .10.Negative ads decreased bothinternal (d = -.10) and external(d = -.14) political efficacy.

3. Ansolabehere,Iyengar, and Simon1999 (UpdatesAnsolabehere,Iyengar, Simon,and Valentino1994)

Positive/negative“tone” of 1992 Senatecampaigns, codedfrom newspaperaccounts; and recall ofpositive or negative adfrom the 1992 and1996 presidentialelections

Aggregate analysis ofturnout in 34 Senateelections; and 2SLSanalysis of 1992 and1996 ANES data

Actual turnout

Reported turnoutfrom survey

States with more negative Senateelection campaigns had lowerturnout, d = -1.27.Recall of negative ad associatedwith lower probability of votingcompared to recall of positive ad,d = -07.

4. Arceneaux andNickerson 2005

A positive or negativemessage about an issueof concern to the voterfrom a Democratic501c3 organization inMinnesota, and apositive or negativemessage about twoballot propositionsfrom a differentnonprofit organizationin California

Two large fieldexperiments deliveringpositive or negativecampaign messages toyoung undecidedvoters in Minnesota,(N = 6206, but mostdata come from asurvey of 1385 ofthem) and to minorityvoters in Los Angeles(N = 63,354, but mostdata come from asurvey of 333 ofthem), during the2004 presidentialelection

Turnout

Vote Choice

Affect for target

Affect for attacker

Knowledge (aboutballot measures)

Negative message associated withtrivially higher turnout in bothMinnesota (d = .01) and LosAngeles (d = .004).Positive message trivially moreeffective in convincing people tovote for Kerry in the Minnesotastudy (d = -.02); negative messageslightly more effective in gettingpeople to support each of the twoballot measures in the Los Angelesstudy, (average) d = .04.Negative message very slightly moreeffective in lowering evaluations ofthe target, d = .02.No effect of negative message onevaluations of attacker, d = -.00.Negative message slightly lesseffective in imparting knowledgeabout one of the ballotpropositions, and slightly moreeffective in imparting knowledgeabout the other, average d = .04.

5. Babbitt and Lau1994

Positive/negative“tone” of 1988 and1990 Senatecampaigns, codedfrom newspaperaccounts

Information aboutcandidates from 1988and 1990 U.S. Senateelections, and surveydata (N = 1,947) fromANES Senate ElectionStudy

General knowledgeabout incumbentsenator running inelection

Negative campaigning associatedwith somewhat less knowledge ofincumbent senator, averaged = -.05.

6. Bartels 2000 Geer’s coding ofproportion of negativeads used by twomajor-partypresidentialcandidates, 1968–1992

14,861 respondents in1968–1992 ANESpresidential yearsurveys

Differentialcandidate affect

Interest incampaign

Negative campaigns associated withslightly less overall positive affectfor the candidates, d = -.01.The more negative the campaign,the higher interest in it, d = .03.


APPENDIX TABLE 1A continued



7. Basil, Schooler,and Reeves 1991

Positive and negativeads from twosenatorial campaignsin another state

Repeated measuresexperimental design;24 communityresidents saw two“campaigns” consistingof three positive orthree negative ads foreach Senate candidate

Affect for attacker

Affect for target

Memory of ad

Candidate liked better andperceived as stronger when hepresented positive ads, d = -.30.Target liked better whenopponent used positive rather thannegative ads, d = .46.Positive ads recalled better thannegative ads, d = -.30.

8. Brader 2005 Positive or negativeissue-based ad, crossedby presence orabsences of music andimages designed toevoke enthusiastic orfearful emotions

286 Massachusettsresidents participatedin experiment in finalweeks of 1998Democraticgubernatorial primary;ad embedded in 30minute local newsbroadcast.

Intended turnout

Vote intention


Memory for ad

Political efficacy

Trust ingovernment

Interest incampaign

Negative ads led to slightly higherintention to vote in the primary,d = .01, and considerably more soin the general election campaigns,d = .20.Subjects more likely to reportintending to vote for candidatesponsoring a positive ad thancandidate sponsoring a negative ad,d = -.14.Subjects reported liking candidatewho sponsored a negative ad a littlemore than candidate whosponsored a positive ad, d = 14.Negative ad recalled slightly morethan positive ad, (estimate) d = .03.Internal and external politicalefficacy slightly higher amongsubjects shown positive adcompared to subjects shownnegative ad, (estimate) d = -.04 andd = -.03.Trust in government slightly higheramong subjects shown positive adcompared to subjects shownnegative ad, (estimate) d = -.03.Negative ads led to somewhathigher interest in the campaign,d = .11.





9. Brader andCorrigan 2006

CMAG data on theproportion of ads thatwere negative, and theproportion of ads thatwere character-basedattacks (which theauthors callmudslinging) thatwere shown during the2000 U.S. federalelection

Representative samplesof 13,311 respondentsduring the primaryelection, and 14,040respondents duringthe general electioncampaign, from the2000 NationalAnnenberg ElectionStudy (NAES)

Intended turnout

Interest in thecampaign

Mudslinging significantly decreasedintention to vote during theprimary (d = -.04) but had aweaker and nonsignificant effectduring the general elections(d = -.01), while exposure toissue-based attacks slightlyincreased intention to vote duringboth the primary and generalelections (d = .01 and d = .02),average effects of d = -.02 and d = 0for the primary and generalelections.Mudslinging significantly decreasedcampaign interest during theprimary (d = -.02) but had aweaker and nonsignificant effectduring the general elections(d = -.01), while exposure toissue-based attacks slightlydecreased interest during theprimary but slightly increasedinterest during the generalelection (d = -.01 and d = .01),average effects of d = -.02 andd = 0 for the primary and generalelections.

10. Bradley,Angelini, and Lee2005

Viewing 18 positive,moderate, or negativeads from Bush andGore in the 2000 U.S.presidential campaign

49 college studentsviewed 18 ads inrepeated measuresexperimental designwhile attached toelectrodes recordingfacial EMG responseand galvanic skinconductance response

Memory Recognition greater for negative adscompared to moderate or positiveads, p < .05 (assumed d = .68).

11. Bratcher 2001 Ratings of tone of1986 through 1998Senate electioncampaigns bycampaign consultantsand content analysis ofcoverage of campaigns

Aggregate analysis ofoutcome of 155 raceswhere the incumbentwas seekingre-election, 40 openseat races; and surveydata from 4476respondents to theANES Senate electionstudy

Vote choice


In aggregate, negative campaigninghurts incumbents (d = -.32), helpschallengers (d = .22), and hurtsopen-seat candidates (d = -.14). Atindividual level, negativism slightlyhurts Democrats (d = -.04) andslightly helps Republicans (d = .08).At individual level, negativismslightly hurts Democrats (d = -.08)and slightly helps Republicans(d = .05).

12. Brians andWattenberg 1996

Recall of exposure totelevision news andpositive and negativeads during 1992presidential campaign

51% of ANESsurvey(N = 1,263) whocould recall somepolitical ad

Memory of ad Negative political ads more likely tobe recalled relative to an estimate oftheir prevalence during campaign,d = .51.





13. Brooks 2000 Estimate of tone of1992 and 1994 Senateelection campaigns,from newspaperaccounts ofcampaigns

Replicates and extendsAnsolabehere andIyengar’s analysis withaggregate (N = 33,from 1994) andindividual-level surveydata (N = 2,019) from1992 and 1994 Senateraces, with adjustedmodel takingcontextual factors intoaccount

Actual turnout At aggregate level, negativecampaigning associated with lowerturnout in 1994 (d = -.24); at theindividual level, almost no effect ofnegative campaigning on turnout in1992 or 1994 (average d = -.01).

14. Brooks 2006(Replaces aggregatedata from 1992originally reportedin Brooks 2000)

Tone of 1992 Senateelection campaigns (asoriginally coded byAnsolabehere et al.1994), from newspaperaccounts of campaigns

Reanalysis/respecification ofAnsolabehere et al.’saggregate analysis of34 1992 Senateelection outcomes

Actual turnout(from both thePresidential andSenate elections)

Averaged across several differentspecifications of the basic turnoutmodel, turnout 1–3% higher, themore positive the campaign(average d = -.60).

15. Brooks andGeer 2007

Positive, civil negative,or uncivil negativeadvertisements usedby hypotheticalcandidate running forCongress

3 ¥ 2 experiment runon KnowledgeNetworks survey(N = 1,748) varyingtone and content(issue- or trait-based)of ads

Intention to vote inupcoming 2004national election

Fairness andinformativenessof adExternal politicalefficacy

Trust ingovernment

Interest incampaign

Civil negative ads resulted insomewhat lower levels of intentionto vote compared to positive adscondition (d = -.04), but uncivilnegative condition resulted insignificantly higher intention tovote, d = .09.Both civil negative (d = -.14) anduncivil negative (d = -.51) ads ratedsignificantly lower than positive ads.External efficacy somewhat lower inboth civil (d = -.11) and uncivil(d = -.05) negative ads conditions.Trust in government somewhatlower in both civil (d = -.11) anduncivil (d = -05) negative adsconditions.Political interest slightly lower inboth civil (d = -.04) and uncivil(d = -.05) negative ads conditions.

16. Bullock 1994 Exposure to ads fortwo hypothetical statesenate candidates,varying by type ofattack ad (image orissue) and ambiguityof ad

Experiment with 451randomly selectedprospective jurorsawaiting assignment

Affect for target

Affect for sponsor

Vote intention

Targeted candidates rated lessfavorably after exposure to negativeads compared to positive ads,d = 1.40.Candidates rated more positivelywhen sponsoring positive ads thanattack ads, d = -1.52.Negative ads caused the likelihoodof voting for the targeted candidateto drop significantly, d = 1.09.

17. Capella andTaylor 1992

Which candidateinitiated negative adsin 25 1986 Senatecampaigns with“substantial amountsof negativeadvertising”

Vote totals in 25Senate elections

Authors’ judgmentof whether negativead campaign“worked” or“failed” (i.e., howfinal resultsdiffered fromprojected resultsbefore ad campaignbegan)

Negative ad campaign decreasedproportion of vote obtained byinitiator of negative ads in 18 of 25elections, d = -.58.





18. Chang 2001 Positive or negativeprint ads created byresearcher

Experiment with 165college students whosaw either positive ornegative ads from eachof two fictitiouscompeting candidates

Differential liking ofcandidates

Memory of ads

Positive ads slightly moreefficacious than negative ads,d = -.04.Memory for negative ads muchgreater than memory for positiveads, d = .90.

19. Chang 2003 Positive or negativeprint ads created byresearcher from actualcampaign material,embedded inmagazine articles

Experiment with mixof 754 college studentsand adults randomlyassigned to see eitherno ads, a positive ad,or a negative ad fromeach of the twoleading candidates inthe 1998 mayoraltyelection in Taipei

Vote intention

Affect for target

Affect for sponsor

Attacking the opponent slightly lesseffective than staying positive forintended vote, d = -.05.Virtually no difference in affect fortarget of ads whether sponsor usespositive or negative ads, d = .00.Sponsor of negative ads likedslightly less than sponsor of positiveads, d = -.04.

20. Chang, Park,and Shim 1998

One candidate ranprominent negativead, the other did not

Survey of 297randomly selectedresidents of Columbia,MO

Affect for targetAffect for attacker

Liking for target decreased, d = .92.Liking for attacker decreased,d = -.67.

21. Chanslor 1995 Ads produced byincumbent in 1992Oklahoma senatorialelection

2 ¥ 2 experiment with166 college studentsubjects, varyingvalence and type (issuevs. image) of ad

Intended votechoice

Affect for target

Affect for sponsor

Intention to vote for incumbentvery slightly less after seeingnegative rather than positive ad,d = -.02.Target (challenger) likedsignificantly less after subjects sawad attacking him rather thanpositive ad from the incumbent,d = .31.Sponsor (incumbent) liked muchless after subjects viewed attack adrather than positive ad, d = -.70.

22. Clinton andLapinski 2004

Actual positive ornegative ads run byBush and Gore in2000 U.S. presidentialelection

Complex experimentaldesign with almost20,000 KnowledgeNetworks respondentswho viewed 0–2positive or negativeads from Gore and/orBush

Intended turnout

Reported turnout

In Wave I, viewing negative Gore adassociated with slightly higherintention to vote compared toseeing positive Gore ad, d = .03.In Wave 2, viewing Gore negativead associated with slightly higherturnout than no-ad control group(d = .05), while viewing both Goreand Bush negative ads associatedwith significantly higher turnoutthan control group, d = .07;reported turnout slightly higher forrespondents who saw negative ad(d = .03) or two negative ads,d = .01.

23. Craig and Kane2000

Respondent’sperceptions ofnegativism of 1998Bush and MacKayFlorida gubernatorialcampaigns

Representative sampleof 613 registeredFlorida voters

External politicalefficacy

Trust ingovernment

Respondents who perceived Bushand MacKay campaigns to benegative reported slightly lowerexternal political efficacy, d = -.04.Perceived campaign negativismassociated with slightly lower trustin government, d = -.08.

24. Craig, Kane,and Gainous 2005

Respondent’sperceptions ofnegativism of 1998Bush and MacKayFlorida gubernatorialcampaigns

Representative sampleof 301 registeredFlorida voters whoresponded to all threewaves of panel study

Knowledge ofcandidates’ issuepositions

Perceived negativism of Bush’s butnot MacKay’s campaignsignificantly associated with greaterknowledge of candidate’s issuepositions, mean d = .19.





25. Crigler, Just,and Belt 2002

Respondent’sperceptions of tone ofDole’s and Clinton’scampaigns, and theirpreferences afterviewing a videotapecontaining twopositive or twonegative ads

Representative panelsurvey of 630respondents, plus fieldexperiment withrespondents viewingvideotape of newsstories and politicalads from 1996 U.S.presidential campaign

Intended turnout

Intended votechoice

Affect for sponsorof ads

Respondents randomly assigned toviewing two positive ads more likelyto say they intended to vote inupcoming election compared torespondents who saw two negativeads, d = -.24.More negative Dole campaignperceived to be, less likely wererespondents to say they intended tovote for him (d = -.21), but noeffect of perceptions of Clinton’scampaign; mean d = -.11.More negative Dole’s campaignperceived to be, lower liking forhim, (d = -.30), but no effect ofClinton’s campaign (assume d = 0);mean d = -.15.

26. Dermody andScullion 2000

Respondent’sperceptions of positiveand negative campaignposters

Within-subjectsexperimental designwith 130 partyactivists reacting topositive and negativecampaign postersshortly before 1997British general election

Recall of ad(poster)

Negative posters much easier torecall than positive posters,d = 2.62.

27. Djupe andPeterson 2002

Coding of tone ofcampaign fromnewspaper articles forthree months prior tocontested Senateprimary

Aggregate analysis ofactual turnout in 33contested Senateprimary elections in1998

Turnout More negative campaigns associatedwith higher turnout rates, d = .92.

28. Finkel andGeer 1998

Proportion of negativeads used by twomajor-partypresidentialcandidates, 1960–1992

Reported turnout by12,252 ANESrespondents, 1960–92

Actual turnout Reported turnout slightly higher forrespondents in election years withhigher proportions of negative ads,d = .01.

29. Freedman andGoldstein 1999

Very sophisticatedestimate of number ofnegative ads seen bysurvey respondents

Second wave(N = 290) ofrepresentative panelstudy of 1997 Virginiagubernatorialcampaign

Reported turnout

Internal politicalefficacy

Viewing more negative adsassociated with higher turnout,d = .24.Exposure to negative ads slightlylowers internal political efficacy,d = -.05.

30. Freedman,Wood, and Lawton1999

Hypothetical vignettesin which fictionalcandidates attack eachother

Experiment withrandom sample ofVirginia votersexposed to fivevignettes

Intended turnout

Intended vote

Intended turnout significantly lowerwhen both candidates attack,d = -.40.Candidates who respond to attackdo worse than candidates who staypositive, d = -.16.

31. Fridkin andKenney 2004

Coding of tone of upto five ads from majorparty candidates for 97contested Senateelections between 1988and 1992

6,110 respondents toANES Senate electionsurveys

Affect for target

Affect for sponsorof ad

Both issue-based attacks (byincumbents and challengers) andcharacter-based attacks moreeffective than positive ads inlowering affect for target, mean ds.03 and .07.Both issue- and character-basedattacks associated with lower affectfor the sponsor compared topositive ads, ds ranging from -.01to -.05.





32. Garand andGraddy 2001

Ansolabehere et al.’scoding of nature of1992 Senatecampaigns fromnewspapers in eachstate

1,074 respondentsfrom 1992 ANESSenate election survey

Reported turnout Negative campaigns associated withsomewhat lower turnout, d = -.05.

33. Garramoneet al. 1990

Exposure to variouscombinations andnumbers of positiveand negativebiographical profilesand commercials fortwo fictional U.S.Senate candidates

Experiment with 372students assigned tocontrol,double-positive,single-positive,negative-positive,single-negative, ordouble-negativecondition

Differential affect

Intended turnout

Exposure to negative ads causedgreater image discrimination (thedifference between candidate imageevaluations) than exposure topositive ads, d = .38.Negative ads did not significantlyaffect intended turnout, d = -.18.

34. Geer 2006(Turnout dataupdates Finkel andGeer 1998).

Proportion of negativeads used by twomajor-partypresidentialcandidates, 1960–2004

Aggregate turnoutlevel from 12presidential electionsbetween 1960 and2004, plussurvey-based meanlevels of political trustand support forelections.

Actual turnout

Trust ingovernment

Faith in elections

Campaign negativism was slightlynegatively associatedwith one measure of turnout (VAP)and slightly positively associatedwith another (VEP), mean effectsize = -.01.Absolutely no relationship betweencampaign negativism and aggregatelevels of trust in government across11 elections, d = 0.Negativism was positively related toaggregate levels of faith in elections,d = .38 (although with only 9 casesthis effect is not at all significant).

35. Geer and Geer2003

Positive or negativeradio ad

Experiment with 121college studentsassigned to hearpositive or negative adfrom Democratic orRepublican candidate

Memory for ad Memory slightly higher for negativeads, d = .08.

36. Geer and Lau2003 (updates Geerand Lau 1998)

State-based estimatesof proportion ofnegative ads used bytwo major-partypresidentialcandidates, 1980–2000

Reported turnout by10,333 ANES surveyrespondents,1980–2000,with results of 90hypothetical modelsadjusted by BayesianModel Averaging

Reported turnout More negative estimated campaigntone in state, higher probability ofvoting, d = .06. Adjusting campaigntone for spending (d = .04) orestimated likelihood of exposure tocampaign ads (d = .05) increasesmobilizing effects of negative ads.

37. Geer and Lau2006 (updates Geerand Lau 1998)

State-based estimatesof proportion ofnegative ads used bytwo major-partypresidentialcandidates, 1980–2000

State-level turnout insix U.S. presidentialelections, 1980–2000,with results of 90different hypotheticalmodels adjusted byBayesian ModelAveraging

State-level turnout More negative estimated campaigntone in a state, higher estimatedturnout, d = .70;With campaign tone weighed byspending, mobilizing effect ofnegative tone decreases somewhat,d = -.12.

38. Globetti andHetherington 2000

Coding ofanti-Congress rhetoricin all ads run by majorparty Senatecandidates in 1994available fromOklahoma archives

420 ANES respondentsin 1992–1997 panelsurvey in states with1994 Senate elections

Trust ingovernment

Approval ofCongress

More anti-government rhetoric bytwo major party candidates, slightlylower trust in government, d = -.02.Anti-Congress rhetoric significantlyrelated to lower approval ofCongress, d = -.19.





39. Goldstein 1997 Number of negativeads shown in 75largest media marketsduring 1996presidential campaign,as recorded byCampaign MediaAdvertising Group(hereafter CMAGdata)

Aggregate analysis of1,588 counties,followed byindividual-levelanalysis of 879 ANESrespondents living in75 largest mediamarkets

County-levelturnoutReported turnout

Reported votechoice


More negative ads run in county,lower turnout, d = -.27.More negative ads exposed to,greater probability of voting,d = .09.Exposure to candidate’s negativeads led to somewhat greaterprobability of voting for him,d = .08.More negative ads a respondentexposed to, lower external politicalefficacy, d = -.05.

40. Greenwald2005

CMAG data onnumber of positiveand negative adsshown by Democraticand Republican Senatecandidates, weightedby likelihood ofexposure

Survey data fromrespondents in 2000National AnnenbergElection Study, brokeninto states withcompetitive andnoncompetitive Senateelections

Vote intention Negative advertising noticeably lesseffective than positive advertising incompetitive elections, d = -.01; butnegative advertising more effectivethan positive advertising amongindependents (d = .07) and for allrespondents in noncompetitiveelections, d = .10.

41. Haddock andZanna 1997

Impressions of actualcandidates before andafter controversialattack ads aired during1993 Canadiannational election

“Natural” experimentwith 110 collegestudents

Affect for attacker

Affect for target

Affect toward attacker decreasedafter airing of ads, d = -.32.Affect toward target increased afterairing, d = -.35.

42. Hill 1989 Positive or negative adfrom Bush or Dukakiscampaign

Experiment with 120college students

Affect for attacker

Affect for target

Sponsor of ad liked less if ad wasnegative rather than positive,d = -.65.Target of ad liked more if ad wasnegative rather than positive,d = -.13.

43. Hitchon andChang 1995

Exposure to positive,neutral, and negativeads from female andmale gubernatorialcandidates

Experiment using 3(positive, negative,neutral) ¥ 2 (female,male) within-subjectfactorial design with75 undergraduates

Affect for attacker

Memory for ads

More negative affect for candidateswho attacked their opponents,d = -.81.Positive ads produced highestcandidate recall, while negative adsproduced lowest candidate recall,d = -58.

44. Hitchon,Chang, and Harris1997

Exposure to positive,neutral, and negativeads in gubernatorialrace

Experiment using a 3(positive, neutral,negative) ¥ 2 (male,female) within-subjectfactorial design with72 undergraduates

Affect for attacker Negative ads produced lessfavorable responses than positive orneutral ads, d = -.80.





45. Houston andDoan 1999

Exposure to positiveor negative ads fromliberal or conservativemock senatecandidates

2 ¥ 2 ¥ 2 experimentwith 173 collegestudent subjectsmanipulatingcandidates, positive ornegative tone of ads,and whether they citeevidence to supporttheir claims

Intended turnout

Affect for target

Affect for attacker

Differential affect

Respondents significantly less likelyto say they would vote (underunfavorable circumstances) whenshared ideology candidate rannegative rather than positivecampaign (d = -.42), but no effectof tone of opposing ideologycandidate’s campaign (assumed = 0), so mean d = -.21.Target of negative ads likedsomewhat less than target ofpositive ads, d = .23.Sponsor of negative ads likedsignificantly less than sponsor ofpositive ads, d = -.51.Both shared and opposing ideologycandidates liked relatively less whenengaged in negative compared topositive campaigns, d = -.16.

46. Houston,Doan, andRoskos-Ewoldsen1999

Exposure to positiveor negative campaignads from liberal orconservative mocksenate candidates

Two experiments withundergraduates assubjects manipulatingagreement withcandidate’s ideologyand nature of politicalads (Ns = 77 and 68)

Affect for targetof ads

Affect for attacker

Vote intention

Target in both experiments likednonsignificantly less when othercandidate attacks than whenopponent positive, d = .26 and .40.Affect for sponsor lower whencandidate uses negative rather thanpositive ads in both experiments, d-.55 and -.36.Subjects in both experimentsreported lower likelihood of votingafter exposed to negative comparedto positive ads, d = -.55 and -.56.

47. Iida 2005 Categorization of 1992Senate electioncampaigns as“positive,” “neutral,” or“negative” byAnsolabehere et al.1994

1307 respondents in1992 ANES Senateelection survey instates whereincumbent seekingre-elections facedmajor party challenger

Candidateknowledge(willingness toplace candidates on7-point liberalism-conservatism scale)

Knowledge of incumbent’s ideologysomewhat less (d = -.02) butknowledge of challenger’s ideologysignificantly greater (d = .08) instates with largely negativecampaigns (mean d = .03).

48. Jackson andCarsey 2007

CMAG data onnumber of positiveand negative adsactually shown in eachstate

550 1998 ANESrespondents, and28,362 CurrentPopulation Surveyrespondents, reportingwhether they voted inSenate election thatyear

Reported turnout In both ANES survey and CPSstudy, exposure to more negativeads led to higher probability ofvoting, d = .22 and d = .09.

49. Jackson,Mondak, andHuckfeldt 2005

CMAG data from 2002House, Senate, andgubernatorialcampaigns, andseparate coding ofmost inflammatory“mudslinging” ads

Representative sampleof 1,219 respondents,asked about theirtelevision viewinghabits

Internal efficacy

External efficacy

Affect forgovernment(Congress)

Viewing more negative ads had nomeasurable effect on internalefficacy, while mudslingingassociated with slightly higher levelsof efficacy, d = .05.Viewing more negative adsassociated with slightly lowerexternal efficacy (d = -.02), whilemudslinging associated with slightlyhigher external efficacy.Viewing more negative adsassociated with slightly higher levelsof affect for Congress, d = .08, whilemudslinging had no measurableeffect.





50. Jackson andSides 2006

Tone of up to five adsfrom 1990 senateelections, coded byKahn and Kenney

Attempt to replicateKahn and Kenney’sanalysis of 1990 Senateelection campaignswith ANES Senateelection survey(N = 2,003)

Reported turnout More negative campaign tone,higher turnout, d = .07; but tonetimes spending depresses turnout,d = -.04 (neither effect statisticallysignificant).

51. Jasperson andFan 2002

Finely measured toneof each ad aired byWellstone andBoschwitz in 1996Minnesota Senate race,weighted by how manytimes each ad aired

Time series analysis ofdaily attitude towardchallenger Boschwitzduring last six monthsof campaign

Affect for target

Affect for attacker

Wellstone’s attack ads moreeffective in lowering affect for targetthan Boschwitz’s positive ads are inincreasing his rating, d = .10.Boschwitz’s attack ads moredeleterious to himself than theydetract from liking of Wellstone,d = -.07.

52. Kahn and Geer1994

Actual positive ornegative ads fromout-of-stategubernatorialcandidates inserted inregular ad breaksduring TV sitcom

Experiment with 209college students whosaw one or twopositive or negativeads

Affect for attacker Sponsor of ad liked less after anegative compared to a positive ad,d = -.28; sponsor liked much lessafter two negative compared to twopositive ads, d = -.74.

53. Kahn andKenney 2004(Updates Kahn andKenney 1998a,1998b, and 2002)

Authors’ coding oftone of ads producedby U.S. Senatecandidates, 1988–1992,and separate coding bycampaign managers ofwhether campaigninvolved mudslinging

ANES Senate ElectionStudy survey(N = 6,110, thoughturnout and campaigninterest effects basedsolely on data from1990 campaign)

Reported turnout

Affect for target

Affect for sponsorof ads

Correctidentification ofcampaign themesemployed bycandidates

Campaign interest

Relatively greater use of negativeads by both candidates associatedwith higher turnout, d = .08, butmudslinging associated with lowerturnout, d = -.10.Both incumbents and challengersliked significantly less whentargeted by negative ads, d = .06.Both incumbents (d = -.05, p < .05)and challengers (d = -.03, ns) likedless when they attack theiropponents.Respondents 17% more likely tocorrectly identify campaign themeswhen campaign primarily negativethan when primarily positive,d = .14 for campaign ads, d = .09for mudslinging.No significant effect of negativismof ads on campaign interest, butmudslinging significantly decreasedinterest, d = -.01 and d = -.08.

54. Kaid 1997 Exposure to actual adsfrom 1996 Clinton orDole campaigns

Experiment with 116undergraduates assubjects

Vote intention

Affect for target

Affect for attacker

Subjects much more likely to intendto vote for attacker after viewingone of his negative ads compared toone positive ad, d = 1.77.Target of ads liked less afternegative ad compared to positivead, . = .68.Sponsor of negative ad liked slightlymore than sponsor of positive ad,d = .28.

55. Kaid andBoydston 1987

One of five actual adsused by congressionalcandidate fromanother district

Experiment withconvenience sample of428 residents ratedcandidate before andafter seeing one of hisads

Affect for targetof ads

Affect for target of ads droppedafter viewing negative ad, d = .36.





56. Kaid, Chanslor,and Hovind 1992

Exposure to differenttypes of actualpolitical ads (positive,negative, issue, image)and type of televisionprogram surroundingthe ad

Experiment with 3 ¥ 3factorial designvarying program andcommercial type,involving conveniencesample of 283members of civicgroups and collegestudents

Vote intention

Affect for attacker

Memory for ad

Positive image ads produced greaterlikelihood of voting than negativeads, d = -2.40.Positive issue ads produced highercandidate evaluations for sponsorthan negative ads, d = -2.05.Aspects of positive issue adsremembered more than aspects ofnegative ads, d = -1.15.

57. Kaid, Leland,and Whitney 1992

Exposure to positiveand negative ads fromBush and Dukakiscampaigns

Experiment with 112undergraduates whosaw 3 Bush ads (2positive, 1 negative), 3Dukakis ads (2positive, 1 negative),or 3 ads from bothcandidates (2/3positive for each)

Memory for ads Positive ads more likely to beremembered than expected bychance (i.e., .67), d = -.30.

58. King andHendersen 1999

Exposure to a positiveor negative ad from1998 Michigangubernatorial election

2 ¥ 2 experimentaldesign using 111undergraduates, varyingpositive/ negative natureof ad, controlling forprior liking of thecandidates (mediansplit)

Affect for attacker

Affect for target

No effect of tone of ad on affect forsponsor (assumed d = 0).No effect of tone of ad on affect fortarget (assumed d = 0).

59. King,Hendersen, andChen 1998

Exposure to singlepositive or negative adfrom Clinton or Dolecampaigns, near endof 1996 U.S.presidential campaign

2 ¥ 2 ¥ 2 experimentusing 137undergraduatesubjects, varyingpositive/ negativenature of ad,Clinton/Dole assponsor of the ad,controlling on priorliking of candidates

Liking for sponsorof ads

Liking for targetof ads

Vote intention

Memory for Ads

Clinton liked less when exposed tohis negative ad, but no effect ofexposure to Dole ads, meand = -.32.Dole liked less after exposure tonegative Clinton ad, but no effect ofexposure to Dole ads, mean d = .31.Likelihood of voting for Clintondecreased after exposure to hisnegative ad, but no effect ofexposure to Dole ads, meand = -.23.Positive Clinton ads better recalledthan negative Clinton ads, but noeffect of exposure to Dole ads,mean d = -.40.

60. King andMcConnell 2003

Negativeadvertisements from1996 Durbin–SalviSenate race in Illinois

Experiment with 121college students whoviewed 0–3 politicalads by Durbinattacking Salvi,inserted into normalcommercial breaks oftelevision program

Vote intention

Affect for target

Affect for attacker

Viewing attack ads increasedexpressed likelihood of voting forattacker, d = .34.More attack ads seen, lower affectfor target, d = -.35.More attack ads seen, more sponsorof attacks liked, d = .39; however, asignificant (negative) “boomerang”effect of number of ads squared,such that viewing three ads resultedin lower affect for attacker thanviewing one or two ads.

61. Landi 2004 Lau and Pomper’scoding of tone of U.S.Senate electioncampaigns, 1988–1998

Aggregate analysis ofturnout in 190 Senateelections, and reportedturnout byrespondents in1988–1992 ANESSenate election survey

Turnout Turnout positively butnonsignificantly related tonegativism of campaigns in bothaggregate (d = .09) andindividual-level analyses (d = .02).





62. Lang 1991 Exposure to 8randomly selectedpositive and negativeads varying emotionalappeal andaudio-visual format

Experiment using 4(order) ¥ 2 (emotion)¥ 2 (format) ¥ 4(repetition) mixedmodel factorial designwith 67undergraduates

Memory for ads More information recalled aboutnegative ads than positive ads,d = .83.

63. Lau andPomper 2004(updates Lau,Pomper, andMumoli 1998)

Positive/negative“tone” of 1992–2002U.S. Senate campaigns,coded from newspaperaccounts

Data from 6,283 ANESrespondents, andaggregate analysis ofvote totals from 191Senate elections

Actual electionoutcomes andreported votechoice

Actual turnoutand reportedvoting


Trust ingovernment

At aggregate and individual levels,incumbent negativism resulted indoing significantly worse at thepolls (d = -.64 and d = -.12,respectively), while challengers didnonsignificantly better (d = .26 andd = .01) and no effect in open-seatraces (d = .03 and d = -.01,respectively).No significant effect of totalcampaign negativism on turnout ataggregate (d = -.07) or individuallevel (d = .01).Very small insignificant negativeeffect of negative campaigning onexternal political efficacy, d = -.01.No effect of negative campaigningon trust in government, d = 0.

64. Lau andRedlawsk 2005

Positive or negativeads from fictitiouscandidates in a mockpresidential electioncampaign

2 ¥ 2 experiment withconvenience sample of407 adult subjects,manipulating positiveor negative tone ofcandidates’ ads

Affect for target

Affect for attacker

Differential affect

Vote choice

Memory for ads

Target liked slightly more whenattacked, d = -.04.Sponsor liked more when attacks,d = .09.Differential affect slightly favorsattacker, d = .05.Attacking produces fewer votes thanstaying positive, d = -.06.Negative ads more easily recalledthan positive ads, d = .39.

65. Lawton andFreedman 2001

Combines ratings ofperceived fairness ofeach ad (fromFreedman, Wood, andLawton 1999) andCMAG satellite addata to estimate actualexposure to fair andunfair ads

Panel survey of 873Virginians of votingage

Turnout Greater exposure to unfair adsdepressed turnout in Senateelection (compared to exposure tofair ads), d = -.21.

66. Lemert et al.1991

Respondents reflectingon positive or negativead they could recallseeing during 1988presidential election

Representative sampleof 1,256 surveyrespondents

Likelihood ofvoting for Bush orDukakis

Type of ad recalled

Respondents much less likely to saythey would vote for candidate whenrecalled attack ad rather thanpositive ad, average d = .24.Negative ads much more likely tobe recalled, d = .25.

67. Lemert, Wanta,and Lee 1999

Recall of positive ornegative ad fromcandidates in specialOregon senateelection, andperception thatDemocrat had lived upto pledge to stopnegative campaigning

Representative surveyof 308 registeredvoters in Eugene areawho had voted in pastthree Senate elections

Vote choice

Turnout

Respondents much less likely toreport voting for candidate theycould remember using negative adscompared to a candidate they couldremember using positive ads,d = -.87.Reported turnout slightly lowerwhen respondents could remembernegative rather than positive ad,d = -.13.





68. Leshner andThorson 2000

Respondent’sperceptions of whethercampaign was toonegative

Representative surveyof 393 registeredvoters in Kansas City,3–4 months after 1996national election

Turnout

Public mood

Trust ingovernment

No significant direct effect ofnegativism of 1996 campaign onturnout, d = .04.Perceived negativism of campaignsignificantly associated withnegativism of public mood,d = -.36.Perceived negativism of campaignsignificantly associated with lowertrust in government, d = -.28

69. Luskin andBratcher 1994

Authors’ rating ofnegativism of1986–1992 U.S. Senatecampaigns, based ontheir reading ofvarious campaignreports

Aggregate analysis ofvote totals from 125Senate elections

Turnout Negative campaigning associatedwith small and nonsignificantdecrease in turnout (d = -.12),though decrease noticeably larger instates with large numbers ofindependents, d = -.30.

70. Martin 1999 Coding of tone ofSenate campaign fromstate’s largestnewspaper

1990 ANES(N = 1,021)

“Collectiveawareness,”essentially countof number ofimportantproblems facingcountry

More negative tone of campaign,more likely respondents to reportimportant problems facing country,d = .12.

71. Martin 2004 CMAG data onnumber of negativeads shown in mediamarket by Clinton andDole in 1996 U.S.presidential election

1996 ANES (N = 522) Awareness ofproblems facingcountry

More negative advertising exposedto, more likely to report importantproblems facing country, d = .15.

72. Martinez andDelegal 1990

Exposure to negativeads from one or bothcandidates inhypothetical election;perceived positive/negative nature of1988 Bush andDukakis campaigns

Pre/post experimentwith 131 collegestudents as subjects,and representativesurvey of 421respondents

Trust ingovernmentAffect for attacker

Affect for target

Trust in government increased afterexposure to negative ads, d = .14.More candidate’s campaignperceived as negative, less sponsorliked, d = -.28.More opposing candidate’scampaign perceived as negative,more target liked, d = -.48.

73. Mathews andDietz-Uhler 1998

Exposure to positiveor negative “familyvalues” ad from mockDemocratic orRepublican senatecandidate

Experiment with 125college students assubjects

Affect towardsponsor

Likelihood ofvoting for attacker

Sponsor of positive ad liked muchmore than sponsor of negative ad,d = -.52.Subjects much more likely to intendto vote for sponsor of positive adthan sponsor of negative ad,d = -.62.

74. McBride,Toburen, andThomas 1993

Exposure to fournegative ads from1990 Louisiana Senaterace for firstexperiment; exposureto description of fournegative ads from1992 presidential racein second experiment

Two experimentsinvolving 223undergraduates, 70 ofwhom recontactedafter election tomeasure actualturnout

Intended turnout

Actual turnout

No significant effect of ad valenceon intention to vote, d = .12.Subjects exposed to negative adsslightly (and nonsignificantly) lesslikely to vote, d = -.06.





75. Meirick andPfau 2005

Positive or contrast adproduced byDemocratic candidatesin 2004 Iowa caucuses

Experiment withconvenience sample of117 Democratsparticipating in Iowacaucuses in early 2004,randomly assigned tosee positive or contrastad inserted intoregular commercialbreak of local newsprogram

Affect for attacker Sponsor of contrast ad attackingPresident Bush liked slightly morethan sponsors of positive ads,d = .08.

76. Merritt 1984 Exposure to negativeand neutral ads fromcandidates in 1982California Assemblyrace

Representative surveyof 314 respondents incandidates’ district

Affect towardattacker

Affect towardtarget

Correct recallof ad

More negative affect towardsponsor when ad negative ratherthan positive, d = -.86.More negative affect toward targetafter negative rather than positivead, d = .77.Negative ad more likely to becorrectly recalled, d = .29.

77. Min 2004 Positive or negativetone and focus(policy/ personality) ofnews coverage ofcampaign

2 ¥ 2 experiment with113 college studentswho read articlesabout two fictitiouscandidates forCongress

Intended turnout

Vote intention

Affect for campaign

Subjects exposed to negativecampaign less likely to say theywould vote in this campaign,d = -.13.Subjects exposed to negativecampaign slightly less likely to saythey would vote for attacker,d = -.04.Negative campaign liked less thanpositive campaign, d = -.44.

78. Naman 2000 Comparative versuspositive ads from 1998Senate election inWashington

Experiment with 63undergraduatesassigned to see eitherone negative directcomparative ad percandidate fromanother state or onepositive ad percandidate

Affect for attacker

Vote intention

On average, each candidate likedslightly more in the positive adcondition than in comparative adcondition, d = -.10.Subjects slightly more likely to saythey would vote for candidate innegative compared to positive adconditions, d = 13.

79. Newhagen andReeves 1991

Reactions to actualBush and Dukakispositive, negative, orcomparative ads

Within-subjectsexperimental design;30 residents reactingto 28 different ads

Memory for each ad Recall more accurate (and quicker)for negative rather than positiveads, d = 1.30.

80. Niven 2006 1–3 negative directmail ads opposingincumbent mayorsponsored by 3rdparty

Field experiment with1,400 registeredFlorida votersrandomly assigned toreceive 0–3 mailingscritical of incumbentmayor; plus follow-uppost-election phonesurvey of 168respondents randomlyselected from originalsample

Turnout

Knowledge ofcandidates

More negative mailings received,more likely to vote: 36% (3 ads) vs34% (2 ads) vs 30% (1 ad) vs 27%(0 ads), overall d = .14.Those receiving negative admailings more likely to say theyknew something about candidatescompared to control group (57% vs43%), d = .09.

81. Niven 2005 Respondent’sperceptions of tone of1998 Bush vs MacKayFlorida gubernatorialcampaign

Panel survey of 301Florida voters duringthe campaign


More negative campaigns perceivedto be, more knowledgeable aboutcandidates’ issue positions, d = .19.





82. Patterson andShea 2001

Rating by local experts(newspaper editors) oftone of just-completedcongressional elections

169 respondents from1998 ANES surveywho lived in any of 56Congressional districtsfrom which expertratings had beenobtained


Governmentresponsiveness

Interest in thecampaign

Negative campaign tone associatedwith slightly higher internalpolitical efficacy, d = .02.Negative campaign tone associatedwith slightly lower beliefs ingovernment responsiveness,d = -.04.Negative tone associated withsomewhat greater attention tocampaign, d = .14.

83. Pfau et al. 1989 Exposure to attack adfrom least preferredcandidate during 1988presidential campaignvs. no-exposurecontrol group

Experiment withrepresentative sampleof 374 likely voters

Affect towardattacker

Vote intention

Sponsor of negative ad was likedmore after exposure to ad,compared to control group, d = .75.Respondents more likely to intendto vote for sponsor of negative adcompared to control group, d = .92.

84. Pinkleton 1997 Amount of negativeinformation abouttarget included in adabout fictitiouscandidates

Experiment with 165college studentsassigned tobetween-groupspre-post design(including no adcontrol group)

Affect towardattackerAffect towardtarget

More negative ad, less sponsorliked, d = -.44.More negative ad, less target liked,d = .67;

85. Pinkleton 1998 Amount of negativeinformation abouttarget included in adabout fictitiouscandidates

Experiment with 165college studentsassigned tobetween-groupspre-post design(including no adcontrol group)

Affect towardattackerAffect towardtargetLikelihood ofvoting for targetor sponsor

Sponsor liked less if attacks,d = -.40.Target liked slightly less if attacked,d = .04.Likelihood of voting for sponsordecreases slightly if attacksopponent, d = -.03.

86. Pinkleton andGarramone 1992

Number of negativeads recalled from twomajor party candidatesfor governor and forsenator

Phone survey of 405likely voters just before1990 Michigan senateand gubernatorialelections

Intended turnout

Interest incampaign

Intention to vote very slightlyhigher, more negative ads recalled,d = .01.More negative ads recalled, greaterinterest in campaigns, d = .20.

87. Pinkleton, Um,and Austin 2002

Simulations of directmail political ads, eachread twice by subjects

Experiment with 236college studentsrandomly assigned topositive, negative,comparative, or no-adcontrol group

Trust ingovernment


Belief in citizen’sduty to vote

Somewhat lower trust ingovernment in negative adscondition compared to positive adcondition, d = -.14.Somewhat lower external politicalefficacy in negative ad conditioncompared to positive ads, d = -.19.No significant relationship betweenbelief in citizen’s duty to vote andad condition, (assumed) d = 0.

88. Rahn andHirshorn 1999(updates Rahn andHirshorn 1995)

Exposure to 4 positiveor 4 negative ads from1988 presidentialcampaign

Experiment withconvenience sample of69 8-13 year-oldchildren

Vote intention

Public mood

Liking for thegovernment

Internal politicalefficacyGovernmentresponsiveness

Children exposed to negative ads9% more likely to say they wouldvote in this election if were oldenough, d = .32.Mood significantly lower forchildren exposed to 4 negative adstwo years after election, d = -.87.Children exposed to negative adshad lower affect for government,d = -.42.No effect of ad manipulation oninternal political efficacy orperceived governmentresponsiveness, assumed ds = 0.





89. Roberts 1995 Reported recall ofBush or Clinton ads

Representative phonesurvey of 931respondents after 1992presidential election

Memory for ad Negative Bush and Clinton adsslightly more likely to be recalledthan would be expected by chance,d = .05.

90. Roddy andGarramone 1988

Positive or negativeresponse toopponent’s attack ad

2 ¥ 2 experiment with274 undergraduatesvarying type of attack(issue or image) andnature of response(positive or negative)

Affect for sponsorof response ad

Affect for target

Intended votefor/againstsponsor ofresponse ad

Candidate who responded positivelyrather than negatively liked more,d = -.09.Target liked less after negativeresponse compared to positiveresponse, d = .06.Intention to vote for candidate whoresponded negatively rather thanpositively higher, d = .10.

91. Sanders andNorris 2002

Advocacy and attackParty ElectionBroadcasts (PEBs)during the 2001British nationalelection

Experiment withrepresentative sampleof 919 London votersexposed to differentcombinations of PEBsactually aired by theparties during thecampaign

Affect for target

Affect for sponsor

Both Conservative (d = -.15) andLabor (d = -.16) attack ads verycounter-productive compared totheir advocacy ads, averaged = -.16.Conservative attack ad resulted inincrease in affect for own party,compared to advocacy ad (d = .16),but Labor attack ad resulted inlower affect for Labor compared topositive ad (d = -.07), averaged = .05.

92. Schultz andPancer 1997

Whether fictitiouscandidate attackscharacter of opponent

134 undergraduatesrandomly assigned to2 ¥ 2 experiment,varying gender ofcandidate and whethers/he attacks opponent

Affect for attacker

Vote intention

No significant difference inevaluations of sponsor, (assumed)d = 0.No significant difference in voteintention, (assumed) d = 0.

93. Shapiro andRieger 1992

Positive or negativeradio ads from twofictitious candidates intwo local elections

106 undergraduates in2 ¥ 2 mixedexperimental design:subjects heard apositive and a negativeimage or issue ad

Affect for attacker

Affect for targetof adVote intention

Memory for ad

Sponsor of negative ads liked lessthan sponsor of positive ads,d = -1.89.Target of negative ad liked less thantarget of positive ad, d = .50.Subjects more likely to intend tovote for sponsor of positive adrather than negative ad, d = -1.29.Negative ads more likely to beremembered, d = .54.

94. Shen andWu 2002

Positive or negativeissue-based newspaperads about fictitiousstate senate candidates

Experiment with 150college studentsubjects randomlyassigned to either ano-ad control group,or to read positive ornegative ads sponsoredby either candidate orfictitious PAC

Affect for target

Affect for attacker

Differential affect

Vote intention

Target liked significantly less whenopponent attacks compared towhen opponent stays positive,(estimated) d = .49.Sponsor liked significantly lesswhen attacks compared to whenstays positive, (estimated) d = -.49.Target of negative ads likedsignificantly less than attacker,estimated d = .55.Higher likelihood of voting forattacker compared to target ofnegative ads, estimated d = .55.





95. Sides 2006 Advertising tone ofBush, Gore,Democratic Party, andRepublican Party adsshown in ninebattleground statesduring the last tenweeks of the 2000 U.S.presidential election(CMAG data)

Time series crosssectional analysis ofweekly tracking pollsconducted by the Bushcampaign

Intended votechoice

Negative advertising by the Bushcampaign was effective in raisinghis vote totals relative to hisopponents (d = .31), but positiveadvertising was much more effectivefor Gore (d = -.87) and somewhatmore effective for the Democraticand Republican parties (ds = -.12and -.22). Mean d = -.23.

96. Sides et al.2003

Advertising tone, asmeasured by numberof ads shown each dayin media market

Representative surveyconducted during1998 Californiagubernatorial election

Perceived usefulnessof information fromcandidates’advertisingcampaigns

More positive ads candidates aired,more useful information theircampaign perceived to provide,average d = -.11.

97. Sides et al.2005

Advertising tone, asmeasured by volumeof positive andnegative themes in adsshown each day inmedia market duringcampaign

Representative surveysof San Francisco andChicago during 2002gubernatorial elections

Intended turnout Intention to vote slightly higher inCalifornia (d = .03) but somewhatlower in Illinois (d = -.11), morenegative the campaign.

98. Sonner 1998 Author’scharacterization of adcampaigns incompetitive primaryelection for governor

Representativestatewide trackingpolls conducted at 16different points duringyear-long primaryelection campaign

Liking for target

Liking for attacker

Vote intention

When challenger attackedincumbent, target’s favorableratings fell slightly (d = .04), butincumbent’s attack on challengerlater in campaign resulted in nochange in target’s favorables (d = 0).Challenger who attacked incumbentlost two points in own favorableratings, d = -.04, while subsequentattack by the incumbent resulted inno change in his ratings, d = 0.Challenger who initially attackedincumbent lost 8 points in polls,d = -.15, while incumbent whosubsequently responded withattacks saw small increase in lead,d = .05.

99. Stevens 2002 Exposure to negativeBush or Gore adduring 2000 U.S.presidential election,vs. no-ad controlgroup

2 ¥ 2 experimentvarying exposure to adand presence ofprompt, innon-random internetsample of 492 people

Reportedprobability ofvotingPublic mood


Reported probability of votingsomewhat higher after exposure tosingle negative ad, d = .07.Public mood significantly lowerafter exposure to single negative ad,d = -.25.External efficacy lower afterexposure to single negative ad,d = -.17.

100. Stevens 2005 Exposure to 0–3negative ads from2000 Minnesota senateelection, and CMAGdata describing adsaired during 1998California, Georgia,and Illinoisgubernatorial elections

Experiment with 190college students whowatched local eveningnews program fromthe previous year intowhich were inserted 0,1, or 3 negative adsfrom the senateelection that wasoccurring at the time;and ANES pilot surveyof California, Georgia,and Illinois

Three measures ofinformation aboutcandidates: quantityof information,perceptions of policydifferences, andaccuracy ofplacement on issues(only first and thirdavailable for surveydata).

Compared to no-ad control group,subjects exposed to negative adshad somewhat lower quantity ofinformation about candidates(d = -.10), perceived somewhatfewer policy differences betweenthem (d = -.07), but weresignificantly more accurate inplacing them on the issues(d = .27), average d = .03; exposureto negative ads similarly related tosomewhat greater knowledge of thegubernatorial candidates, d = .07.





101. Stevens, n.d. CMAG data from 2000U.S. presidentialelection, distinguishingnumber of negativeads from proportionof negativism in mediamarket. [Reportedeffects in table are forproportion of negativeads, controlling onnumber of positive,negative, and contrastads to whichrespondent wasexposed.]

2000 ANES survey(N = 1,205)

Reported turnout

Memory for ad

Candidate knowledge

Interest in campaign



Satisfaction withdemocracy

Turnout slightly higher with greaterproportion of negative ads, d = .02.Memory slightly higher with greaterproportion of negative ads, d = .01.Averaged across three differentmeasures, candidate knowledgesomewhat higher with greaterproportion of negative ads, d = .07.Reported interest in campaignsignificantly lower with higherproportion of negative ads,d = -.13.Greater proportion on negative adsassociated with significantly lowerpolitical efficacy, d = -.12.Greater proportion of negative adsassociated with somewhat higherexternal efficacy, d = .03.Exposure to more negative adsassociated with somewhat highersatisfaction with democracy,d = .08.

102. Sulfaro 1998 Reported recall ofpositive or negative adfrom 1992 and 1996U.S. presidentialcampaigns

1992 and 1996 ANESsurveys, N = 4,054

Affect for target

Affect for sponsor

Memory for ads

Negative ads increased liking oftarget for both low education(d = -.02) and high educationrespondents (d = -.01); weightedmean d = -.02.Affect toward sponsor of negativead decreased for low education(d = -.03) but not high educationrespondents (d = 0); weighted meand = -.02.Negative ads recalled better thanpositive ads by both low education(d = .39) and high educationrespondents, d = .39.

103. Thorson,Christ, andCaywood 1991

Fictitious support orattack ads created foractual Senatecandidates

161 undergraduatesassigned to 2 (issue vsimage) ¥ 2 (support orattack) ¥ 2 (presenceof music) ¥ 2 (visualbackground)experiment

Affect for sponsorof adVote intention

Memory for ad

Sponsor of ad liked less if attacking,d = -.35.No significant difference in voteintention, (assumed) d = 0.Memory better for support thanattack ad, d = -.35.

104. Thorson et al.2000 (updates1996)

Perceived exposure topositive and negativeads during campaign

Random survey of 657residents of a northerncity after gubernatorialand Senate electioncampaigns

Turnout

Public mood


Trust in government


No significant relationship betweenrelative exposure to positive andnegative ads and reported turnout,(assumed) d = 0.Exposure to negative ads negativelyrelated to several measures of publicmood, average d = -.09.Relatively greater exposure tonegative ads related to somewhatlower internal political efficacy,d = -.06.Negative ads significantly negativelyrelated to trust in government,d = -.16.Negative ads slightly decreasedknowledge of candidates comparedto positive ads, d = -.06.





105. Tinkham andWeaver-Lariscy1991

Media strategy, asreported by actualcongressionalcandidates (positiveissue, positive image,or focus on opponent)

242 respondents insurvey of both majorparty candidates fromall 333 competitivecongressional races in1982

Actual outcome (i.e.,did attacker win orlose election?)

Challengers who went negativemore likely to win, d = .14;incumbents who went negativemore likely to lose, d = -.16;open-seat candidates who wentnegative much more likely to lose,d = -.68; weighted average,d = -.10.

106. Tinkham andWeaver-Lariscy1993

Positive or negativetone of 10 actualpolitical ads

Within-subjectsexperimental designwith 201undergraduates

Differential affect(“Source utility”minus “Targetutility”)

Positive ads produced greaterdifferential affect for sponsor of ad,d = -4.38.

107. Wadsworthet al. 1987

Aggressive (attack) ornonaggressive(positive) ad

Simple comparisonbetween 44undergraduatesassigned to eitherexperimentalcondition

Affect towardsponsor of ad

Sponsor liked slightly more ifattacked, d = .30.

108. Wanta,Lemert, and Lee1999

Respondent’sperception of exposureto negative ads inspecial 1995 Oregonsenate election

Random samplesurvey of 147registered voters inEugene area who hadvoted in past threesenate elections

Trust in government


Small negative correlation betweenexposure to negative ads and trustin government, d = -.12.No correlation between exposure tonegative ads and knowledge of thetwo major party candidates, d = 0.

109. Wattenbergand Brians 1999

Respondent’s recall ofpositive or negativeads from the 1992 and1996 presidentialelections, plusAnsolabehere et al.’s(1994) categorizationof tone of 1992 Senateelection campaigns

Nationallyrepresentative surveyof 3,216 ANESrespondents; andaggregate level turnoutin 34 1992 Senateelections (which,unlike the original,includes absenteeballots in totals)

Actual turnout

Reported turnout

Political efficacy(combined internaland external efficacyitems)

Turnout slightly lower in states withnegative Senate election campaignscompared to states with positiveSenate campaigns, d = -.04.Recall of negative politicaladvertising positively associatedwith voter turnout compared torecall of positive ad, d = .02.Recall of negative political adssignificantly related to higherpolitical efficacy in 1992 but not atall in 1996, mean d = .05.

110. Weaver-Lariscyand Tinkham 1996

Media strategy, asreported by actual1990 congressionalcandidates (positiveissue, positive image,focus on opponent,response toopponent’s attacks)

295 responses tosurvey of both majorparty candidates in all310 competitivecongressional races in1990

Percentage of totalvote received byrespondent

Negative campaigning less effectivethan positive campaigning forchallengers (d = -.06), open-seatcandidates (d = -.18), andsignificantly for incumbents(d = -.56); weighted mean d = -.30.

111. Weigold 1992 Positive or negative adby fictitiousCongressionalcandidate

116 undergraduatesparticipating in2 ¥ 2 ¥ 2 ¥ 2 mixedfactorial experimentaldesign

Affect for attacker

Affect for target

Differential affect(Attacker–Target)

Attacker liked less when usingnegative ad, d = -1.18.Target liked less after negative ad,d = 1.90.Taken together, negative ad moreeffective than positive ad, d = .72.


Acknowledgments

An earlier version of this paper was presented at the101st annual meeting of the annual meeting of theAmerican Political Science Association, Washington.We thank Betsy Becker and Robert Rosenthal foradvice on some tricky issues in the analysis; anyremaining mistakes are certainly our own.

Manuscript submitted 31 August 2006Manuscript accepted for publication 3 November 2006

References

Abbe, Owen, Paul Herrnson, David Magleby, and Kelly Patterson.2000. “Campaign Professionalism, Negative Advertising, andElectoral Success in U.S. House Races.” Presented at the annualmeeting of the Midwest Political Science Association.

Ansolabehere, Stephen, and Shanto Iyengar. 1995. Going Negative:How Political Advertisements Shrink and Polarize the Electorate.New York: Free Press.

Ansolabehere, Stephen, Shanto Iyengar, Adam Simon, and Nicho-las Valentino. 1994. “Does Attack Advertising Demobilize theElectorate?” American Political Science Review 88 (December):829–38.

Ansolabehere, Stephen, Shanto Iyengar, and Adam Simon. 1999.“Replicating Experiments Using Aggregate and Survey Data:The Case of Negative Advertising and Turnout.” AmericanPolitical Science Review 93 (December): 901–909.

Arceneaux, Kevin, and David W. Nickerson. 2005. “Two FieldExperiments Testing Negative Campaign Tactics.” Presented atthe annual meeting of the American Political Science Associa-tion, Washington.

Babbitt, Paul R., and Richard R. Lau. 1994. “The Impact of Nega-tive Political Campaigns on Political Knowledge.” Presented atthe annual meeting of the Southern Political Science Associa-tion, Atlanta.

Bartels, Larry M. 2000. “Campaign Quality: Standards for Evalu-ation, Benchmarks for Reform.” In Campaign Reform: Insightsand Evidence, ed. Larry M. Bartels and Lynn Vavreck. AnnArbor: University of Michigan Press, 1–61.

Basil, Michael, Caroline Schooler, and Byron Reeves. 1991. “Posi-tive and Negative Political Advertising: Effectiveness of Ads andPerceptions of Candidates.” In Television and Political Advertis-ing, Vol. 1, ed. Frank Biocca. Hillsdale, NJ: Lawrence Erlbaum,245–62.

Becker, Betsy Jane. 1994. “Combining Significance Levels.” In TheHandbook of Research Synthesis, ed. Harris Cooper and Larry V.Hedges. New York: Russell Sage Foundation, 215–30.

Brader, Ted. 2005.“Striking a Responsive Chord: How Political AdsMotivate and Persuade Voters by Appealing to Emotions.”American Journal of Political Science 49 (April): 388–405.

Brader, Ted, and Bryce Corrigan. 2006. “How the Emotional Tenorof Ad Campaigns Affects Political Participation.” Presented atthe annual meeting of the American Political Science Associa-tion, Philadelphia.

Bradley, Samuel D., James R. Angelini, and Sungkyoung Lee. 2005.Psychophsysiological and Memory Effects of Negative PoliticalAds: Aversive, Arousing, and Well Remembered. Unpublishedmanuscript. The Ohio State University.

Bratcher, Chris N. 2001.“The Effectiveness of Negative Campaign-ing in U.S. Senate Elections.” Presented at the annual meetingof the Midwest Political Science Association, Chicago.

Brians, Craig L., and Martin P. Wattenberg. 1996. “Campaign IssueKnowledge and Salience: Comparing Reception from TV Com-mercials, TV News, and Newspapers.” American Journal ofPolitical Science 40 (February): 172–93.

Brooks, Deborah J. 2000. When Candidates Attack: The Effects ofNegative Campaigning on Voter Turnout in Senate Elections.Ph.D. dissertation. Yale University.

Brooks, Deborah J. 2006. “The Resilient Voter: Moving TowardClosure in the Debate over Negative Campaigning andTurnout.” Journal of Politics 68 (August): 684–96.

Brooks, Deborah J., and John G. Geer. 2007. “Beyond Negativity:The Effects of Incivility on the Electorate.” American Journal ofPolitical Science 51 (January): 1–16.

Bullock, David A. 1994. The Influence of Political Attack Advertisingon Undecided Voters: An Experimental Study of CampaignMessage Strategy. Unpublished Ph.D. dissertation. University ofArizona.

Capella, Louis, and Ronald D. Taylor. 1992. “An Analysis of theEffectiveness of Negative Political Campaigning.” Business andPublic Affairs 18 (Spring): 10–17.

Chang, Chingching. 2001. “The Impacts of Emotion Elicited byPrint Political Advertising on Candidate Evaluation.” MediaPsychology 3 (May): 91–118.

Chang, Chingching. 2003. “Party Bias in Political-Advertising Pro-cessing.” Journal of Advertising 32 (Summer): 55–67.

Chang, Won Ho, Jae-Jin Park, and Sung Wook Shim. 1998. “Effec-tiveness of Negative Political Advertising.” Unpublished manu-script. Ohio University.

Chanslor, Mike. 1995. The Effects of Televised Political Advertise-ments on Candidate Image. Ph.D. dissertation, University ofOklahoma.

Clinton, Joshua D., and John S. Lapinski. 2004. “ ‘Targeted’ Adver-tising and Voter Turnout: An Experimental Study of the 2000Presidential Election.” Journal of Politics 66 (February): 69–96.

Cooper, Harris, and Larry V. Hedges, eds. 1994. The Handbook ofResearch Synthesis. New York: Russell Sage Foundation.

Craig, Stephen C., and James G. Kane. 2000. “Winning and Losing,Sour Grapes, and Negative Ads: The Impact of Election Cam-paigns on Political Support.” Presented at the annual meetingof the Midwest Political Science Association.

Craig, Stephen C., James G. Kane, and Jason Gainous. 2005.“Issue-Related Learning in a Gubernatorial Campaign: A Panel Study.”Political Communication 22 (October–December): 483–503.

Crigler, Ann, Marion Just, and Todd Belt. 2002. “The Three Facesof Negative Campaigning: The Democratic Implications ofAttack Ads, Cynical News, and Fear Arousing Messages.” Pre-sented at the annual meeting of the American Political ScienceAssociation, Boston.

Dermody, Janine, and Richard Scullion. 2000. “Perceptions ofNegative Political Advertising: Meaningful or Menacing? AnEmpirical Study of the 1997 British General Election Cam-paign.” International Journal of Advertising 19 (2): 201–23.

Djupe, Paul A., and David A. M. Peterson. 2002. “The Timingand Impact of Negative Campaigning: Evidence from the1998 Senatorial Primaries.” Political Research Quarterly 55(December): 845–60.

Finkel, Steven E., and John Geer. 1998. “A Spot Check: CastingDoubt on the Demobilizing Effect of Attack Advertising.”American Journal of Political Science 42 (April): 573–95.

Fisher, R.A. 1932. Statistical Methods for Research Workers, 4th ed.London: Oliver & Boyd.


Freedman, Paul, and Kenneth M. Goldstein. 1999. “MeasuringMedia Exposure and the Effects of Negative Campaign Ads.”American Journal of Political Science 43 (October): 1189–1208.

Freedman, Paul, William Wood, and Dale Lawton. 1999.“Do’s andDon’ts of Negative Ads: What Voters Say.” Campaigns & Elec-tions 20 (October/November): 20–36.

Fridkin, Kim Leslie, and Patrick J. Kenney. 2004. “Do NegativeMessages Work? The Impact of Negativity on Citizens’ Evalu-ations of Candidates.” American Politics Research 32 (Septem-ber): 570–605.

Garand, James C., and Kristy L. Graddy. 2001. “Explaining theMediated Effects of Negative Campaign Advertising on VoterTurnout.” Presented at the annual meeting of the MidwestPolitical Science Association.

Garramone, Gina M., Charles T. Atkin, Bruce E. Pinkleton, andRichard T. Cole. 1990. “Effects of Negative Political Advertisingon the Political Process.” Journal of Broadcasting & ElectronicMedia 34 (Summer): 299–311.

Geer, John G. 2006. In Defense of Negativity: Attack Ads in Presi-dential Campaigns. Chicago: University of Chicago Press.

Geer, John G., and James H. Geer. 2003. “Remembering AttackAds: An Experimental Investigation of Radio.” Political Behav-ior 25 (March): 69–95.

Geer, John G., and Richard R. Lau. 1998.“A New Way to Model theEffects of Campaigns.” Presented at the 1998 annual meeting ofthe American Political Science Association, Boston.

Geer, John G., and Richard R. Lau. 2003. “Filling in the Blanks: ANew Approach for Estimating Campaign Effects.” Unpublishedmanuscript. Vanderbilt University.

Geer, John G., and Richard R. Lau. 2006. “A New Approach forStudying Campaign Effects.” British Journal of Political Science35 (April): 269–90.

Glass, Gene V., Barry McGaw, and Mary Lee Smith. 1981. Meta-Analysis in Social Research. Beverly Hills, CA: Sage.

Globetti, Suzanne, and Marc J. Hetherington. 2000. “The NegativeImplications of Anti-Government Campaign Rhetoric.” Pre-sented at the annual meeting of the Midwest Political ScienceAssociation.

Goldstein, Kenneth M. 1997. “Political Advertising and PoliticalPersuasion in the 1996 Presidential Campaign.” Presented atthe annual meeting of the American Political Science Associa-tion, Washington.

Goldstein, Kenneth M., and Patricia Strach. 2004. The Medium andthe Message, Upper Saddle River, NJ: Pearson.

Greenwald, David. 2005. “Opinion Effects: Timing versus Tone inCampaign Advertising.” Presented at the annual meeting of theMidwest Political Science Association.

Grunwald, Michael. 2006. “The Year of Playing Dirtier: NegativeAds Get Positively Surreal.” Washington Post, October 27.

Haddock, Geoffrey, and Mark P. Zanna. 1997. “Impact of NegativeAdvertising on Evaluations of Political Candidates: The 1993Canadian Federal Election.” Basic and Applied Social Psychology19 (June): 204–23.

Hill, Ronald P. 1989. “An Exploration of Voter Responses to Politi-cal Advertisements.” Journal of Advertising 18 (Winter): 14–22.

Hitchon, Jacqueline C., and Chingching Chang. 1995. “Effects ofGender Schematic Processing on the Reception of PoliticalCommercials for Men and Women Candidates.” Communica-tion Research 22 (August): 430–58.

Hitchon, Jacqueline C., Chingching Chang, and Rhonda Harris.1997. “Should Women Emote? Perceptual Bias and OpinionChange in Response to Political Ads for Candidates of DifferentGenders.” Political Communication 14 (January): 49–69.

Houston, David A., and Kelly Doan. 1999. “Can You Back that Up?Media Psychology 1 (3): 191–206.

Houston, David A., Kelly Doan, and David Roskos-Ewoldsen.1999. “Negative Political Advertising and Choice Conflict.”Journal of Experimental Psychology: Applied 5 (March): 3–16.

Hunter, John E., and Frank L. Schmidt. 1990. Methods of Meta-Analysis: Correcting Error and Bias in Research Findings.Newbury Park, CA: Sage.

Iida, Takeshi. 2005. “Does Negative Campaigning Reduce Uncer-tainty?” Presented at the annual meeting of the Midwest Politi-cal Science Association.

Jackson, Robert A., and Thomas Carsey. 2007. “U.S. Senate Cam-paigns, Negative Advertising, and Voter Mobilization in the1998 Midterm Election.” Electoral Studies 26 (March): 180–95.

Jackson, Robert A., Jeffery J. Mondak, and Robert Huckfeldt. 2005.“Campaign Advertising and Democratic Citizenship: Evaluat-ing the Case against Negative Political Ads.” Presented at theannual meeting of the American Political Science Association,Washington.

Jackson, Robert A., and Sides, Jason. 2006. “Revisiting the Influ-ence of Campaign Tone in Senate Elections.” Political Analysis14 (Spring): 206–18.

Jasperson, Amy E., and David P. Fan. 2002. “An Aggregate Exami-nation of the Backlash Effect in Political Advertising: The Caseof the 1996 U.S. Senate Race in Minnesota.” Journal of Adver-tising 31 (Spring): 1–12.

Kahn, Kim Fridkin and John G. Geer. 1994.“Creating Impressions:An Experimental Investigation of Political Advertising on Tele-vision.” Political Behavior 16 (March): 93–116.

Kahn, Kim Fridkin, and Patrick J. Kenney. 1998a. “Do NegativeCampaigns Mobilize or Suppress Turnout? Clarifying the Rela-tionship between Negativity and Participation.” Unpublishedmanuscript. Arizona State University.

Kahn, Kim Fridkin, and Patrick J. Kenney. 1998b.“Negative Adver-tising and an Informed Electorate: How Negative CampaigningEnhances Knowledge of Senate Elections.” Presented at theConference on Political Advertising in Election Campaigns,Washington.

Kahn, Kim, and Patrick J. Kenney. 2002. “Do Negative MessagesWork? An Examination of the Impact of Negativity in U.S.Senate Campaigns.” Presented at the annual meeting of theAmerican Political Science Association, Boston.

Kahn, Kim, and Patrick J. Kenney. 2004. No Holds Barred: Nega-tivity in U.S. Senate Campaigns. Upper Saddle River, NJ:Pearson Prentice Hall.

Kaid, Lynda Lee. 1997.“Effects of the Television Spots on Images ofDole and Clinton.” American Behavioral Scientist 40 (August):1085–94.

Kaid, Lynda Lee, and John Boydston. 1987. “An ExperimentalStudy of the Effectiveness of Negative Political Advertise-ments.” Communication Quarterly 35 (Spring): 193–201.

Kaid, Lynda Lee, Mike Chanslor, and Mark Hovind. 1992. “TheInfluence of Program and Commercial Type on PoliticalAdvertising Effectiveness.” Journal of Broadcasting & ElectronicMedia 36 (Summer): 303–20.

Kaid, Lynda Lee, Chris M. Leland, and Susan Whitney. 1992. “TheImpact of Televised Political Ads: Evoking Viewer Responses inthe 1988 Presidential Campaign.” Southern Speech Communi-cation Journal 57 (Summer): 285–95.

King, Erika G., and Robert W. Hendersen. 1999.“Effects of Tone ofCampaign Ads and Viewers’ Initial Attitudes on Ad Efficacy inthe 1998 Michigan Gubernatorial Campaign.” Presented at theannual meeting of the Midwest Political Science Association.


King, Erika G., Robert W. Hendersen, and Hong C. Chen. 1998.“Viewer Response to Positive vs. Negative Ads in the 1996 Presi-dential Campaign.” Presented at the annual meeting of theMidwest Political Science Association.

King, James D., and Jason B. McConnell. 2003. “The Effect ofNegative Campaign Advertising on Vote Choice: The Mediat-ing Influence of Gender.” Social Science Quarterly 84 (Decem-ber): 843–57.

Landi, Massimiliano. 2004. Negative Campaigning and ElectoralCompetition. Masters thesis. University of Pennsylvania.

Lang, Annie. 1991. “Emotion, Formal Features, and Memory forTelevised Political Advertisements.” In Television and PoliticalAdvertising, Vol. 1, ed. Frank Biocca. Hillsdale, NJ: LawrenceErlbaum, 221–43.

Lau, Richard R., and Gerald Pomper. 2004. Negative Campaigning:An Analysis of U.S. Senate Elections. Lanham, MD: Rowmanand Littlefield.

Lau, Richard R., Gerald Pomper, and Grace A. Mumoli. 1998.“Effects of Negative Campaigning on Senate Election Out-comes: 1988, 1990, 1994, & 1996.” Presented at the annualmeeting of the Midwest Political Science Association.

Lau, Richard R., and David P. Redlawsk. 2005. “Effects of Positiveand Negative Political Advertisements on Information Process-ing.” Unpublished manuscript, Rutgers University.

Lau, Richard R., Lee Sigelman, Caroline Heldman, and PaulBabbitt. 1999. “The Effects of Negative Political Advertise-ments: A Meta-Analytic Review.” American Political ScienceReview 93 (December): 851–75.

Lawton, L. Dale, and Paul Freedman. 2001. “Beyond Negativity:Advertising Effects in the 2000 Virginia Senate Race.” Presentedat the annual meeting of the Midwest Political Science Asso-ciation.

Lemert, James B., William R. Elliot, James M. Bernstein, William L.Rosenberg, and Karl J. Nestvold. 1991. News Verdicts, theDebates, and Presidential Campaigns. New York: Praeger.

Lemert, J., W. Wanta, and T-T Lee. 1999. “Party Identification andNegative Advertising in a U.S. Senate Election.” Journal of Com-munication 49 (2): 123–36.

Leshner, Glenn, and Esther Thorson. 2000.“Overreporting Voting:Campaign Media Public Mood, and the Vote.” Political Com-munication 17 (July–September): 263–78.

Luskin, Robert C., and Christopher Bratcher. 1994. “NegativeCampaigning, Partisanship, and Turnout.” Presented at theannual meeting of the Midwest Political Science Association.

Mansnerus, Laura. 2005. “Forrester Tosses His ‘Hail Mary’ Ad.”New York Times (23 October): Section 14NJ; Column 5; NewJersey Weekly Desk; ON POLITICS; Pg. 2.

Martin, Paul S. 1999. “The Underwhelmed Citizen: A Theory ofMedia Influence on Conditional Political Participation.” Pre-sented at the annual meeting of the American Political ScienceAssociation, Atlanta.

Martin, Paul S. 2004. “Inside the Black Box of Negative CampaignEffects: Three Reasons Why Negative Campaigns Mobilize.”Political Psychology 25 (August): 545–62.

Martinez, Michael D. and Tad Delegal. 1990. “The Irrelevance ofNegative Campaigns to Political Trust: Experimental andSurvey Results.” Political Communication and Persuasion 7(January/March): 25–40.

Mathews, Douglas, and Beth Dietz-Uhler. 1998. “The Black-SheepEffect: How Positive and Negative Advertisements AffectVoters’ Perceptions of the Sponsor of the Advertisement.”Journal of Applied Social Psychology 28 (October 16): 1903–15.

May, Patrick. 2006.“Ads Reach New Lows.” San Jose Mercury News,June 2.

McBride, Allan, Robert Toburen, and Dan Thomas. 1993. “DoesNegative Campaign Advertising Depress Voter Turnout? Evi-dence from Two Election Campaigns.” Unpublished manu-script. Grambling State University.

Meirick, Patrick C., and Michael W. Pfau. 2005. “Tone, Target andIssue Ownership in Presidential Primary Advertising Effects.”Presented at the annual meeting of the Political Communica-tion Division of the International Communication Associa-tion, New York.

Merritt, Sharyne. 1984. “Negative Political Advertising: SomeEmpirical Findings.” Journal of Advertising 13 (Fall): 27–38.

Min, Young. 2004. “News Coverage of Negative Political Cam-paigns: An Experiment of Negative Campaign Effects onTurnout and Candidate Preference.” Harvard InternationalJournal of Press/Politics 9 (4): 95–111.

Mosteller, Fred, and R. R. Bush. 1954. “Selected Quantitative Tech-niques.” In Handbook of Social Psychology, Vol. 1, ed. GardnerLindzey. Cambridge, MA: Addison-Wesley, 289–334.

Nagourney, Adam. 2006. “New Campaign Ads Have a Theme:Don’t Be Nice.” New York Times, September 27.

Naman, Marcelle Rathle. 2000. The Effects of Negative PoliticalDirect Comparative Television Advertisements on Voter Evalua-tions and Voting Decisions of Viewers. Masters thesis. Universityof South Alabama.

Newhagen, John E., and Byron Reeves. 1991. “Emotion andMemory Responses for Negative Political Advertising: A Studyof Television Commercials Used in the 1988 Presidential Elec-tion.” In Television and Political Advertising, Vol. 1, ed. FrankBiocca. Hillsdale, NJ: Lawrence Erlbaum, 197–220.

Niven, David. 2006.“A Field Experiment on the Effects of NegativeCampaign Mail on Voter Turnout in a Municipal Election.”Political Research Quarterly, 59 (June): 203–210.

Niven, David. 2005. “Issue-Related Learning in a GubernatorialCampaign: A Panel Study.” Unpublished manuscript. FloridaAtlantic University.

Patterson, Kelly D., and Dan Shea. 2001.“Local Norms and NeutralObservers: Rethinking the Effects of Negative Campaigning.”Presented at the annual meeting of the American PoliticalScience Association, San Francisco.

Pfau, Michael, and Michael Burgoon. 1989. “The Efficacy of Issueand Character Attack Message Strategies in Political CampaignCommunication.” Communication Reports (Summer): 53–61.

Pfau, Michael, Henry C. Kenski, Michael Nitz, and John Sorenson.1989. “Use of the Attack Message Strategy in Political Cam-paign Communication.” Presented at the annual meeting of theSpeech Communication Association, San Francisco.

Pinkleton, Bruce. 1997. “The Effects of Negative ComparativePolitical Advertising on Candidate Evaluations and AdvertisingEvaluations: An Exploration.” Journal of Advertising 26(Spring): 19–29.

Pinkleton, Bruce E. 1998. “Effects of Print Comparative PoliticalAdvertising on Political Decision-Making and Participation.”Journal of Communications 48 (Autumn): 24–36.

Pinkleton, Bruce E., and Gina M. Garramone. 1992. “A Survey ofResponses to Negative Political Advertising: Voter Cognition,Affect, and Behavior.” Proceedings of the 1992 Conference of theAmerican Academy of Advertising: 127–33.

Pinkleton, Bruce E., Nam-hyun Um, and Erica Weintraub Austin.2002. “An Exploration of the Effects of Negative PoliticalAdvertising on Political Decision Making.” Journal of Advertis-ing 31 (Spring): 13–26.

Quigley, Winthrop. 2006. “Why Mud Works in Political but NotProduct Ads.” Albuquerque Journal, August 3, 2006, BusinessOutlook, p. 3.


Rahn, Wendy M., and Rebecca Hirshorn. 1995. “Political Advertis-ing and Public Mood: An Experimental Study of Children’sPolitical Orientations.” Presented at the annual meeting of theAmerican Political Science Association.

Rahn, Wendy M., and Rebecca M. Hirshorn. 1999. “PoliticalAdvertising and Public Mood: A Study of Children’s PoliticalOrientations.” Political Communication 16 (October–December): 387–407.

Roberts, Marilyn S. 1995. “Political Advertising: Strategies forInfluence.” In Presidential Campaign Discourse: Strategic Com-munication Problems, ed. Kathleen E. Kendall. Albany: SUNYPress, 179–99.

Roddy, Brian L., and Gina M. Garramone. 1988. “Appeals andStrategies of Negative Political Advertising.” Journal of Broad-casting & Electronic Media 32 (Fall): 415–27.

Roese, Neal J., and Gerald N. Sande. 1993. “Backlash Effects inAttack Politics.” Journal of Applied Social Psychology 23(August): 632–53.

Rosenthal, Robert. 1979. “The ‘File Drawer’ Problem and Toler-ance for Null Results.” Psychological Bulletin 86 (May): 638–41.

Rosenthal, Robert. 1984. Meta-analytic Procedures for SocialResearch. Beverly Hills, CA: Sage.

Rosenthal, Robert, and Donald B. Rubin. 2003. “requivalent: A SimpleEffect Size Indicator.” Psychological Methods 8 (4): 492–96.

Sanders, David, and Pippa Norris. 2002. “Advocacy versus Attack:The Impact of Political Advertising in the 2001 UK GeneralElection.” Presented at the ECPR Joint Sessions, Turin.

Schultz, Cindy, and S. Mark Pancer. 1997. “Character Attacks andTheir Effects on Perceptions of Male and Female Political Can-didates.” Political Psychology 18 (March): 93–102.

Shapiro, Michael A., and Robert H. Rieger. 1992. “ComparingPositive and Negative Political Advertising on Radio.” Journal-ism Quarterly 69 (Spring): 135–45.

Shea, Christopher. 2006. “Go Negative! Are Attack Ads Good forDemocracy?” Boston Globe, May 21.

Shen, Fuyuan, and H. Denis Wu. 2002. “Effects of Soft–MoneyIssue Advertisements on Candidate Evaluation and VotingPreference: An Exploration.” Mass Communication & Society 5(4): 395–410.

Sides, John. 2006. “Swing Your Partner: The Importance of Inter-actions in Campaigns.” Presented at the annual meeting of theAmerican Political Science Association, Washington.

Sides, John, Matthew Grossmann, Christine Trost, and KeenaLipsitz. 2003. “Candidate Attacks and Voter Aversion: TheUncertain Link between Negativity and Campaign Satisfac-tion.” Presented at the annual meeting of the American PoliticalScience Association, Philadelphia.

Sides, John, Keena Lipsitz, Matthew Grossmann, and ChristineTrost. 2005.“Candidate Attacks and Voter Aversion: The Uncer-tain Effects of Negative Campaigning.” Unpublished manu-script. George Washington University.

Sonner, Brenda S. 1998. “The Effectiveness of Negative PoliticalAdvertising: A Case Study.” Journal of Advertising Research 38(December): 37–42.

Stanley, T. D., and Stephen B. Jarrell. 1989. “Meta-RegressionAnalysis: A Quantitative Method of Literature Surveys.” Journalof Economic Surveys 3 (April): 161–70.

Stevens, Daniel. 2002. “Explaining Heterogeneity in the Effects ofNegative Advertising.” Presented at the annual meeting of theMidwest Political Science Association.

Stevens, Daniel. 2005. “The Two Routes for Effects of NegativeAdvertising.” Presented at the annual meeting of the MidwestPolitical Science Association.

Stevens, Daniel. No Date. “More than Volume: Proportion,Volume, and the Effects of Exposure to Negative Advertising.”Unpublished manuscript. Hartwick College.

Sulfaro, Valerie A. 1998.“Political Sophistication and the Presiden-tial Campaign: Citizen Reactions to Campaign Advertise-ments.” Presented at the annual meeting of the MidwestPolitical Science Association.

Thorson, Esther, William G. Christ, and Clarke Caywood. 1991.“Selling Candidates Like Tubes of Toothpaste: Is the Compari-son Apt?” In Television and Political Advertising, Vol. 1, ed.Frank Biocca. Hillsdale, NJ: Lawrence Erlbaum, 145–72.

Thorson, Esther, Ekaterina Ognianova, James Coyle, and FrankDenton. 2000. “Negative Political Ads and Negative CitizenOrientations toward Politics.” Journal of Current Issues &Research in Advertising 22 (Spring): 13–41.

Tinkham, Spencer F., and Ruth Ann Weaver-Lariscy. 1991. “Adver-tising Message Strategy in U.S. Congressional Campaigns: ItsImpact on Election Outcome.” Current Issues and Research inAdvertising 13 (Spring–Summer): 207–26.

Tinkham, Spencer F., and Ruth Ann Weaver-Lariscy. 1993. “ADiagnostic Approach to Assessing the Impact of NegativePolitical Television Commercials.” Journal of Broadcasting &Electronic Media 37 (Fall): 377–400.

Tucker, Brian. 2006. “It’s Election Season, for Better or Worse.”Crain’s Cleveland Business, October 23, p. 10.

Wadsworth, Anne Johnston, Philip Patterson, Lynda Lee Kaid,Ginger Cullers, Drew Malcomb, and Linda Lamirand. 1987.“ ‘Masculine’ vs. ‘Feminine’ Strategies in Political Ads: Implica-tions for Female Candidates.” Journal of Applied Communica-tion 15 (Spring and Fall): 77–94.

Wanta, Wane, James B. Lemert, and Tien-tsung Lee. 1999. “Con-sequences of Negative Political Advertising Exposure.” InEngaging the Public: How Government and the Media Can Rein-vigorate American Democracy, eds. Thomas J. Johnson, Carol ElHays, and Scott P. Hays. Boulder, CO: Rowman & Littlefield,97–109.

Wattenberg, Martin P., and Craig L. Brians. 1999. “Negative Cam-paign Advertising: Demobilizer or Mobilizer?” American Politi-cal Science Review 93 (December): 891–99.

Weaver-Lariscy, Ruth Ann, and Spencer F. Tinkham. 1996. “Adver-tising Message Strategies in U.S. Congressional Campaigns:1982, 1990.” Journal of Current Issues and Research in Advertis-ing 18 (Spring): 53–66.

Weigold, Michael F. 1992. “Negative Political Advertising: Indi-vidual Differences in Responses to Issue vs. Image Ads.” Pro-ceedings of the 1992 Conference of the American Academy ofAdvertising.

Whelan, Jeff, and Josh Margolin. 2005. “For Underdog, ThoseBiting Ads Proved No Help.” Star-Ledger (9 November): 1.

Richard R. Lau is professor of political science,Rutgers University, New Brunswick, NJ 08901. LeeSigelman is professor of political science, The GeorgeWashington University, Washington, D.C., 20052. IvyBrown Rovner is a graduate student in politicalscience, Rutgers University, New Brunswick, NJ 08901.


theeffectsof negativepoliticalcampaigns: ameta...

Documents